Development of a Fully Monolithic Microfluidic Device for Complete Blood Count
by
John Nguyen
A thesis submitted in conformity with the requirements for the degree of Master of Applied Science (M.A.Sc.) Graduate Department of Mechanical and Industrial Engineering University of Toronto
c Copyright 2014 by John Nguyen Abstract
Development of a Fully Monolithic Microfluidic Device for Complete Blood Count
John Nguyen
Master of Applied Science (M.A.Sc.)
Graduate Department of Mechanical and Industrial Engineering
University of Toronto
2014
This thesis describes a monolithic microfluidic device capable of complete blood constituent enumeration from whole blood. For the first time, on-chip sample processing (e.g. dilution, lysis, and filtration) and downstream single cell analysis were fully integrated on device to enable complete blood cell count. The microfluidic device consists of two parallel sub-systems that perform sample processing and electrical analysis for simultaneous measurement of red
(RBC) and white blood cell (WBC) parameters. The system provides a modular and adaptable environment capable of handling solutions of various viscosities and mixing ratios and features a new offset filter configuration for increased experimental duration. RBC concentration, mean corpuscular volume (MCV), cell distribution width, WBC concentration and differential are determined by electrical impedance measurements. Experimental characterization results of
97,305 WBCs and 104,735 RBCs from 10 patient blood samples demonstrate that the system is capable of performing high volume enumeration and complete blood count with accuracy.
ii Acknowledgements
I would like to thank Professor Sun for his guidance, perspective on life and passion for research.
I would like to thank all the members of the Advanced Micro and Nanosystems Laboratory
(AMNL), both past and present, for their support and company. I would also like to give special acknowledgments to all the members of the microfluidic group: Mark, Yi and Yuan for all of their experimental help, advice and long discussions.
I would like to acknowledge the financial support I have received from the NSERC Create
Program in Microfluidic Applications and Training in Cardiovascular Health (MATCH) and the Ontario Graduate Scholarship (OGS).
I would like to thank my parents for everything they’ve provided me through the years.
Lastly, thank you to my partner and best friend Jenny, for her never ending support, encour- agement and love.
iii Contents
List of Figures vi
List of Tables x
1 Introduction 1
1.1 Background...... 1
1.2 Blood separation on microfluidics...... 4
1.3 Microfluidic coulter counter...... 7
1.4 Fluidic Sample Processing for Blood...... 8
1.4.1 Dilution...... 9
1.4.2 Lysis...... 10
1.5 Motivation...... 11
1.6 Thesis outline...... 13
2 System Overview 14
2.1 System Flow...... 15
2.2 Device Design...... 19
2.3 Filter Design and Implementation...... 22
3 Fabrication and Experimental Methods 28
3.1 Fabrication...... 28
3.2 Experimental Procedure...... 30
3.3 Signal Analysis...... 30
iv 4 Experimental Results and Discussion 35
4.1 Device Characterization...... 35
4.1.1 Pressure Independence...... 35
4.1.2 Flowrate Control...... 37
4.2 Red Blood Cell Analysis...... 38
4.2.1 Red Blood Cell Enumeration...... 38
4.2.2 Red Blood Cell Characterization...... 39
4.3 White Blood Cell Analysis...... 44
5 Conclusion 52
6 Future Directions 53
Bibliography 55
A Circuit Simulation 61
B Previous Revisions of Device Design 63
v List of Figures
1.1 Relative concentrations of blood constituents in a single µL of blood. Red blood
cells - RBCs, white blood cells - WBCs and platelets are highlighted...... 4
1.2 Various techniques to isolate/separate blood cells...... 5
1.3 System schematics from various microfluidic devices utilizing Coulter counter
technology...... 9
1.4 Overview of existing microfluidic technologies with on-chip sample processing
and integration with measurement units...... 11
2.1 System Flow Chart. Whole blood is siphoned into two streams for RBC and
WBC measurement. WBC analysis requires lysis and quench to eliminate trou-
blesome RBCs before measurement...... 16
2.2 Device Overview. 3D and 2D schematic of different fluidic modules required for
RBC and WBC analysis...... 17
2.3 Raw RBC and WBC Impedance Peaks. Two 500 ms segments raw data show-
ing detection threshold (green), enumerated cells circled red and highlighting
effective electrical volume difference between WBCs (left) and RBCs (right)... 18
2.4 Simulation Results: Expected Impedance Change. Simulation results highlight-
ing the impedance change of spherical cells of different diameters for various
channel dimensions...... 19
2.5 Fluidically Analogous Circuit Model. Model used to determine channel geome-
tries and configuration to ensure precise fluidic mixing ratios. Subscripts denote
fluidic resistors in various modules...... 20
vi 2.6 Types of Debris.Images of various types of debris which can occlude constriction
measurement channels and limit experimental throughput...... 23
2.7 Summary of simulation results comparing unwanted fluidic resistance change as
increasing percentage of orifices becomes clogged for existing and our new offset
filter configurations...... 24
2.8 Reconfigured Plug-In Electrodes. Schematic highlighting reconfiguration of con-
ventional plug-in electrodes. Use of the new electrode placement as a reservoir
for debris. Fluid flow can intermittently be adjusted to clear obstructions and
increase device throughput...... 26
2.9 Electrode Configurations. Schematic showing various electrode configurations
used in previous microfluidics...... 27
3.1 Schematic of multi-layered fabrication sequence. Two SU-8 feature layers are
used to generate channel heights of 15 µm and 40 µm...... 29
3.2 Images of device operation and device schematic highlighting key modules for
RBC (right) and WBC (left) measurements. Device dimensions, RBC lysis,
filtration and measurement are highlighted...... 31
3.3 Limits of detection from RBC Analysis. Typical non-gated RBC histogram high-
lighting limits of detection for low electrical volume cells. Platelets or particles
sized lower than the limit of detection are ignored from analysis due to difficulties
in distinguishing them from baseline noise...... 32
3.4 Image of mixing channel for incubation post-lysis and quench.RBC ghosts and
cellular components are visible and can contribute electrical interference. WBCs
are circled...... 33
3.5 Baseline Correction Histogram. RBC size histograms for raw and post base-
line correction of the windowed sample are presented.Corrected size histogram
improves size accuracy and produces expected RBC distribution...... 34
vii 4.1 Pressure Independence: Mixing Ratios. Relative ratios of the fluidic widths
( W1 ) were directly correlated to their relative volumetric flow rate. The use of WT an inlet-outlet configuration for serial dilution is used to enable the device to
behave consistently independent of the applied pressure...... 36
4.2 Flow Metering. Effects of varying flow rates on basal impedance are observed.
Linear trend indicates the possibility of using observed impedance change to
approximate volumetric flow rate...... 38
4.3 Effect of dilution ratio on RBC enumeration accuracy. Both maximum values and
distribution shape of the size histogram are evidence of incorrect enumeration of
clustered cells at lower dilution ratios. Concentration results for varying dilution
ratios, compared with reference concentration measured by hematology analyzer. 40
4.4 RBC Deformability and its effect on MCV. Size histograms change in terms of
µ and distribution for various pressures due to increasing RBC deformation.... 41
4.5 RBC Characterization Results: Size histograms highlighting specific measured
RBC indices: distribution width (RDW-SD) and mean corpuscular volume (MCV).
Correlation plot for reference hematology analyzer vs. microfluidically measured
MCV...... 43
4.6 Concordance and Bland-Altman Plots for RBC Enumeration...... 44
4.7 WBC Subtype Size Difference. Optical confirmation of size difference of different
WBC subtypes (e.g. lymphocytes, neutrophils and monocytes)...... 45
4.8 Sample Multi-Frequency Measurements for WBC Analysis. 10 kHz to 990 kHz
signals were used for WBC analysis. Lower frequency provides purely size infor-
mation, while higher frequencies give insight on intracellular properties...... 46
4.9 Sample Scatter Graphs from WBC Analysis. Shown are opacity and electri-
cal volume measurements from healthy patient samples of various lymphocyte
populations (8.3%-31.8%)...... 47
4.10 Concordance and Bland-Altman Plots for WBC Analysis...... 48
A.1 Spice Simulation used to confirm circuit analysis techniques. WBC and RBC
modules are highlighted...... 62
viii B.1 Photolithographic masks for various iterations of microfluidic device used for
complete blood count...... 64
B.2 Photolithographic masks for various filter configuration designs...... 65
B.3 Device operation of previous version of microfluidic device...... 66
ix List of Tables
1.1 Anomalies of blood count as indicator for different diseases...... 2
3.1 Summary of Fabrication Parameters...... 28
3.2 Summary of Fabrication Sequence...... 29
3.3 Summary of Impedance Analyzer settings...... 31
4.1 RBC concentration: Bland Altman parameters for concordance of measurements
from standard analyzer and microfluidic device...... 39
4.2 WBC Parameters; Bland Altman parameters for concordance of measurements
from standard analyzer and microfluidic device...... 49
4.3 Comparative study for Bland Altman parameters between existing microfluidic
enumeration techniques...... 50
x Chapter 1
Introduction
1.1 Background
Blood is among the most important biological fluids and appropriately one of the most studied.
It is a complex heterogeneous biological fluid consisting of mostly water but is a carrier of proteins, glucose, mineral ions, hormones and even dissolved gases such as carbon dioxide.
It is a necessity in the transport of nutrients and oxygen to and metabolic waste away from tissues and organs and in maintenance of homeostasis, regulating physiological temperature and pH. It provides a carrier medium for the vascular network to transport immune cells for defense against infection by foreign bodies. Whole blood is a mixture of plasma and blood cells. Plasma is the liquid component of blood and is a blend of water, sugar, fat, protein and salts. More importantly, blood cells can be separated into three main categories based on bodily function: erythrocytes or red blood cells (RBCs), leukocytes or white blood cells (WBCs) and thrombocytes or platelets.
One of the largest constituents of whole blood is the erythrocytes, or red blood cells. Roughly
45% of blood and over 99% of all blood cells are made up of RBCs [1]. RBCs are primarily responsible for delivering oxygen to all other cells in the body. As such, RBCs are an important indicator of health, and can be used to diagnose many diseases such as anemia, malaria, and sickle-cell disease. RBCs are without a nucleus and are thus easily deformable. On the other hand, leukocytes, or white blood cells, constitute roughly 0.1% of all blood cells, existing in an approximate 1:1000 ratio with RBCs. Besides concentration, WBCs and RBCs differ in shape
1 Chapter 1. Introduction 2
Increased Parameter Potential Conditions
Red blood cells Tolycythemia Hematocrit EPO doping Neutrophils Bacterial infection Lymphocytes Lymphocytic leukemia Monocytes Bacterial infection Eosinophil Parasitic infection
Table 1.1: Anomalies of blood count as indicator for different diseases and size. WBCs are often irregular in shape but for theoretical purposes are assumed to be spherical in shape. They have a nucleus and an outer buffer coat resulting in greater stiffness and less deformability in comparison to RBCs. While there is overlap in diameter ranges of leukocytes and erythrocytes, the majority of WBCs are larger in size making bulk separation of WBCs due to size possible.
The most commonly conducted medical blood tests are blood cell enumeration (red blood cells RBCs; white blood cells WBCs and platelets) and WBC differential. They are rou- tine parts of preliminary diagnostic procedures for sick individuals and as screening tools for those who are asymptomatic[1]. Measuring complete blood count (CBC) can provide insight on primary conditions of the blood and bone marrow including disorders such as anemia [2,3], leukemia [3,4] and polycythemia [3,5]. CBC also enables evaluation of medical conditions that secondarily affect the blood and the immune system, resulting in hematologic expressions such as infection or inflammation[3]. Elevated counts in certain blood constituents can be indicative of other various diseases highlighting the importance of blood analysis [3] (Table 1.1). Currently, conducting CBC relies on costly and complex hematology analyzers located in centralized fa- cilities. The need for portable, rapid, accurate and quantitative evaluation of RBCs and WBCs necessitates the development of devices for point-of-care blood cell measurements.
Hemocytometers were originally designed to count blood cells but require manually iden- tifying cells within a controlled cavity with a known volume [6]. While manual enumeration is useful for counting abnormal cells, manual means are often subject to sampling errors be- cause of low cell counts compared to automated methods. The process can also be subject Chapter 1. Introduction 3 to operator variability and is immensely time consuming. Thus, the current gold standard for
CBC is the use of automated hematology analyzers. The working principles of these analyzers vary, but typically involve the use of the Coulter principle to provide sizing information and optical detection with light scattering for more specific cell enumeration [7,8]. However, both manual and automated complete blood count techniques rely on centralized laboratories, bulky and costly equipment and skilled operators. Microfluidics can address these issues by reducing the complexity, time and costs of tests while enabling, for example, real-time monitoring and providing immediate access to critical information for remote patients or in emergency rooms
[9]. A rapid and reliable blood count for emergency room patients upon admission can provide considerable information compared to the conventional practices where complete blood work is requested much later in the stay.
There are numerous difficulties in identifying these micron sized particles particularly due to the ratio of RBCs to the remaining cellular components. RBCs constitute >99% of all cellular components. Fig. 1.1 highlights the discrepancy in relative cell densities. Although, the majority of each blood cell group varies in properties sufficient enough to allow for moderate identification, provided that the relative concentrations of cell groups can be adjusted via sample handling processes such as separation or dilution.
The application of microfluidics for blood analysis is enticing due to many inherent advan- tages. Small working volumes allows microfluidics optimal control of the microenvironment for pH, temperature, dissolved solutes and flow. Bodily fluids are precious and essential to bodily function hence a low sample volume required for processing blood is highly beneficial.
Microfluidic based blood separation techniques typically rely on differences in both biophysical and electrical parameters such as volume, deformability, conductivity and permittivity. These techniques have shown good separation and isolation of red and white blood cells and can be well suited for implementation in analysis platforms. Additionally, dilution or targeted chemical processes such as lysis can be used to remove unwanted cells. This can provide the cell group isolation necessary for enumeration of specific populations. For enumeration and analysis, the
Coulter counter is a well established technology and is trusted in the medical community. Chapter 1. Introduction 4
Figure 1.1: Relative concentrations of blood constituents in a single µL of blood. Red blood cells - RBCs, white blood cells - WBCs and platelets are highlighted.
1.2 Blood separation on microfluidics
Inherent differences in cell size, deformability, affinity, magnetic and electrical properties can be used to provide mechanisms for both passive and active separation. Passive separation techniques rely on size and deformability differences and utilize fluid structures to either invoke certain forces or retard certain cell migration, while active methods rely on external force fields.
Due to the differences in relative cell density, blood cell separation techniques typically involve the isolation of WBCs from the rest of blood constituents.
The inherent size and deformability difference between leukocytes and erythrocytes make
filter based separation a logical mechanism for separation tasks. Microstructures of various shapes with specific gap sizes allow for ranges of cell diameters and deformability to either pass through or be captured. The very small size range of blood cells often limit filters to be silicon based and MEMS fabrication techniques [14, 15]. Microfilters can still serve as effective means to increase leukocyte concentration by passively removing cellular components less than a critical diameter, however silicon fabrication can be time consuming and costly. Elastomeric Chapter 1. Introduction 5
A
B D C
Figure 1.2: (A) Bump array. Cells above a critical radii will be displaced laterally, separating them from the main stream flow [10] (B) Working principle of spiral channel with trapezoid cross section leukocyte separation. [11]. (C) 3D and 2D scheme of particle sorter based on PDMS porous membrane[12]. (D) Principle of separation for T-channel bifurcation [13]. device can serve as an alternative to reduce fabrication efforts [12, 16]. Wei et al. reported a
PDMS based porous membrane which addresses difficulties of achieving extremely small pore sizes (2.5-3.3 um) in an elastomeric based microfluidic device (Fig. 1.2(C)) [12]. However, high
fluidic shear required for high throughput causes deformation of the substrate making gap size difficult to control thus reducing separation efficiency. A universal problem of dead-end filters is the accumulation of the trapped cells at the pore openings severally affecting subsequent separation processes. Cross filtration which involves filter features to be parallel to the feed stream as opposed to being perpendicular can alleviate clogging issues. Chen et al. have reported silicon based cross filtration device which allows larger particles to stay suspended in the feed stream instead of being deposited after filtration [17].
Devices which use of hydrodynamic effects which are dependent on size and deformability Chapter 1. Introduction 6 can be a substitute for filter based designs by having much larger channels sizes and thus completely removing clogging potential. Wu et al. have reported a spiral channel device which is capable of separating and recovering leukocytes using the interplay between inertial life and viscous drag (Fig. 1.2(B))[11]. The leveraging of biological phenomena (e.g. the bifurcation law and the Fahreus effect) have led to microfluidics that produce comparable or superior throughput to that of filter based devices but without clogging and degrading efficiency [13,18,19]. Most recently Cupelli et al. have utilized hydrodynamic effects to separate leukocytes [20]. Cells with radii above a critical radius will continue to flow in the feed channel whereas smaller cells are separated via the separation channel, Fig 1.2(D). Two phenomena contribute to this effect, otherwise known as the Zweifach−Fung effect: difference in pressure between channels of varying flow rates, and asymmetric shear experienced by particles thus producing a torque and pulling it towards faster flow rates. However, such techniques are highly flow rate dependent. Under high flow rates, cell deformation due to shear would change their effective radius and consequently limit separation efficiency.
Another popular mechanism for separation of cells in solution according to size is known as deterministic lateral displacement (DLD)(Fig. 1.2(A)). Inglis et al. have demonstrated the use of DLD arrays to separate and enrich WBCs[10, 21, 22]. These operate similarly to channels utilizing hydrodynamic effects for separating; namely using a critical radii. As particles flow through an array of staggered posts, particles below a critical size will pass through unaffected or otherwise would moved laterally. These devices feature a large gap size which is less prone to clogging. Sequential filtration with varying gap sizes allows for increased filtration efficiency in later stages by having a smaller gap size but to avoid clogging by pre-filtering the entering
fluid.
While mechanical methods of separation and isolation of cell groups are well developed, the clinical applicability and reliability of chemical processing is unparalleled. Chemical processes such as dilution or lysis can effectively isolate cell groups such as WBCs with high reliability.
The use of chemical processes in an integrated blood analyzer can eliminate additional and unnecessary complexities in isolating cell groups before enumeration. Chapter 1. Introduction 7
1.3 Microfluidic coulter counter
Inherent volume differences between cells have been leveraged by Coulter counters. The Coulter counter measures the impedance of a small orifice/channel as particles of comparable size pass.
At DC frequencies, the insulating layer that is the cell membrane and the presence of the cell alters the impedance of the orifice by replacing much of the conductive liquid. Well established models are available for correlating impedance changes to cell volumes. Several designs have re- cently been proposed to improve the microfluidic Coulter counters’ performance. For instance, two-dimensional sheath flow focusing was demonstrated to increase measurement repeatabil- ity [23]; and throughput of microfluidic Coulter counters was improved using multiple-orifice designs [24].
A significant challenge for microfluidic Coulter counters is the choice of electrode material.
An ideal choice are Ag/AgCl non-polarizable electrodes, which work well in the conventional
Coulter counter. However, effectively integrating Ag/ AgCl electrodes into microfluidic chan- nels is complex, and Ag/ AgCl electrodes have a limited lifetime [25, 26]. Modular Ag/AgCl electrodes can be implemented with the use of microchannels acting as salt bridges to provide electrical connection [27–29] (Fig.1.3(B)). Although other metal electrodes can be more readily integrated into microfluidic channels using existing MEMS fabrication techniques, additionally difficulties can arise. The mostly capacitive electrical double layer created at interface of elec- trode and liquid severely affects DC signals; limiting the use of low frequencies. Methods for minimizing the electrical double layer effect include the modification of the electrode surface roughness in order to increase the surface area [13] and the utilization of polyelectrolytic salt bridges (PSBEs)[30,31] or polyelectrolyte gel electrodes (PGEs) [32](Fig.1.3(D)). Most recently, a DC impedance-based microcytometer device integrating PGEs was reported for CTC cell de- tection. CTCs were successfully detected in blood samples from breast cancer patients [33]. In addition to the electrode design, attention must be paid to volume correlation from electrical measurements due varying different electrode configuration and channel geometry [34, 35].
Although the Coulter counter is a widely used technique in clinical instruments (e.g., hema- tology analyzers), it is limited to classifying cells of distinct volume differences. Some of the most recent commercial hematology analyzers adopted have both DC and RF measurements. Chapter 1. Introduction 8
The ratio of the RF signal to the DC signal is defined as opacity, which is used as a volume independent electrical measurement of cells. Along with the volume characterization, WBC differential count can be achieved. This technique was recently demonstrated on microfluidic devices. In addition to WBC count, microfluidic flow cytometry has also been used for analyzing a variety of particles [36–40]. An impedance flow cytometry design using coplanar electrodes was demonstrated by Gawad et al [36]. Three microfabricated electrodes were integrated on the bottom of the microfluidic channel. Due to the non-uniform electrical field distribution in the channel, any positional variation of tested cells undesirably affected impedance measurement.
To overcome this issue, parallel facing electrodes were developed [37](Fig.1.3(A)). A single pair of electrodes are used for measurements while the other used as reference allowing for differen- tial measurement of electrical signals.The device was demonstrated to be capable of detecting electrical property differences between normal RBCs and glutaraldehyde-fixed RBCs.
Holmes et al. incorporated a fluorescence measurement unit on their microfluidic flow cytometry device[39](Fig.1.3(C)). This device design enabled direct correlation of impedance signals from individual cells with their biochemical phenotypes. A 3-part WBC differential count of human blood was well achieved. Besides counting blood cells, microfluidic impedance
flow cytometry has also found applications in other areas, such as the measurement of cell viability and physiological differences of cells.
1.4 Fluidic Sample Processing for Blood
Conventional bench-top blood analysis procedures necessitate the dilution of whole blood prior to measurement due to the high density of RBCs. Typically, this dilution can range from
100-10,000x depending on the technology used. Conversely, for WBC analysis, RBC lysis is necessary to eliminate any measurement obstruction from other cells groups, either optically or electrically. A brief overview of microfluidic techniques used to perform large scale dilution and blood lysis are presented here. Chapter 1. Introduction 9
A B
C D
Figure 1.3: System schematics from various microfluidic devices utilizing Coulter counter tech- nology. (A) Planar electrodes implemented in a microfluidic channel for differentiating RBCs and glutaraldehyde-fixed RBCs [37]. (B)Microfluidic system using Ag/AgCl electrodes for bio- physical characterization of red blood cells (RBCs) and characterization of adult and neonatal RBCs [27]. (C)Fluorescence measurement unit integrated with parallel electrodes for WBC differentiation [39]. (D) PGE electrodes used in DC impedance-based cell counter for CTC enumeration [33].
1.4.1 Dilution
For investigating RBCs, conventional techniques require logarithmic dilution used in series, which can span 2-6 orders of magnitude. This preparation is tedious, time-consuming and lead to an accumulation of non-systematic user dilution error. Several microfluidic techniques exist to address these issues. For large scale microfludic dilution, two prevalent configurations are pyramidal and serial[41–44]. Pyramidal networks generate concentration gradients based on diffusive mixing of fluids by repeatedly splitting and mixing fluid streams. Irimia et al. have used pyramidal dilution schemes to generate accurate chemical gradients to study of cellular responses [43]. The versatility of pyramidal networks in generating gradients of various solutions has allowed it to be applicable for studies in chemotaxis of leukemic cell lines [44]. Pyramidal microfluidic networks, however, are limited by a lack of device compactness and dilution range. Chapter 1. Introduction 10
Typically, the total network area of a pyramidal network increases proportionally to the number of fluidic exits. An alternative to pyramidal networks is the use of serial dilution networks which are capable of generating several order of magnitude concentration profiles. Hattori et al. have demonstrated the use of a serial dilution network capable of generating both linear and logarithmic dilution ratios [42]. Additional modifications to improve performance and compactness such as fluidic mixers have been implemented to eliminate the long serpentine channels required for diffusive mixing [41].
1.4.2 Lysis
WBC analysis requires either the removal of RBCs through either mechanical or chemical means. RBC lysis can be used to eliminate unwanted cell groups which can affect electrical measurements and accuracy. Microfluidic lysis can offer advantages in maintaining a controlled environment which can precisely control mixing ratios and incubation time. Sethu et al. have demonstrated a microfluidic device for rapid RBC lysis under continuous flow (Fig. 1.4(C)).
This was accomplished with a single step lysis procedure. Han et al. improved upon this with a microfluidic device with similar architecture employing a two stage lysis protocol, with lysis and quench to maintain WBC integrity [45]. In both technologies, post-processed lysate could be removed from the device and processed elsewhere. Van Berkel et al. have demonstrated an integrated system for RBC and WBC analysis [46](Fig. 1.4(B)). The system features multiple devices each with specialized functions. Both dilution and lysis is performed on a fluidic process- ing platform before manually transferring blood lysate and diluent to two separate microfluidic analysis chips for perform enumeration.
The integration of lysis and analysis has advantages in reducing operational complexity and device bulk. To enable complete blood count in a point-of-care (POC) environment necessitates a platform which is compact and completely integrated to eliminate user induced difficulties from transferring. For instance, Chempaq (UNITECH, Denmark) is a portable hematology an- alyzer capable of 3-part WBC differential count and hemoglobin concentration measurements
[47](Fig. 1.4(A)). The 3-part WBC differential count is based on the modification of cell volume from a proprietary chemical treatment. Other microfluidic technologies have enabled the inte- gration of cell lysis and downstream electrical measurements to enumerate rare cells [49](Fig. Chapter 1. Introduction 11
A B
C D
Figure 1.4: Overview of existing microfluidic technologies with on-chip sample processing and integration with measurement units. (A)Disposable cartridge capable of accepting whole blood and producing WBC enumeration and hemoglobin measurements. Proprietary lysis solutions induce further size change to enable 3-WBC differential. [47](B)Integrated POC unit for sample processing before user transfer to dedicated electrical analysis devices. [46]. (C)Single stage microfluidic lysis of RBCs under continuous flow using NH4Cl. [48]. (D)Microfluidic device integrating sample lysis from whole blood and downstream electrical measurements for CD4+ and CD8+ T Lymphocytes. [49].
1.4(D)). A completely integrated compact microfluidic platform which can accept whole blood and perform complete blood count for both RBC and WBC populations is still lacking.
1.5 Motivation
There are several microfluidic approaches for achieving sample preparation from whole blood and blood constituent characterization. Sample preparation and pre-processing (e.g., serial dilution and chemical lysis) is important in the autonomous processing of blood to eliminate the need for skilled operators and human induced errors. However, previously microfluidic Chapter 1. Introduction 12 pre-processing of whole blood has been accomplished without the integration of single cell analysis [48, 49]. Other current microfluidic devices using the Coulter principle to measure blood cell sizes and enumerate blood cells target specific cells providing incomplete blood count information[47, 49–52]. These devices also lack on-chip processing[39], necessitate proprietary or fluorescent dye solutions [47,53], and require multiple specialized devices [45]. Additionally, these devices require stringent flow rate controls which need bulky pumping mechanisms. There is an apparent lack of a portable and miniaturized technology capable of providing complete blood enumeration.
Here, I propose a fully monolithic microfluidic device capable of providing complete blood enumeration which integrates on-chip sample preparation and single cell electrical impedance measurements on a simple and compact platform. The monolithic device is capable of accept- ing raw/whole blood for processing and measurement because of integrated sample preparation modules. The use of analogous circuit theory for channel design simplifies typical microfluidic design processes and enables all fluidic flow to be driven by a single pneumatic pressure source.
Thus, this reduces bulk and operational complexity compared to other microfluidic devices.
Integrated filtration and analysis permits cell analysis on the order of 10,000 cells per exper- iment; critical to obtaining statistically significant and clinically relevant sample sizes. The device’s complete functionality (e.g., sample preparation, filtration, analysis), compactness and simplicity of measurement can allow the realization of a portable, hand held and disposable complete blood analyzer unit.
Specific objectives of the complete blood analyzer include:
• Conducting on-chip sample preparation
• Integrating on-chip sample preparation with downstream electrical measurements com-
pactly
• Electrically enumerating and characterizing blood cell constituents (e.g., red blood cell,
white blood cell and platelets)
• Producing CBC results comparable to reference hematology analyzers Chapter 1. Introduction 13
1.6 Thesis outline
An overview of the ensuing chapters is as follows: Chapter2 provides a system overview and design methodology of the fully monolithic for complete blood enumeration. In Chapter3, fabrication and experimental techniques and procedures are discussed. Experimental results and discussion for blood cell enumeration and differential are presented in Chapter4. The thesis is concluded with a summary of contributions in Chapter5 and suggested directions for future work in Chapter6. Chapter 2
System Overview
To accomplish a fully monolithic microfluidic device capable of complete blood count, many existing microfluidic technologies can be integrated into a consolidated platform. This however introduces additional limitations by inheriting each modules’ systematic difficulties thus am- plifying overall complexity. The major difficulties with integration can be separated into two categories: flow requirements and cell densities.
Flow Rate Requirements
Overall flow rate, mixing ratios, incubation time and measurement flow rate need to be finely controlled. RBC and WBC analysis have very different sample preparations and requirements.
According to establish protocols [54], RBC lysis requires specific solution ratios and mixing times, for example:
• 1 : 12 mixing ratio for blood to lysis solution
• 6 seconds incubation
• 1 : 5.3 mixing ratio blood to quench solution
• > 30 seconds incubation
Sub-optimal incubation times can be detrimental to enumeration accuracy. Over exposure to the lysis solution, thus delaying the sample’s exposure to quench solutions and negatively affecting WBC integrity. Under exposure results in un-lyzed RBCs which share similar volume
14 Chapter 2. System Overview 15 characteristics with lymphocytes. Also, cell velocities through the constriction channel used for measurement need to be optimized to ensure a balance between maximizing cell throughput and limitations of the electrical system. Ideally, fluidic flow can be operated with a singular source; either syringe pump (constant flow) or constant pressure thus reducing the additional bulk required by individual solution dedicated fluidic pumps. This reduces operational complexity and results in an ideal platform for P.O.C. measurements. Therefore, significant fluidic design considerations are needed to meet all fluidic and measurement requirements.
Cell Densities
For clinical applications, cell samples blood are highly heterogeneous, and the number of cells of interest can be low (e.g. Post-lysis WBCs). Thus, low-throughput techniques that are only able to test tens or hundreds of cells have little clinical relevance. In order to obtain statistically meaningful data, testing a large number of cells is necessary. RBC-free samples for
WBC analysis have a fixed dilution ratio as per established protocols (1:23) [54]. Conventional bench-top lysis procedures typically rely on centrifugation post-lysis to concentrate and increase
WBC density. While applicable in clinical settings, this luxury is not readily available on a microfluidic platform. The relatively low density of WBCs requires large sample throughput to enumerate upwards of 10,000 cells for clinical relevance. Clogging becomes increasingly more likely for such large volumetric flow of sample through constriction channels. Debris induced clogging can severely limit constriction channel throughput [27,28,55–57] and render the device effectively useless thereafter. Thus, a means to reduce the total number of clogging events or mechanism to circumvent throughput limiting debris is very appreciated.
2.1 System Flow
Fig. 2.1 shows a flow schematic of the microfluidic device that sequentially performs whole blood preprocessing and blood cell enumeration. The device operates in a parallel fashion, siphoning raw/whole blood from a single sample inlet into two separate streams for concurrent RBC and WBC processing and analysis [Fig. 2.2]. A limitation of the coulter principle is coincident events otherwise known as coincidence; a consequence of multiple objects simultaneously present Chapter 2. System Overview 16 in the detection region which makes accurate enumeration difficult. To limit the number of coincident events from the overwhelmingly large concentration of RBCs (3,500-5,000 cells/µL) in raw blood, dilution is necessary for RBC enumeration. For RBC analysis, four separate 1:10 dilution modules are serially employed to effectively dilute RBCs, before the sample reaches in-line filtration modules [Fig.2.2(B)]. A modified serial dilution scheme employing multiple inlets of similar solutions is used in our device design. As seen in Fig. 2.2(B) each dilution module connects to its own individual buffer and waste outlet allowing modifications to overall operation. Two examples of modifications are, firstly, to introduce different buffer solutions at various inlets for multiple solution mixing and secondly, to reduce the intended dilution ratio by dead-ending both inlets and outlets.
Lysis Quench Measurement WBC Whole 1:25 Blood 1:10,000 RBC Dilution Measurement
Figure 2.1: System Flow Chart. Whole blood is siphoned into two streams for RBC and WBC measurement. WBC analysis requires lysis and quench to eliminate troublesome RBCs before measurement.
To properly analyze WBCs, sample lysis and quench is necessary for removing RBCs due to the large density discrepancy between RBCs and WBCs. Following this, flow is filtered and before finally undergoing multi-frequency electrical measurements (10-990 kHz) by the constric- tion region. The size difference between WBCs and RBCs is substantial to differentiate the two cell types (10-15 µm for WBCs vs. 6.2-8.2 µm for RBCs)[58, 59]. This size difference is even more pronounced under flow conditions due to the folding deformation of RBCs causing in an even lower experimental electrical volume [28, 55]. Fig. 2.3 shows representative experi- mental data highlighting the effective electrical volume difference between the two cell groups.
Complete lysis of RBCs is still preferred due to the large discrepancy in cell densities between
RBCs and WBCs. Chapter 2. System Overview 17
(3) (4) A (1)
(2)
(5) B
Raw(Blood
1:10(Dilution( Module Lysis
Quench
In-line( Filtration Measurement
Outlet Inlet Outlet
Figure 2.2: Device Overview. (Left) 3D model depicting the devices geometry and world-to- chip interfaces. Inlets (1-4) denote the location of whole blood, buffer, lysis and quench solution reservoirs respectively. Modular Ag/AgCl electrodes are plugged in and provide downstream electrical analysis. A single negative pressure source (5) is applied to provide the driving force for all fluidic handling on-chip. (Right) 2D schematic of different fluidic modules required for
RBC and WBC analysis. Chapter 2. System Overview 18
To effectively design constriction channels for electrical measurement, the dimensions of constriction channels should be firstly, sufficiently large to avoid physical squeezing cells to ensure precise size correlation from electrical measurements, and secondly, not overly large as to affect electrical detection limits and size resolution. Fig. 2.4 presents quantitative impedance simulation results based on the Coulter principle for varying constriction channel geometries and cells sizes. The overall curvature of each plot exhibits the ability to distinguish between different cell sizes based on an observed impedance change. For the particular measurement channel geometry used here (15×15 µm2) simulation results confirm high volume resolution for cells ¿ 10µm, vital in efforts to differentiate WBCs.
Figure 2.3: Raw Data. Two 500 ms segments raw data showing detection threshold (green), enu- merated cells circled red and highlighting effective electrical volume difference between WBCs (left) and RBCs (right). Chapter 2. System Overview 19
30 25x25bDoublebLayer 25x25bSinglebLayer 25 15x15bDoublebLayer 15x15bSinglebLayer
20
DoublebLayer 15 SinglebLayer
10 ImpedancebChangeb[%] 5
0 8 9 10 11 12 13 14 CellbDiameter
Figure 2.4: Summary of simulation results highlighting the impedance change of spherical cells of different diameters for various channel dimensions.
2.2 Device Design
The relative ratio of flow rates of buffer, sample, lysis and quench is determined by the desired dilution ratio and established lysis protocols [54]. The large number of fluidic processes and the complexity of the network requires the use of an electrical circuit analogy particularly for miniaturization and reducing fluidic flow sources. Analytical solutions describing fluid flow are derived from equivalent circuit equations, which conveniently reduce to a system of linear algebraic equations [60].
12µL R =∼ (2.1) h 3 h wh (1 − 0.63 w
As illustrated in Fig.2.5 an analogous circuit model is used in our work to generate specific mixing ratios, incubation time and preferred flow rates. The fluidic resistance of a rectangular microfluidic channel is shown in Eq. 2.1 provided that the channel width is much larger than Chapter 2. System Overview 20 its height (i.e., w >> h)[61]. If the cross sectional dimensions of all channels and fluid viscosity
µ are constant, the approximation can be reduced to the product of a summarized coefficient
α and the channel length, Rh = αL. Channel lengths can then be used instead of the more P a·s physically applicable but more unwieldy fluidic resistances [ m3 ] as resistor values in the circuit model. Here, all fluidic handling channels are 200 x 40 µm2 as a compromise between minimizing
fluidic resistance and maximizing device compactness. In cases where the channel function (e.g. constriction channels, filtration modules) demanded variation in the cross sectional dimension or fluidic viscosities (e.g. buffer, whole blood), a comparable channel length, of standardized dimensions (200 x 40 µm2) and equivalent fluidic resistance was used as a model input. In our device for blood testing, viscosities of raw blood and solutions can vary greatly. Blood viscosity
WBC RBC
Rl Rm Rq RB,w RB,r Rb,1 i1 i2 i3 RB,w
Rl Rw,1
Rq
Rp,1 Rm Rw,1 4x
i1 i2 i3 Rm,w Rm,r RB,r
Va Rc Rc
Va
Figure 2.5: Analogous Circuit Model. Model used to determine channel geometries and con- figuration to ensure precise fluidic mixing ratios. Subscripts denote fluidic resistors in various modules (B: blood, b: buffer, w: waste, l: lysis, q: quench, m: mixing and c: constriction). Channel lengths are directly used as resistor values. Chapter 2. System Overview 21 is approximately 5.3 times greater than that of buffer solution [62] (5.3 P a · s vs. 1 P a · s).
Consequently, resistor values in the circuit model can be scaled by this difference in viscosity to reflect differences in the summarized coefficient, α between channels handling whole blood and buffer solution. Closed form equations obtained from circuit analysis of Fig. 2.5 are iteratively solved to yield channel lengths for various fluidic resistors (e.g., Rl; lysis resistor, Rb,1; dilution buffer 1 and Rw,1; dilution waste 1) to ensure specific mixing/dilution ratios (dr, 0 < dr < 1). An important objective of the design process was to ensure device operation is independent of applied pressure. Eq. 3.2-3.7 are determined for a fixed ratio of V , where γ = 1 − 1 and ib dr β = 1 . Additionally, sample flow rate can be directly controlled by modulation of applied dr pressure (V ) without consequence to any of the on-chip mixing ratios due to the fixed nature of V . ib
γRb,1 = RB (2.2) V γRb,1 + βRm + γRw,1 = (2.3) ib V γRw,1 + γRb,2 − Ri = (2.4) ib . . V γRb,n + βRm + γRw,n = (2.5) ib V γRw,n + γRb,n+1 − Ri = (2.6) ib Ri + Rc V γRb,n + β Rm + = (2.7) Rc ib
RBC/WBC measurement modules can be operated simultaneously with a single pressure source. This has advantages in terms of portability and compactness over microfluidic devices requiring stringent and bulky flow rate control, and typically requires an individual source for each solution (e.g. lysis solution, dilution buffer, quench solution). In addition, the short total channel length and hence lower pressure requirement of our device compared to other technologies [42,63], makes it more suited for portable applications especially since the operation pressure of 1.5 kPa is easily attainable without the need of compressed gases and cumbersome Chapter 2. System Overview 22 equipment.
This technique of using an electric circuit analogy permits the design of microfluidic networks that are independent of flow rates, applied pressure, mixing ratios and viscosities. The modified serial dilution scheme designed via circuit analogies allows for a modularity system capable of adjusting multiple mixing solutions and ratios post fabrication. By scaling channels of various dimensions, functions and viscosities to a length of standardized dimensions, we address limitations of previous design efforts in pressure driven microfluidics. Users can capitalize on the versatility of both the device design process and device operation to develop complex microfluidic systems capable of handling countless biological tasks.
2.3 Filter Design and Implementation
Debris can plague the use of constriction channels for measurement and can vary in size, de- formability and material make-up (Fig. 2.6). The lysis of RBCs is rarely 100% debris free and produces highly deformable RBC ghosts - remnant RBC membranes without internal compo- nents [64].Individually the ghosts can freely travel but can cluster and collectively impede elec- trical measurements by clogging the constriction channel. Increasingly higher negative pressure can be applied to dislodge the clog but due to the porous nature of a debris cluster, remov- ing the clog by applying high pressure is not always effective. The use of an in-line filtration module can drastically reduce the occurrences of constriction channel clogging by forcing the sample flow through numerous small and parallel orifices of comparable size to the downstream constriction channel[64–67].
Implementing filtration structures can increase device fluidic resistance and can limit the devices ability to operate with low pneumatic pressures. The overall fluidic resistance of a filtra- tion module can be reduced by a factor of n by maximizing the number of parallel orifices/filter
Ri channels (n), resulting in a local resistance of n , where Ri is the resistance of a single filter channel. The filter module uses 100 parallel 15×15 µm2 channels. This produces an equivalent
12 P a·s 2 effective resistance of 5.05 × 10 m3 , while the fluidic resistance of a 200 × 40 µm channel of 12 P a·s equivalent length is comparable at 1.61 × 10 m3 . Minimizing overall fluidic resistance can be accomplished by using numerous filter channels in parallel. Fig. 2.7 summarizes finite element Chapter 2. System Overview 23
Cellular Environmental
Figure 2.6: Various types of debris which can occlude constriction measurements channels and limit experimental throughput. simulation results used to theoretically compare filter performance, of several existing filtration configurations [65–67] with our new offset filter configuration.
The offset filter configuration has module inlets and outlets positioned rotationally sym- metric from each other. As more filter orifices are clogged, there is an undesirable increase in fluidic resistance. Inevitably, filter modules will become clogged but its fluidic behavior in terms of resistance can be optimized to allow for even longer experimental durations. Numeric simulations used to determine fluidic resistance can be of assistance. First, fluid flow behavior is simulated for all filter configurations and filter channels experiencing the largest volumetric are determined. Second, we preferentially occlude filter channels in decreasing order of volumetric
flow rate based on the previous results to determine filter behavior over its lifetime. An advan- tage of the offset configuration compared to existing filter configurations is the uniform pressure distribution across the length of the array (i.e. the pressure drop is equal for all filter channels).
This results in less severe resistance changes as increasingly more filter orifices/channels are occluded. At a 50% clogging state, the resistance of the offset design increases by only 26.8% in comparison to 35.2%, 68.7% and 166% for other configurations. This improvement allowed for more accurate fluidic handling, longer experimental durations and increased total sample throughput.
Plug-in Ag/AgCl electrodes were used for measurement through fluidic reservoirs (Fig. 2.2).
The implementation of filtration modules negatively obscures electrical measurements in tra- ditional in-line plug-in electrodes used elsewhere [27, 28, 55–57]. We reconfigured the electrode Chapter 2. System Overview 24
Square Triangle T Wave Offset
Inlet Outlet
250 T Triangle SquareSWave 200 Offset
150
100
50 RelativeSResistanceSChangeS[%]
0 10 20 30 40 50 60 70 TotalSOrficesSCloggedS[%]
Figure 2.7: Summary of simulation results comparing unwanted fluidic resistance change as increasing percentage of orifices becoming clogged for existing and our new offset filter configu- rations. When 50% of filter orifices become clogged, our offset filter increases fluidic resistance only by 26.8%. placement to maintain the simplicity of plug-in electrodes but still enabled the integration of fil- tration structures and the circumvention of cell settling at electrode regions for previous designs
Fig.2.8(A). For conventional plug-in electrodes, the sudden expansion of channel width/height produces local velocities less than the critical cell velocity which can cause cell sedimentation.
With the electrode placed perpendicular to the fluid flow direction, it remains electrically con- nected to measurement regions through an intermediate fluid filled channel acting as a salt bridge. The fluidic reservoir for the electrode experiences zero fluid flow under normal device operation. This conveniently provides a reservoir for removing foreign debris which may have bypassed the filtration module and obstructed the constriction channel. While reversible flow Chapter 2. System Overview 25 has been used to unclog microchannels [68,69] debris is still present within the fluidic network.
Instead, we modify the applied pressure on inlets to various states (−p, 0 or dead ended) and subtly redirect fluid flow into the electrode reservoir to unclog the measurement channel and clear obstructions (Fig. 2.8(B)). This eliminates problematic debris allowing experimentation to continue as usual. Chapter 2. System Overview 26
A New Reconfigured Conventional Electrode Electrode
B 0 0 ~ ~ ~ -p
-p -p 0
Figure 2.8: Schematic highlighting reconfiguration of conventional plug-in electrodes.(A) Com- parison of new reconfigured electrode placement vs. conventional plug-in electrodes. (B) Use of the new electrode placement as a reservoir for debris. Fluid flow can intermittently be adjusted to clear obstructions and increase device throughput. Chapter 2. System Overview 27
In summary, maximizing parallel filter channels and employing a new offset configuration can minimize both the initial fluidic resistance and its undesirable increase over time. Plug-in electrodes utilizes a measurement mechanism which shares many advantages with aligned and parallel electrodes [36](Fig. 2.9). Integration of these beneficial electrodes into conventional microfluidics can be achieved by reconfiguring the electrode placement which conveniently also promotes additional techniques for mitigating clogging of constructional channels. The combi- nation of integrated filtration structures and an unclogging mechanism can be broadly applicable and would enable constriction channel operation for extended durations (e.g., measurement of
10,000 cells per experiment).
Aligned Parallel Planar
Figure 2.9: Schematic showing various electrode configurations used in previous microfluidics. Chapter 3
Fabrication and Experimental Methods
3.1 Fabrication
Devices consisted of a single layer of polydimethylsiloxane (PDMS) and was fabricated using standard soft lithography. Overall, the device consists of three separate areas each with dis- tinct cross sectional dimensions: the constriction region (15 x 15 µm2 and 20 µm in length), intermediate channels between electrodes (1000 x 40 µm2) and all remaining fluidic and reagent handling channels (200 x 40 µm2). Although the basal impedance is largely controlled by the geometric dimensions of the constriction channel, the intermediary channels contributed sub- stantial impedance and cannot be ignored particularly when considering relative impedance change.
Briefly, two layers of SU-8 are sequential developed to create the multi-layered device (15 µm and 40 µm in height). The entirety of the fabrication process is summarized in Table 3.2 with
PR Type Spin [RPM, s] Exposure [s] Bake (tpre, tpost)[min]
Seed SU8-5 3000, 35 6 4, 2 15 µm SU8-25 3000, 35 10 10, 4 25 µm SU8-25 2000, 35 12 10, 4
Table 3.1: Summary of Fabrication Parameters
28 Chapter 3. Fabrication and Experimental Methods 29
Step Procedure
1 Dehydration of glass slides in the hotplate of 150oC for 30 min 2 Pour and cover glass slide surface with SU-8 25 negative photoresist 3 Program spinner - Step one: 500 rpm, 1 acceleration, spin time of 10 sec, step two: (15 µm - 3000 rpm, 25 µm - 2000 rpm), 3 acceleration, spin time of 35 sec 4 Pre Exposure Bake (65oC for 3 min, 95oC for 7 min) 5 Place slide and chrome mask in mask aligner. Program mask aligner (hard-expose, expose time = (15 µm - 10 s, 25 µm - 12 s). Expose glass slide. 6 Post Exposure Bake (step one: 65oC for 1 min; step two: 95oC for 3 min) 7 Develop exposed glass slide in SU-8 developer solution for 30 seconds under strong agitation.
Table 3.2: Summary of Fabrication Sequence specific fabrication parameters in Table 3.1. Fig.3.1 highlights schematically the fabrication sequence.
GlassmSlide
1stmLayermSU-8
Constrictionm LayermMask
Exposem&mDevelop
2ndmLayermSU-8
Mixingm LayermMask
Exposem&mDevelop
CompletedmMaster
Figure 3.1: Schematic of multi-layered fabrication sequence. Two SU-8 feature layers are used to generate channel heights of 15 µm and 40 µm. Chapter 3. Fabrication and Experimental Methods 30
3.2 Experimental Procedure
Blood samples were obtained from healthy donors via venipuncture (Mount Sinai Hospital,
Toronto, Canada). Hematological parameters of whole blood samples varied within healthy physiological ranges (Lymph %: 8-32, WBC concentration: 10-13 cells/nL, MCV: 83-96 fL, RBC concentration: 3-5 cells/pL). Blood samples were anticoagulated with EDTA anticoagulant
(ethylenediaminetetraacetic acid 1.5 mg ml−1) (Sigma-Aldrich, Oakville, ON, Canada) and stored at room temperature prior to use within 12 hours of withdrawal. For each patient sample, an aliquot of the sample was also measured by a standard hematology analyzer (XN-
9000 , Sysmex America, Illinois) for reference.
PBS with 1% w/v BSA (New England Biolabs Inc., Herts, UK) was used for device in- cubation and as RBC diluent. Lysis and quench solutions were prepared based on previously established bulk lysis protocols[39,54]. Briefly, 10 L of whole blood is mixed on-chip in a 1:12 ratio with lysis solution (0.12% v/v formic acid and 0.05% saponin in DI) and subsequently neutralized by 1:5.3 with a modified quench solution (0.6% w/v sodium carbonate in 1X PBS).
Prior to experiments, devices were incubated with PBS, all inlets and outlets were sealed and the device was pressurized to remove trapped gas pockets via diffusion through the PDMS channel wall. Whole blood, lysis and quench solutions were pipetted into their respective inlet ports/reservoirs. Two sets of Ag/AgCl electrodes, one for each measurement module, were plugged into the device through fluidic reservoirs. Sample and dilution outlets were then connected externally to a custom pneumatic pressure source that drives fluidic flow. Fig.3.2 highlights various microfluidic modules used to perform lysis, filtration and measurement.
3.3 Signal Analysis
Sinusoidal voltage signals at multiple frequencies (10, 403, 700 and 990 kHz at 500 mVpp) were applied through Ag/AgCl electrodes. Raw experimental voltage data was collected (Zurich
Instruments, HF2IS, Switzerland) and converted to an impedance reading. A summary of experimental parameters used for the impedance analyzer can be found in Table 3.3. Entire impedance data files would be windowed into shortened segments, 7.2 kSa in length. Firstly, a histogram of impedance values was generated for the entire window. Secondly, a Gaussian Chapter 3. Fabrication and Experimental Methods 31
A B
C D
15x15 μm2 V constriction channel
Inlet Outlet
Figure 3.2: Images of device operation and device schematic highlighting key modules for RBC (right) and WBC (left) measurements. (A) Module used to lyse RBCs with lysis solution. (B) Fully monolithic microfluidic device for blood cell enumeration on 18 x 48 mm2 footprint. (C) Downstream electrical measurement. WBCs are circled. (D) Inline filtration structures used to eliminate debris. RBCs are circled. All scale bars are 100 m. distribution is then fitted to the histogram data and µ value would be the extracted basal impedance. Peak analysis would be conducted using simple thresholding techniques. Peaks greater than the set threshold, a numeric value above the basal impedance, are kept and counted as cells.
During signal processing for RBC analysis, WBCs and platelets were excluded based on distinct impedance signatures. Platelets in particular reflect <1% of the detection volume making their enumeration challenging. It is difficult to distinguish small platelet sized particles
Settings
Analyzer Range: 490 mVpk Freq: 10k, 403k, 700k and 990k Sampling rate: 14.4k Sa/s Filter 48 dB/Oct, Bandwidth: 120 Hz Amplifier R(V/A) = 10k, G=1 Total gain = 10k(V/A)
Table 3.3: Summary of Impedance Analyzer settings Chapter 3. Fabrication and Experimental Methods 32 with the current measurement channel geometry. However, due to their relatively small size, the presence of platelets can easily be accounted for by identifying minima in the RBC size histogram. This represents a limit of detection for RBC analysis (Fig.3.3). WBCs, on the other hand, can easily be distinguished using both their large relative volume and rare occurrence rate.
400
350
300
250 Limitssofsdetection
200
150 FrequencysCount 100
50
0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 ImpedancesChanges(Ω)
Figure 3.3: Typical non-gated RBC histogram highlighting limits of detection for low electrical volume cells. Platelets or particles sized lower than the limit of detection are ignored from analysis due to difficulties in distinguishing them from baseline noise.
The inherent nature of a negative pressure driven device is that 0 gauge pressure is applied to the fluidic reservoirs and are subject to fluid evaporation over extended periods. Along with variability in post-lysis release of cellular components and cell densities, basal impedances could shift over time. Fig. 3.4 shows a mixing channel used for incubation post-lysis and quench.
Both released cellular components of RBC and RBC ghosts negatively obscure electrical mea- surements and contribute to shifting baselines. Typical histogram based methods of detecting basal impedances are not suitable to use because of baseline shifting. A baseline correction algorithm was developed. It consisted of fitting and then subtracting a low ordered polynomial from windowed segment of raw data Fig. 3.5(Inset). A comparison between size histograms Chapter 3. Fabrication and Experimental Methods 33 produced from raw impedance signals and baseline corrected signals can be found in Fig. 3.5.
Without any corrections, baselines are often incorrectly calculated which leads to numerous peaks at lower electrical volumes mis-identified as cells. Size measurement accuracy improves with the baseline correction as evidenced by the characteristic RBC distribution.
For smaller cells typically those <4 kΩ in WBC analysis, an additional signal processing algorithm was implemented to eliminate false identification. Identified peaks are isolated and compared to a signal template. Cells passing through the constriction region are predicted to generate a purely Gaussian signal due to the symmetric nature of measurement. Cells which have clustered or have been affected by RBC ghosts can exhibit non-symmetric electrical signals. Raw and template signals are correlated to generate a correlation coefficient (R2).
Detected peaks which produce a R2 > 0.95, are categorized as cells while others are omitted from statistical analysis.
Figure 3.4: Mixing channel for incubation post-lysis and quench. RBC ghosts and cellular components are visible and can contribute electrical interference. WBCs are circled. Scale bar 200 µm. Chapter 3. Fabrication and Experimental Methods 34
25 Raw Corrected
20 Ω
15
t 10 Frequency [%] Frequency
5
0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Impedance [Ω]
Figure 3.5: RBC size histograms for raw and post baseline correction of the windowed sample are presented. Raw histogram (blue) has large number of cells at lower electrical volumes due to mis-identification. Size histogram from baseline correction shows improvement in size accuracy and produces expected RBC distribution.(Inset) Summary of baseline correction algorithm. Low ordered polynomial (Red) is fitted and subtracted from raw data to eliminate baseline trends. Chapter 4
Experimental Results and Discussion
4.1 Device Characterization
4.1.1 Pressure Independence
The device was operated at varying pressures to determine its performance variation and to validate correct mixing ratios based on the circuit model analogy. Negative pressures ranging from 1,200 to 2,100 Pa were applied via both the dilution and outlet ports. Two junctions were observed via microscopy imaging to confirm agreement between experimental and theoretical mixing ratios. The apparent mixing ratio was calculated by measuring the width of the merging
fluidic streams immediately adjacent to junction W1 (Fig. 4.1 Inset). Relative ratios of the WT fluidic widths were directly correlated to their relative volumetric flow rate [70]. Observing a location adjacent and downstream from the junction ensures that there is insufficient time and channel length to allow diffusion, and volumetric flow is reflected entirely in the measured ratio of fluid widths. The theoretical mixing ratios for RBC dilution and WBC lysis are 1:10 (0.1) and 1:12 (0.083), respectively. The apparent mixing ratio remains relatively constant for both blood-lysis and blood-buffer mixing. The measured mixing ratio for all applied pressures was
0.11 ± 0.03 and 0.089 ± 0.012 for RBC and WBC junctions, respectively (n = 3) (Fig. 4.1).
35 Chapter 4. Experimental Results and Discussion 36
0.2 RBCLDilution WT W1 WBCLLysis 0.15
0.1
ApparentLMixingLRatio 0.05
0 1200 1500 1800 2100 AppliedLPressureL[Pa]
Figure 4.1: Mixing ratios are independent of pressure variations. The use of an inlet-outlet configuration for serial dilution is used to enable the device to behave consistently independent of the applied pressure. According to theoretical calculation, the expected mixing ratios are 0.1 and 0.083 for RBC dilution and WBC lysis, respectively. (Inset) Relative ratios of the fluidic widths ( W1 ) were directly correlated to their relative volumetric flow rate. WT
Inherently linked to the applied pressure is the total volumetric flow through the device which varies proportionally to the applied pressure by a ratio ( V ) that is fixed during device ib design (see Eq. 2.3). While the presence of this ratio ensures the relative volumetric flow of each solution remains constant, device operational success and accuracy deviates only at non-optimal pressures. Fluidic mixing relies on diffusion in the process of flowing through a long serpentine channel. For WBC lysis operated at 1,500 Pa, the average residence time in the fluidic mixing after introducing lysis solution was 5.8 s. The whole blood mixture ideally requires 6 seconds of incubation to ensure complete lysis of RBCs [54]. At applied pressures of 1,800 and 2,100 Pa, the average residence time is 4.65 s and 4.01 s, which resulted in many Chapter 4. Experimental Results and Discussion 37 un-lysed RBCs observed in both the filter and measurement regions.
4.1.2 Flowrate Control
Ions present in the measurement channel can respond to the electrical fields upon application of an AC voltage along the measurement channel [50, 71]. This results in an ionic redistribution, which effectively changes the observed impedance of the constriction channel. Flow rate through measurement channel has been shown to be proportional to the ionic concentration and mass transport. The mass transport limited current iL is directly proportional to the volumetric flow rate Q. Eq. 4.1 describes this relationship where, n is the number of electrons transferred, F is the Faraday constant,DA is the diffusion coefficient and l, h, w, we are various electrode and cell geometries. The use of this relation provides a technique to determine volumetric flow rate from observed impedance changes similar to that of Hassan et al.[50]. Relative impedance change
(Zf /Zo) from steady state impedance (Zo) is observed for various flow rates established using a controlled flow rate. Fig 4.2 summarizes results of relative impedance change for varying
flow rates from 0.5-30 µL/min. Although experimental results show strong linear correlation between flow rate and impedance change, actual implementation using cell lysate is challenging.
The lysis of cells causes the release of intracellular ionic contents which affect the overall basal impedance and conductivity of the medium. There is large variation between patient samples in both the reference basal and basal impedance changes over time. While there is a positive correlation between observed basal impedance and applied pressure, approximated flow rates were difficult to confirm. The current applicability of using impedance change to determine
flow rate while using the same signal for electrical analysis of cells is minimal. However, with additional fine calibration, especially using alternative means to establish flow rates on even smaller scales, accurate flow metering can be accomplished.
2/3 1/3 2 2 1/3 iL = 0.925nF [A]bulkDA Q we(t /h w) (4.1) Chapter 4. Experimental Results and Discussion 38
1
0.99
0.98
0.97 o
/Z 0.96 f Z
0.95
0.94
0.93
0.92 0 5 10 15 20 25 30 35 Flow Rate [µL/min]
Figure 4.2: Flow Metering. Effects of varying flow rates on basal impedance are observed. Linear trend indicates the possibility of using observed impedance change to approximate vol- umetric flow rate. Flow rates ranging from 0.5-30 µL/min.
4.2 Red Blood Cell Analysis
4.2.1 Red Blood Cell Enumeration
RBC enumeration was conducted using various dilution ratios of 100×, 1, 000× and 10, 000×.
Fig. 4.3(A) shows raw impedance data captured during one second of RBC measurement and the corresponding frequency histogram for extended experimentation of 2000 cells. For lower dilution ratios, as expected, there is a high number of coincidence events. The aligned measurement mechanism coupled with the high cell density presents difficulties in the form of baseline identification and distinguishing between multiple cells travelling successively or simultaneously. Fig. 4.3(A) shows that both size histograms for 100× and 1, 000× dilution ratios are right skewed and have a maximum frequency (%) at impedance values greater than the expected impedance change for RBCs. This combination confirms that multiple cells are often incorrectly enumerated as a single large cell. For a dilution ratio of 10, 000×, the system Chapter 4. Experimental Results and Discussion 39
Measured range r2 Bias ±σ Bias % 95% LOA LOA %
RBC [c/pL] 3.5-4.8 0.83 0.1±0.3 <3% [-0.4,0.6] <26%
Table 4.1: RBC concentration: Bland Altman parameters for concordance of measurements from standard analyzer and microfluidic device. is able to reestablish a baseline from which we can detect successive cells flowing through the measurement channel. The absence of larger peaks > 4kΩ indicates minimal coincidence. The resultant size histogram concurs with reference hematology results for both distribution shape and mean.
As shown in Fig. 4.3(B), high dilution ratios (10, 000×) best mitigated the effect of co- incidence and produced results most close to reference RBC concentration benchmarked by a commercial hematology analyzer. Particularly at 100× and 1, 000× dilutions, the mistaken identification of multiple cells as a single cell results in much lower concentration measurement results. The stochastic nature of coincidence makes it difficult to adjust using a correction factor. The combined average standard deviation for dilution ratios of 1,000 and 100 was 1.076 cells/pL. 104,735 total RBCs (n = 10) were subsequently analyzed to produce concentration results with strong correlation with reference results (R2 = 0.83) and an average difference of
+0.1 cells/pL (Fig. 4.6(A)).
4.2.2 Red Blood Cell Characterization
The measured electrical volume is difficult to correlate with its single cell volume when consid- ering its resting biconcave shape and its deformed parachute shape in microfluidic channels[55].
However, the electrical volume exhibited by a population of RBCs can still be correlated to its mean cell volume, which is similar to the procedures performed in hematology analyzers.
In addition to effect of dilution presented in Sec. 4.2, RBC characterization requires fine con- trol over flow rate. For increasing applied pressures and thus increasing flow rate through the constriction measurement channel; increased RBC deformation is observed. This results in varying mean cell size for identical samples(Fig. 4.4). Predictably, mean peak size decreases for increasing applied pressures. Fig 4.4 (Inset Top) shows two RBC size histograms produced from various pressures. The deformation of RBCs under flow can affect both the mean value Chapter 4. Experimental Results and Discussion 40
A 20 3 10 2 100x
0 1
-10 0 0 0.2 0.4 0.6 0.8 1 0 5 10 15 20 15 6
10 4 10B000x 1B000x 5 2 0
-5 0 0 0.2 0.4 0.6 0.8 1 Frequencyx][R 0 2 4 6 8 10 Impedancex]k Ω R 6 10
4
2 5
0
-2 0 0 0.2 0.4 0.6 0.8 1 0 2 4 6 8 Timex]sR Impedancex]kΩR B 5
4 /L] 12
3
2 Concentrationx[10 1
0 Reference 10B000x 1B000x 100x
Figure 4.3: Effect of dilution ratio on RBC enumeration accuracy. (A) 1 second window of raw data at varying dilution ratios (100−10000×) and corresponding size histograms for 2,000 cells per sample. 10, 000× dilution allows for accurate detection of single RBCs. Both maximum values and distribution shape of the size histogram are evidence of incorrect enumeration of clustered cells at lower dilution ratios. (B) Concentration results for varying dilution ratios, compared with reference concentration measured by hematology analyzer. Chapter 4. Experimental Results and Discussion 41 of the histogram (µ) and the distribution width. Thus, well controlled pressure is imperative to accurate RBC characterization. For pressures >2000 Pa, the mean peak size appears to approach a steady value indicating RBCs reaching a deformed folded state where no further deformation can occur. RBC size characterization for these pressures will likely yield compa- rable results. In experiments, the pressure is well controlled and has an maximum/minimum range of ±50 Pa. Particularly, for applied pressures > 1500 the mean peak size is comparable.
The applied pressure chosen for experimental use was 1500 Pa as a balance between reasonable cell throughput and electrical enumeration limits.
25 2500 1800%Pa 1200%Pa 20
15 2000
10 Frequency%[%]
5 1500
0 0 500 1000 1500 2000 2500 3000 3500 Peak%Size%[Ω] 1000 Average%Peak%Size%[ Ω ] 500
0 1000 1200 1400 1600 1800 2000 2200 Applied%Pressure%[Pa]
Figure 4.4: RBC Deformability and its effect on MCV. Mean/average peak size for an RBC sample consisting of 2000 cells. Vertical error bars denote standard deviation (σ) and horizontal error bars denote maximum and minimum pressure deviations. (Inset Bottom) A schematic showing the deformed parachute shape of an RBC in a small microfluidic channel. (Inset Top) Size histograms for various pressures. Both µ and distribution are affected by increasing pressure.
Fig. 4.5(A) is an experimentally generated RBC size histogram showing various RBC indices that were measured. Mean corpuscular volume (MCV) was quantified by determining the mean Chapter 4. Experimental Results and Discussion 42 value of the fitted Gaussian curve, while the distribution width was determined by the width of the fit at 20% of the maximum height. Reference MCV measurements and measured mean electrical volume were fitted produce a correlation coefficient (R2) of 0.97 [Fig. 4.5(B)]. The hematocrit of blood samples encompassed a wide range varying from 32% to 46%. The MCV results correlated independently of the sample hematocrit. The accurate measurement of MCV demonstrates just one of many hematologic parameters capable of being determined by the analysis of experimental RBC size histograms. Chapter 4. Experimental Results and Discussion 43
8
RDWHSD A 7
MCV 6
5
4
3 Frequency1[7]
B n = 9264
A
i i 5ii Aiii A5ii Biii B5ii Impedance1Change1fΩL B Aii 5i
4i 95
3i
B 9i R =1it97
Bi Reference1MCV1[fL]
85 Reference1Hematocrit1[7] Ai
MCV Hematocrit 8i i 6ii 65i 7ii 75i 8ii 85i 9ii 95i Aiii Measured1MCV1[Ω]
Figure 4.5: RBC characterization. (A) RBC size histogram highlighting specific measured RBC indices: distribution width (RDW-SD) and mean corpuscular volume (MCV).(B) Correlation plot for mean corpuscular volume (MCV): reference hematology analyzer (y) vs. microfluidically measured MCV (x). Line fitting generates a correlation coefficient of 0.97. Sample hematocrit
(Green) is plotted to show independence of measured MCV. Chapter 4. Experimental Results and Discussion 44
6 0.8
5.5 B A 0.6 Bias0(0.11 5 95B0Limits0(,0.4%0.61 0.4 R2=00.83 4.5 0.2 4 0 3.5
,0.2 3
2.5 ,0.4 Reference0RBC0Conc.0[cells/pL]
2 Difference0between0methods0[cells/pL] ,0.6 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5 3 3.5 4 4.5 5 5.5 Device0RBC0Conc.0[cells/pL] Average0of0both0methods0[cells/pL]
Figure 4.6: Concordance and Bland-Altman Plots for RBC Enumeration. (A) Concordance plot for device and reference hematology analyzer produced RBC concentration (B) Bland-Altman plot comparing two techniques. Bias and limits of agreement (bias +2σ, LOA) are shown in blue and red respectively.
4.3 White Blood Cell Analysis
Post-lysis RBC-free sample flows to the WBC constriction channel for enumerating and dif- ferentiating two major WBC cell types: lymphocytes and a grouped category of granulocytes and monocytes. The inherent cell size difference between lymphocytes and non-lymphocytes in a post-lysis sample was first confirmed using microscopy imaging (see Fig. 4.7). There is an apparent size difference which generates a bimodal distribution of WBCs, centered around
70 fL for lymphocytes and 210 fL for non-lymphocytes, which concurs with size characteristics determined by hematology analyzers [72]. Chapter 4. Experimental Results and Discussion 45
LymphocyteL(~8Lμm)L NeutrophilL(~15Lμm)L MonocytesL(~18Lμm)L
7 SampleL1
6 SampleL2 Fit
5
4
3
FrequencyLCount 2
1
0 6 8 10 12 14 16 18 20 22 DiameterL[µm]
Figure 4.7: Optical confirmation of size difference of different WBC subtypes (e.g. lymphocytes, neutrophils and monocytes). Scale bars are 15 µm. (Below) Size histogram from optical mea- surements from two separate samples (n = 50). Fitted curve indicates a bi-modal distribution.
Enumeration and size characterization of WBCs (97,305 total from 10 samples) was done using the system. Individually RBC ghosts produce minimal electrical signals, but the rel- atively high concentration of RBC ghosts can collectively cause false identification with the smallest WBC sized group (i.e., lymphocytes). Hence, low and high frequency impedance data is used in tandem to selectively enumerate WBCs from RBC ghosts. Moderately high fre- quency impedance data (100-400 kHz) is used to identify WBC peaks and synchronized with both lower frequencies (10 kHz) and higher frequencies (990 kHz) to generate multi-frequency information (Fig. 4.8). At lower measurement frequencies, impedance data most accurately reflects cell volume but is also most susceptible to false identification of RBC ghosts and debris.
Higher frequencies provide impedance data that is dependent on both cell size and intracellular Chapter 4. Experimental Results and Discussion 46 properties. At these frequencies RBC ghosts exhibit very minimal electrical signatures and can easily be excluded from measurements. Opacity, which is the ratio of cell impedance at high frequency and low frequencies, is used to differentiate WBC subgroups [39, 45, 73]. Fig. 4.9 shows scatter graphs of low frequency (i.e., cell electrical volume) and opacity data for various lymphocyte ratios within and out of normal physiological ranges. Opacity data here cannot conclusively differentiate between granulocytes and monocytes, but low frequency data in com- bination with high frequency data for cell identification is sufficient to generate a predictable bimodal size histogram and provide 2-WBC differentials. Minimums in the size histograms are identified as lymphocyte flags, and cells are categorized according to their relative position to these generated gating flags.
4 × 10 3.5 10 kHz 400 kHz 3 700 kHz
2.5 990 kHz
2
1.5
1 Impedance [ Ω ]
0.5
0
-0.5 0 50 100 150 200 250 300 350 400 450 500 Time [ms]
Figure 4.8: Sample Multi-Frequency Measurements for WBC Analysis. 10 kHz to 990 kHz signals were used for WBC analysis. Identified peaks are circled in red. Lower frequency provides purely size information, while higher frequencies give insight on intracellular properties. Chapter 4. Experimental Results and Discussion 47 A B
C D
Figure 4.9: Scatter graph for WBC measurement showing opacity and electrical volume from healthy patient samples of various lymphocyte populations. Abnormally low 8.3% (C), within healthy range 22.5-23% (A-B), and abnormally high 31.8% (D).
Blood samples from 10 healthy patients were analyzed for total WBC concentration and
WBC differential using our monolithic microfluidic device. Electrical measurements were com- pared to reference results from a standard hematology analyzer. Bland Altman analysis is used to assess the agreement between the two analysis methods. Fig. 4.10 shows concordance
(left) and Bland-Altman plots (right) for various blood cell parameters where the discrepancy between device and reference measurements is plotted against the average of the two (n = 10).
Bias and the 95% limits of agreement LOA, (bias 1.96σ) are plotted in blue and red respec- tively. WBC concentration measurements show a strong linear correlation (R2 = 0.94) and has Chapter 4. Experimental Results and Discussion 48 an average discrepancy that is close to zero (-0.5 cells/pL) (Fig. 4.10(B)). On average, there is a slight undercounting of WBCs by the microfluidic device relative to reference analyzers.
The limits of agreement are relatively narrow [+0.0,-1.1] and is < 16% of the healthy WBC concentration range[59]. Device generated total WBC concentration and WBC differential bias values were within 4.5% and 6.8% of reference measurements. Additionally, the LOA widths were within ±9.9% and ±30.9%. The device also shows the ability to accurately differentiate
WBCs for lymphocyte ratios ranging from 8-32% with high correlation (R2 = 0.97).
A 14 7y2 7 13 Bias3R,7y5=
,7y2 95W3Limits3R,1y1%7y7= R2=37y94 12 ,7y4
,7y6 11
,7y8
17 ,1 ReferenceWBC3Concy3[cellspnL]
9 Difference3between3methods3[cellspnL] ,1y2 9y5 17 17y5 11 11y5 12 9 9y5 17 17y5 11 11y5 12 12y5 13 Device3WBC3Concentration3[cellspnL] Average3of3both3methods3[cellspnL] B 35
4 Bias3R1y4= 37 95W3Limits3R,1y8%4y5= R2=37y97
25
2
27
15 7
17 Difference3between3methods3[W] Reference3Lymphocyte3Differential3[W] 5 ,2 17 12 14 16 18 27 22 24 26 28 37 7 5 17 15 27 25 37 35 Device3Lymphocyte3Differential3[W] Average3of3both3methods3[W]
Figure 4.10: Concordance and Bland-Altman Plots for WBC enumeration and differentiation.
(A) WBC Concentration comparisons (B) WBC differential comparisons. Bias and limits of agreement (bias +2σ, LOA) are shown in blue and red respectively.
Bias % can be used an overall indicator of device accuracy when compared to reference values. Currently, there does not exist a universal standard/guideline outlining the limits of bias % for clinical relevance or reliability. While the bias % of all three measured blood cell Chapter 4. Experimental Results and Discussion 49
Measured range r2 Bias ±σ Bias % 95% LOA LOA %
WBC [c/nL] 9.7-11.2 0.94 -0.5±0.3 <5% [-1.1,0.0] <10% WBC diff [%] 11.3-30.1 0.97 1.4±0.7 <7% [-1.8,4.5] <31%
Table 4.2: WBC Parameters; Bland Altman parameters for concordance of measurements from standard analyzer and microfluidic device. parameters are relatively low, additionally studies giving insight on the current microfluidic standard for bias % can be helpful. Hence, a comparative study of percentage bias and other microfluidic technologies is conducted. The most directly applicable study was done by Hollis et al. [73]. Hollis et al. conducted a comparison of venous and capillary differential WBC counts using both standard hemotology analyzer and a microfluidic impedance cytometer. For total WBC concentration results (n = 9), Hollis et al. produced bia % and LOA widths of
7% and ±22% respectively when compared to reference analyzer results. Additional studies enumerating CD4+/CD8+ cells have used Bland Altman analysis for comparative purposes.
Watkins et al. produced bias % of 2.8% and 1.5% for CD4+ and CD8+ cells respectively [49].
Microdevices used at the point-of-care for CD4+ enumeration from Tanzanian HIV-Infected patient samples (n = 130) produced a bias % of 5.1% [74]. Consistently, the bias % for microfluidic devices is small such that 1% < microfluidic measurement < 10%. A summary of this comparative study can be found in Table 4.3. Bias % produced here of 5% and 7% is of the same order of magnitude as other microfluidic enumeration technologies. While there is a lack of a commercial/clinical standard for bias %, the comparable values to other accepted microfluidic enumeration platforms give confidence in the device’s ability to accurately perform complete blood counts.
On-chip sample preparation used in this device proved effective in both the dilution and lysis of RBCs to enable accurate enumeration of RBCs and WBCs. Final concentration results can
n be represented by C = f·t dr, where n: number of detected cells, f: approximated flow rate, t: experiment duration and dr being the dilution ratio. Two major difficulties exist in predicting cell concentrations from electrical measurements; firstly, the approximation of volumetric flow on-chip and secondly, overall cell loss.
It is important in determining cell concentrations to accurately calculate the sample volume Chapter 4. Experimental Results and Discussion 50
Techniques Targeted Cells Summary Bias %
Coplanar electrodes CD4+/CD8+ Point of care HIV diagnostics 2.8% using whole blood [49] Parallel electrodes WBC Comparison between capillary 7% and venous blood counts [73] Optical Enumeration CD4+ Enumeration of CD4+ T-cells 5.1% in Tanzanian HIV-infected patients [74] Fluorescent Antibodies CD4+ Microchip CD4 counting for 2.7% HIV Monitoring [75] Microfluidic ELISA CD4+ ’Moving the substrate’ ELISA 6.1% [76]
Table 4.3: Comparative study for Bland Altman parameters between existing microfluidic enu- meration techniques. which can be calculated using the approximated flow rate and elapsed time. Both fluidic and electrical simulations rely on the prior knowledge fluid viscosities. Although blood viscosity can be approximated to be 5.3 Pa·s at room temperature (20oC), it is highly dependent on sample temperature[64] and can contribute to errors in flow approximation. Cell loss from sedimentation and unwanted chemical lysis can contribute to undercounting. Particularly in the analysis of WBC concentrations, the negative bias is indicative of a systematic trend. The combined effect of using aligned measurements (see Fig. 2.9) and a relatively short constriction channel limits the maximum flow rate on device which can promote cell sedimentation if cell velocities are too low [77, 78]. Cell settlement in tubular connections that affect other similar microfluidics is lessened here by having on-chip reservoirs. Strong linear correlations between device and reference results indicate the use of a correction factor or internal calibration can be used to account for cell loss from both sedimentation and chemical procedures.
From the Bland Altman plot in Fig. 4.10(B) there appears to be a negative sloped skew for WBC differential measurements. This deviation becomes most apparent in abnormally low or high lymphocyte percentages. Higher than expected lymphocyte percentages could be the consequence of lysis resistant RBCs being mistakenly enumerated as a small WBC [79].
The discrepancy between measurement techniques was consistently positive for low-normal lymphocyte percentages. Lymphocytes have been shown to slightly increase in average diameter Chapter 4. Experimental Results and Discussion 51 while certain non-lymphocytes (neutrophils) have decreased in diameter following treatment with saponin [39]. The inherent overlap in diameters of non-lymphocytes and lymphocytes are problematic in size differentiation. Either the use of additional intracellular information or alternative lysis solutions such as NH4Cl can alleviate these effects. Our difficulty in applying higher frequencies restricts our differential ability to perform 2-
WBC categorization. PDMS immediately adjacent to the constriction channel contribute a small parallel capacitance to the measurement circuit and limits our usable frequencies to the sub-MHz range. For frequencies in the MHz range, the generated size histogram from WBC analysis converges towards a single peak distribution, as in, all cells exhibit similar electrical characteristics due to the loss of measurement sensitivity. Irrespective of these difficulties, the device shows strong linear correlation for WBC differential independent of initial lymphocyte or total WBC concentrations. Chapter 5
Conclusion
This thesis reported a fully monolithic microfluidic system for sequentially performing dilution and lysis of whole blood, electrical enumeration and white blood cell differential. Guidelines for both filter and channel design can be broadly applicable and can extend constriction channel lifetimes enabling the characterization of upwards of 10,000 cells per device. As a demonstra- tion of this technology, 10 blood samples from healthy patients were fluidically processed on device before electrical analysis was conducted to produce various blood cell concentrations and parameters. Comparative results between reference hematology analyzers and the microfluidic device for specific RBC indices, cell concentrations and differentials show strong correlation and support for the ability to perform accurate complete blood count. Additionally, experimental evidence proves the device can operate with use of a single pneumatic pressure source and mixing ratios can be established independent of applied pressure. More generally, this platform can be used to handle solutions of varied viscosities and mixing ratios and can provide electrical analysis of single blood cells in a portable and miniaturized form factor.
52 Chapter 6
Future Directions
The following are examples of future works that can be undertaken:
• On-chip integration of temperature control. This can reduce temperature variability for
both blood and solutions samples. Blood viscosity can greatly vary with temperature,
while RBC lysis has been known to be most effective at room temperature.
• Using the existing platform to enumerate/target rare cell groups (e.g., CTCs). Existing
CTC targeting microfluidic platforms often require RBC lysis.
• Development of a manifold to seal the microfluidic device and provide additional fluidic
reservoirs for extended experiment durations. Sample and solution loading can be com-
pletely automated with use of off-chip fluidic valving and manifold. This can provide a
better world-to-chip interface which is better suited for remote applications
• Use of a different device building material. The device currently suffers from electrical
permeation at higher frequencies. A different material better suited for high frequency
measurements would enable more interrogation opportunities such as measuring cell per-
mittivity and conductivity.
• Miniaturize the existing 15×15 µm2 measurement channel to 5×5 for RBC enumera-
tion. This still allows for RBC enumeration since RBCs can easily deform and squeeze
past. However this allows for an additional advantage of being able to enumerate and
characterize the smallest blood cell group: platelets.
53 Chapter 6. Future Directions 54
• Development of additional parallel modules to further increase the capabilities of the
system. Typical complete blood counts give chemical analysis yielding hemoglobin con-
centration. Bibliography
[1] Martin Shapiro and Sheldon Greenfield. The Complete Blood Count and Leukocyte Dif- ferential Count: An Approach to Their Rational Application. Annals of Internal Medicine, 106(17):65–74, 1987.
[2] Gurpreet Dhaliwal, Patricia Cornett, and Lawrence Tierney. Hemolytic Anemia. American Family Physician, 69(11):2599–2606, 2004.
[3] Beverly George-Gay and Katherine Parker. Understanding the complete blood count with differential. Journal of Perianesthesia Nursing, 18(2):96–117, April 2003.
[4] Tait D Shanafelt and Timothy G Call. Current approach to diagnosis and management of chronic lymphocytic leukemia. Mayo Clinic proceedings, 79(3):388–98, March 2004.
[5] Michael B Streiff, Brad Smith, and Jerry L Spivak. The diagnosis and management of poly- cythemia vera in the era since the Polycythemia Vera Study Group: a survey of American Society of Hematology members’ practice patterns. Blood, 99(4):1144–9, February 2002.
[6] P Lutz and W H Dzik. Large-volume hemocytometer chamber for accurate counting of white cells (WBCs) in WBC-reduced platelets: validation. pages 409–412, 1993.
[7] R Siekmeier, a Bierlich, and W Jaross. The white blood cell differential: three methods compared. Clinical chemistry and laboratory medicine : CCLM / FESCC, 39(5):432–45, May 2001.
[8] Keiji Fujimoto. Principles of Measurement in Hematology Analyzers Manufactured by Sysmex Corporation. Sysmex Journal International, 9(1):31–44, 1999.
[9] Daniel Heikali and Dino Di Carlo. A Niche for Microfluidics in Portable Hematology Analyzers. Journal of the Association for Laboratory Automation, 15(4):319–328, August 2010.
[10] David W Inglis, Megan Lord, and Robert E Nordon. Scaling deterministic lateral displace- ment arrays for high throughput and dilution-free enrichment of leukocytes. Journal of Micromechanics and Microengineering, 21(5):054024, May 2011.
[11] Lidan Wu, Guofeng Guan, Han Wei Hou, Ali Asgar S Bhagat, and Jongyoon Han. Separa- tion of leukocytes from blood using spiral channel with trapezoid cross-section. Analytical chemistry, 84(21):9324–31, November 2012.
[12] Huibin Wei, Bor-han Chueh, Huiling Wu, Eric W Hall, Cheuk-wing Li, Romana Schirhagl, Jin-Ming Lin, and Richard N Zare. Particle sorting using a porous membrane in a mi- crofluidic device. Lab on a chip, 11(2):238–45, January 2011.
55 Bibliography 56
[13] Siyang Zheng, Jing-quan Liu, and Yu-chong Tai. Streamline-Based Microfluidic Devices for Erythrocytes and Leukocytes Separation. Journal of Microelectromechanical Systems, 17(4):1029–1038, August 2008.
[14] Hong Miao Ji, Victor Samper, Yu Chen, Chew Kiat Heng, Tit Meng Lim, and Levent Yobas. Silicon-based microfilters for whole blood cell separation. Biomedical microdevices, 10(2):251–7, April 2008.
[15] Xing Chen, Dafu Cui, and Jian Chen. Design, fabrication and characterization of nano-filters in silicon microfluidic channels based on MEMS technology. Electrophoresis, 30(18):3168–73, September 2009.
[16] Jafar Alvankarian, Alireza Bahadorimehr, and Burhanuddin Yeop Majlis. A pillar-based microfilter for isolation of white blood cells on elastomeric substrate. Biomicrofluidics, 7(1):014102, 2013.
[17] X Chen, D Cui, C Liu, and H Li. Microfluidic chip for blood cell separation and collection based on crossflow filtration. Sensors and Actuators B: Chemical, 130(1):216–221, March 2008.
[18] Myounggon Kim, Seung Mo Jung, Kyeong-Hwan Lee, Yang Jun Kang, and Sung Yang. A microfluidic device for continuous white blood cell separation and lysis from whole blood. Artificial organs, 34(11):996–1002, November 2010.
[19] Miyuki Matsuda, Masumi Yamada, and Minoru Seki. Blood cell classification utilizing hydrodynamic filtration. Electronics and Communications in Japan, 94(1):1–6, January 2011.
[20] Claudio Cupelli, Thorsten Borchardt, Thomas Steiner, Nils Paust, Roland Zengerle, and Mark Santer. Leukocyte enrichment based on a modified pinched flow fractionation ap- proach. Microfluidics and Nanofluidics, 14(3-4):551–563, November 2012.
[21] David W Inglis, John a Davis, Robert H Austin, and James C Sturm. Critical particle size for fractionation by deterministic lateral displacement. Lab on a chip, 6(5):655–8, May 2006.
[22] John a Davis, David W Inglis, Keith J Morton, David a Lawrence, Lotien R Huang, Stephen Y Chou, James C Sturm, and Robert H Austin. Deterministic hydrodynamics: taking blood apart. Proceedings of the National Academy of Sciences of the United States of America, 103(40):14779–84, October 2006.
[23] J.H. Nieuwenhuis, F. Kohl, J. Bastemeijer, P.M. Sarro, and M.J. Vellekoop. Integrated Coulter counter based on 2-dimensional liquid aperture control. Sensors and Actuators B: Chemical, 102(1):44–50, September 2004.
[24] Jiang Zhe, Ashish Jagtiani, Prashanta Dutta, Jun Hu, and Joan Carletta. A microma- chined high throughput Coulter counter for bioparticle detection and counting. Journal of Micromechanics and Microengineering, 17(2):304–313, February 2007.
[25] M Waleed Shinwari, David Zhitomirsky, Imran a Deen, P R Selvaganapathy, M Jamal Deen, and D Landheer. Microfabricated reference electrodes and their biosensing applica- tions. Sensors (Basel, Switzerland), 10(3):1679–715, January 2010. Bibliography 57
[26] Jianhua Zhou, Kangning Ren, Yizhe Zheng, Jing Su, Yihua Zhao, Declan Ryan, and Hongkai Wu. Fabrication of a microfluidic Ag/AgCl reference electrode and its application for portable and disposable electrochemical microchips. Electrophoresis, 31(18):3083–9, September 2010.
[27] Yi Zheng, Ehsan Shojaei-Baghini, Azar Azad, Chen Wang, and Yu Sun. High-throughput biophysical measurement of human red blood cells. Lab on a chip, 12(14):2560–7, July 2012.
[28] Yi Zheng, John Nguyen, Chen Wang, and Yu Sun. Electrical measurement of red blood cell deformability on a microfluidic device. Lab on a chip, 13(16):3275–83, August 2013.
[29] Yi Zheng, Ehsan Shojaei-Baghini, Chen Wang, and Yu Sun. Microfluidic characterization of specific membrane capacitance and cytoplasm conductivity of singlecells. Biosensors & bioelectronics, 42:496–502, April 2013.
[30] Honggu Chun, Taek Dong Chung, and Hee Chan Kim. Cytometry and velocimetry on a microfluidic chip using polyelectrolytic salt bridges. Analytical chemistry, 77(8):2490–5, April 2005.
[31] Honggu Chun, Hee Chan Kim, and Taek Dong Chung. Ultrafast active mixer using poly- electrolytic ion extractor. Lab on a chip, 8(5):764–71, May 2008.
[32] Kwang Bok Kim, Honggu Chun, Hee Chan Kim, and Taek Dong Chung. Red blood cell quantification microfluidic chip using polyelectrolytic gel electrodes. Electrophoresis, 30(9):1464–9, May 2009.
[33] Hyoungseon Choi, Kwang Bok Kim, Chang Su Jeon, Inseong Hwang, Saram Lee, Hark Kyun Kim, Hee Chan Kim, and Taek Dong Chung. A label-free DC impedance- based microcytometer for circulating rare cancer cell counting. Lab on a chip, 13(5):970–7, February 2013.
[34] R. W. DeBlois. Counting and Sizing of Submicron Particles by the Resistive Pulse Tech- nique. Review of Scientific Instruments, 41(7):909, 1970.
[35] O. a. Saleh and L. L. Sohn. Quantitative sensing of nanoscale colloids using a microchip Coulter counter. Review of Scientific Instruments, 72(12):4449, 2001.
[36] S Gawad, L Schild, and P H Renaud. Micromachined impedance spectroscopy flow cy- tometer for cell analysis and particle sizing. Lab on a chip, 1(1):76–82, September 2001.
[37] Karen Cheung, Shady Gawad, and Philippe Renaud. Impedance spectroscopy flow cy- tometry: on-chip label-free cell differentiation. Cytometry. Part A : the journal of the International Society for Analytical Cytology, 65(2):124–32, June 2005.
[38] G Benazzi, D Holmes, T Sun, M C Mowlem, and H Morgan. Discrimination and analysis of phytoplankton using a microfluidic cytometer. IET nanobiotechnology / IET, 1(6):94–101, December 2007.
[39] David Holmes, David Pettigrew, Christian H Reccius, James D Gwyer, Cees van Berkel, Judith Holloway, Donna E Davies, and Hywel Morgan. Leukocyte analysis and differ- entiation using high speed microfluidic single cell impedance cytometry. Lab on a chip, 9(20):2881–9, October 2009. Bibliography 58
[40] Mikael Evander, Antonio J Ricco, John Morser, Gregory T a Kovacs, Lawrence L K Leung, and Laurent Giovangrandi. Microfluidic impedance cytometer for platelet analysis. Lab on a chip, 13(4):722–9, January 2013.
[41] Choong Kim, Kangsun Lee, Jong Hyun Kim, Kyeong Sik Shin, Kyu-Jung Lee, Tae Song Kim, and Ji Yoon Kang. A serial dilution microfluidic device using a ladder network generating logarithmic or linear concentrations. Lab on a chip, 8(3):473–9, March 2008.
[42] Koji Hattori, Shinji Sugiura, and Toshiyuki Kanamori. Generation of arbitrary monotonic concentration profiles by a serial dilution microfluidic network composed of microchannels with a high fluidic-resistance ratio. Lab on a chip, 9(12):1763–72, June 2009.
[43] Daniel Irimia, Dan a Geba, and Mehmet Toner. Universal microfluidic gradient generator. Analytical chemistry, 78(10):3472–7, May 2006.
[44] Glenn M Walker, Jiqing Sai, Ann Richmond, Mark Stremler, Chang Y Chung, and John P Wikswo. Effects of flow and diffusion on chemotaxis studies in a microfabricated gradient generator. Lab on a chip, 5(6):611–8, June 2005.
[45] Xiaojun Han, Cees van Berkel, James Gwyer, Lorenzo Capretto, and Hywel Morgan. Mi- crofluidic lysis of human blood for leukocyte analysis using single cell impedance cytometry. Analytical chemistry, 84(2):1070–5, January 2012.
[46] Cees van Berkel, James D Gwyer, Steve Deane, Nicolas G Green, Nicolas Green, Judith Holloway, Veronica Hollis, and Hywel Morgan. Integrated systems for rapid point of care (PoC) blood cell analysis. Lab on a chip, 11(7):1249–55, April 2011.
[47] Bjorn Ekberg, Ulrik Larsen, and Niels Fogh-Andersen. A Real Point-of-Care System for Complete Blood Counting. Point of Care, 4(1):64–65, 2005.
[48] Palaniappan Sethu, Melis Anahtar, Lyle L Moldawer, Ronald G Tompkins, and Mehmet Toner. Continuous flow microfluidic device for rapid erythrocyte lysis. Analytical chemistry, 76(21):6247–53, November 2004.
[49] N. N. Watkins, U. Hassan, G. Damhorst, H. Ni, A. Vaid, W. Rodriguez, and R. Bashir. Microfluidic CD4+ and CD8+ T Lymphocyte Counters for Point-of-Care HIV Diagnostics Using Whole Blood. Science Translational Medicine, 5(214):214ra170–214ra170, December 2013.
[50] U. Hassan, N. N. Watkins, C. Edwards, and R. Bashir. Flow metering characterization within an electrical cell counting microfluidic device. Lab on a Chip, 14(8):1469, 2014.
[51] Nicholas N. Watkins, Supriya Sridhar, Xuanhong Cheng, Grace D. Chen, Mehmet Toner, William Rodriguez, and Rashid Bashir. A microfabricated electrical differential counter for the selective enumeration of CD4+ T lymphocytes. Lab on a Chip, 11(8):1437, 2011.
[52] Zhe Mei, Sung Hwan Cho, Arthur Zhang, Jie Dai, Tsung-Feng Wu, and Yu-Hwa Lo. Count- ing leukocytes from whole blood using a lab-on-a-chip Coulter counter. Conference proceed- ings : ... Annual International Conference of the IEEE Engineering in Medicine and Biol- ogy Society. IEEE Engineering in Medicine and Biology Society. Conference, 2012:6277–80, January 2012. Bibliography 59
[53] Wendian Shi, Luke Guo, Harvey Kasdan, and Yu-Chong Tai. Four-part leukocyte differ- ential count based on sheathless microflow cytometer and fluorescent dye assay. Lab on a chip, 13(7):1257–65, April 2013.
[54] Yi Li, Carole Young, Timothy J Fischer, James H Carter, and Sergio C Veulens. Reagent and method for differential determination of leukocytes in blood, 1998.
[55] Yi Zheng, Jun Chen, Tony Cui, Nadine Shehata, Chen Wang, and Yu Sun. Charac- terization of red blood cell deformability change during blood storage. Lab on a chip, 14(3):577–83, February 2014.
[56] Yi Zheng, John Nguyen, Yuan Wei, and Yu Sun. Recent advances in microfluidic techniques for single-cell biophysical characterization. Lab on a chip, 13(13):2464–83, July 2013.
[57] Jian Chen, Yi Zheng, Qingyuan Tan, Yan Liang Zhang, Jason Li, William R Geddie, Michael a S Jewett, and Yu Sun. A microfluidic device for simultaneous electrical and mechanical measurements on single cells. Biomicrofluidics, 5(1):14113, January 2011.
[58] Barbara Young, James Lowe, Alan Stevens, and John Heath. Wheater’s Functional His- tology: A Text and Colour Atlas, volume 14. Elsevier Health Sciences, 5 edition, August 2006.
[59] Mary Louise Turgeon. Clinical Hematology Theory and Procedures. Lippincott Williams & Wilkins, 5 edition, 2011.
[60] Kwang W Oh, Kangsun Lee, Byungwook Ahn, and Edward P Furlani. Design of pressure- driven microfluidic networks using electric circuit analogy. Lab on a chip, 12(3):515–45, February 2012.
[61] Christopher J. Morris and Fred K. Forster. Oscillatory flow in microchannels. Experiments in Fluids, 36(6):928–937, June 2004.
[62] M V Kameneva, M J Watach, and H S Borovetz. Gender difference in rheologic properties of blood and risk of cardiovascular diseases. Clinical hemorheology and microcirculation, 21(3-4):357–63, January 1999.
[63] Jitae Kim, Horacio Kido, Roger H. Rangel, and Marc J. Madou. Passive flow switch- ing valves on a centrifugal microfluidic platform. Sensors and Actuators B: Chemical, 128(2):613–621, January 2008.
[64] Tomohiro Tsuji, Takashi Sakata, Yukio Hamaguchi, Fu-sheng Wang, and Berend Houwen. New Rapid Flow Cytometric Method for the Enumeration of Nucleated Red Blood Cells. Cytometry, 37:291–301, 1999.
[65] Jingdong Chen, Di Chen, Tao Yuan, Xiang Chen, Yao Xie, Hualin Fu, Daxiang Cui, Xudong Fan, and Maung Kyaw Khaing Oo. Blood plasma separation microfluidic chip with gradual filtration. Microelectronic Engineering, 128:36–41, October 2014.
[66] Andrea Carbonaro, Swomitra K Mohanty, Haiyan Huang, Lucy a Godley, and Lydia L Sohn. Cell characterization using a protein-functionalized pore. Lab on a chip, 8(9):1478– 85, September 2008. Bibliography 60
[67] Adrian T O’Neill, Nancy a Monteiro-Riviere, and Glenn M Walker. Microfabricated cur- tains for controlled cell seeding in high throughput microfluidic systems. Lab on a chip, 9(12):1756–62, June 2009. [68] Yun-Hsuan Chang, Chao-Jyun Huang, and Gwo-Bin Lee. A tunable microfluidic-based filter modulated by pneumatic pressure for separation of blood cells. Microfluidics and Nanofluidics, 12(1-4):85–94, July 2011. [69] Song-Bin Huang, Min-Hsien Wu, and Gwo-Bin Lee. A tunable micro filter modulated by pneumatic pressure for cell separation. Sensors and Actuators B: Chemical, 142(1):389– 399, October 2009. [70] Sungyoung Choi, Myung Gwon Lee, and Je-Kyun Park. Microfluidic parallel circuit for measurement of hydraulic resistance. Biomicrofluidics, 4(3), January 2010. [71] Jonathan a. Cooper and Richard G. Compton. Channel Electrodes A Review. Electro- analysis, 10(3):141–155, March 1998. [72] Sysmex Xtra. WBC Histogram Interpretations of 3-Part Differentiation. Technical Report July, 2011. [73] Veronica S Hollis, Judith a Holloway, Scott Harris, Daniel Spencer, Cees van Berkel, and Hywel Morgan. Comparison of venous and capillary differential leukocyte counts using a standard hematology analyzer and a novel microfluidic impedance cytometer. PloS one, 7(9):e43702, January 2012. [74] SangJun Moon, Umut Atakan Gurkan, Jeffrey Blander, Wafaie W Fawzi, Said Aboud, Ferdinand Mugusi, Daniel R Kuritzkes, and Utkan Demirci. Enumeration of CD4+ T-cells using a portable microchip count platform in Tanzanian HIV-infected patients. PloS one, 6(7):e21409, January 2011. [75] William R Rodriguez, Nicolaos Christodoulides, Pierre N Floriano, Susan Graham, Sang- hamitra Mohanty, Meredith Dixon, Mina Hsiang, Trevor Peter, Shabnam Zavahir, Ibou Thior, Dwight Romanovicz, Bruce Bernard, Adrian P Goodey, Bruce D Walker, and John T Mcdevitt. A Microchip CD4 Counting Method for HIV Monitoring in Resource- Poor Settings. PLoS MEDICINE, 2(7):e812, 2005. [76] Shuqi Wang, Savas Tasoglu, Paul Z Chen, Michael Chen, Ragip Akbas, Sonya Wach, Cenk Ibrahim Ozdemir, Umut Atakan Gurkan, Francoise F Giguel, and Daniel R Kuritzkes. Micro-a-fluidics ELISA for Rapid CD4 Cell Count at the Point-of-Care. Nature Scientific Reports, 4(3796):1–9, 2014. [77] Jay Warrick, Ben Casavant, Megan Frisk, and David Beebe. A Microfluidic Cell Concen- trator. Analytical chemistry, 82(19):8320–8326, 2011. [78] Yi Zhang and Sriram Neelamegham. Estimating the efficiency of cell capture and arrest in flow chambers: study of neutrophil binding via E-selectin and ICAM-1. Biophysical journal, 83(4):1934–52, October 2002. [79] M Zandecki, F Genevieve, J Gerard, and a Godon. Spurious counts and spurious results on haematology analysers: a review. Part II: white blood cells, red blood cells, haemoglobin, red cell indices and reticulocytes. International journal of laboratory hematology, 29(1):21– 41, February 2007. Appendix A
Circuit Simulation
61 Appendix A. Circuit Simulation 62 Figure A.1: Spicedilution. Simulation used to confirm circuit analysis techniques. (Top) WBC analysis module. (Bottom) RBC analysis and Appendix B
Previous Revisions of Device Design
63 Appendix B. Previous Revisions of Device Design 64 C AB Figure B.1: Photolithographic1a masks and for 1b variousdifferences. with iterations a (C) of shows single microfluidic version dilution device 2 module used where both for for viscosity complete RBC and blood analysis. dilution count. differences Fluidic are (A-B) better resistors handled. show for version blood were experimentally too long due to viscosity Appendix B. Previous Revisions of Device Design 65
A Sample Analysis Preparation Filtration B C
D E
F G
Figure B.2: (A) Flow diagram for sample flow. In order, sample flow is subject to sample preparation, filtration and analysis. Photolithographic masks for various filter configuration designs: (B) Line, (C) square wave, (D) line with triangular channel profile, (E) offset pillars, (F) geometric pattern and (G) offset - rotationally symmetric. Appendix B. Previous Revisions of Device Design 66
B CD
A B
C D
Figure B.3: Device operation of previous version of microfluidic device. (A) 2D schematic of previous version. (B) Whole blood flowing through fluidic resistor. (C) Blood merging with lysis stream for RBC lysis. (D) Blood lysate flowing through long mixing serpentine channel.