Grip Exertion Measurement System (GEMS)

SANJAY HEGDE†, AMIT KAMAT† RAHUL MISHRA†, HEEMESH SETH‡

† DEPARTMENT OF BIOMEDICAL ENGINEERING

‡ DEPARTMENT OF ELECTRICAL ENGINEERING

VANDERBILT UNIVERSITY SCHOOL OF ENGINEERING

APRIL 27, 2004

ADVISOR: MARK RICHTER BENEFICIAL DESIGNS

INSTRUCTOR: DR. PAUL KING ASSOCIATE PROFESSOR OF BIOMEDICAL AND MECHANICAL ENGINEERING VANDERBILT UNIVERSITY SCHOOL OF ENGINEERING TABLE OF CONTENTS

3 1. ABSTRACT……………………………………………………………………………….…..... 4 2. INTRODUCTION………………………………………………………………………………. 3. METHODOLOGY 6 3.1. INITIAL WORK…………..……………………………………………………………….. 7 3.2. PROJECT CIRCUITRY…………………………………………………………………….. 8 3.3. CIRCUIT FABRICATION……………..………………………………………………….. 9 3.4. PROCEDURE AND DATA COLLECTION…………………………….…..…………… 4. RESULTS 10 4.1. DATA……………………………………………………………………..…….…………. 13 4.2. ANALYSIS……….…………………………………….………….………………………. 14 4.3. MISCELLANEOUS CONSIDERATIONS……………………………….….…………… 15 5. SAFETY ISSUES……………………………………………………….……………….……… 6. MARKETING AND ECONOMICS 16 6.1. ECONOMICS…..…………………………………………………………………………. 16 6.2. MARKETING……………………………………………………………………………… 18 7. CONCLUSIONS…………………………………………………………….………………….. 19 8. RECOMMENDATIONS………………………………………………….…………………… 20 9. REFERENCES…………………………………………………………...…………………….

APPENDICES 21 1. CALIBRATION CURVES…………..……………………………………………………… 22 2. LABVIEW EXPLANATION………………………………………………...…………….. 23 3. MATLAB CODE AND EXPLANATION…………………………………………………. 24 4. GRAPHICAL OUTPUT…………………………………………………...……………….. 30 5. DESIGNSAFE REPORT…………………………………………………...……………….

2 Abstract

There is a high incidence of upper-extremity injury associated with manual wheelchair users

(MWUs). Additionally, there is a high incidence rate of carpal tunnel syndrome (CTS) with a prevalence of between 49% and 73%. Some methods consist of computational modeling, qualitative analysis, and quantitative analysis. Computational modeling has proven to be effective yet is not completely accurate due to the fact biomechanics literature argue that the wrist should not be modeled as a fixed joint based on the fact the joint center migrates in proximity. Qualitative methods have been proven to deliver coarse data in determining where forces are applied on the hand from the hand rim and vice-versa. With quantitative analysis researchers are able to pinpoint with great accuracy what spots on the hand are undergoing the most force at each phase of the three-stage propulsion process.

Hence, our goal was to pursue a quantitative approach in accurate measurement of grip forces applied to a wheelchair hand rim during propulsion. The implemented force glove consisted of CUI sensors model IESF-R-5L.

A CUI, Inc. force sensor, which is a passive resistive circuit element, was used in the model and the force applied resulted in lower resistance. We arranged the sensors to where we felt were strong contact points between the hand rim and the user. The eight leads were then encased in plastic wire wrap and attached to a breadboard for data collection. With the use of the LabView virtual instrument (VI) and a MATLAB script data was outputted for further analysis. Quantitative analysis has proven to be very effective in determining the key locations on the hand that exert forces on the hand rim and vice-versa. The data gathered from our initial trials have shown that the ring finger, middle palm, and index finger undergo extensive force in relation to the other locations on the hand.

3 Introduction

There is a high incidence of upper-extremity injury associated with manual wheelchair users

(MWUs). The site predominately affected of musculoskeletal injury is the shoulder (Rao152). More specifically surveys have shown that 31% to 73% MWUs have experienced shoulder pain (Boninger et al. 718). Additionally, there is a high incidence rate of carpal tunnel syndrome (CTS) with a prevalence of between 49% and 73% (Boninger et al. 718). Based on these surveys and various clinical and research case studies there is a strong correlation between activities of daily living, such as propelling and moving to and from a wheelchair, to repetitive strain injuries that cause pain (Boninger et al. 718, Shimada 275). It has been suggested that the repetitive actions in propelling a wheelchair contributes to the effects of CTS (Armstrong 830, Burnham 513, Shimada 275). From this, researchers have pursued various methods of determining what causes the various pains at the systemic level up to the shoulder joint.

Various methods of determining the cause of CTS and shoulder injuries specific to wheelchair use have been devised. Some methods consist of computational modeling, qualitative analysis, and quantitative analysis. Computational modeling has been of great interest to many researchers due to the associated benefits of dynamically applying various model approaches with very little difficulty or cost. Many models have been devised with very similar approaches and concepts. The two major models that have been proposed with entirely different approaches: 1) Model the wrist with individual carpal bones and 2) model the wrist with the movement of the hand in relation to the forearm (Shimada

4 275). Computational modeling has proven to be effective yet is not completely accurate due to the fact biomechanics literature argue that the wrist should not be modeled as a fixed joint based on the fact the joint center migrates in proximity (Shimada 275, Veeger 305). Based on this known problem other methods of determining the cause of CTS and shoulder injuries have been devised.

Qualitative methods have been proven to deliver coarse data in determining where forces are applied on the hand from the hand rim and vice-versa. Some methods consist of video capture, pressure paper, and ink analysis (Davis 61). With qualitative data coupled with a strong understanding of hand physiology one could hypothesize and test what particular muscle groups and areas on the hand undergo various wheelchair activity forces. From this, a correlation could be made to determine the cause of CTS and shoulder injuries that have been associated to MWUs. Another method that is slow emerging in determining the cause of CTS and shoulder injuries is quantitative methods.

Various devices have been devised to measure quantitative force or pressure data during wheelchair use. Many of these particular devices come in various forms from load cell to strain gauge based instruments. With quantitative data researchers are able to pinpoint with great accuracy what spots on the hand are undergoing the most force at each phase of the three-stage propulsion process.

From this, a quantitative proves to be promising in determining the cause of CTS and shoulder injuries.

Our goal was to pursue a quantitative approach in accurate measurement of grip forces applied to a wheelchair hand rim during propulsion. This task involved a step-by-step methodology process that involved intensive background paper research, qualitative testing, fabrication of a glove with force sensors, and acquisition of voltage correlated force data. The implemented force glove consisted of

CUI sensors model IESF-R-5L. This fabricated glove aided in the collection of various voltage data, which correlated to a specific calibration curve to determine force values. Further analysis with this

5 force data helped to determine what parts of the hand and fingers were involved in the three stage propulsion process.

Methodology

3.1 Initial Work

We originally purchased eight Motorola MPX2050 pressure sensors for use with our glove. Based on our research, we had concluded that the sensors would meet our budget constraints as well as provide a linear output voltage directly proportional to increasing pressure. Unfortunately, we found that the MPX2050 necessitated a fluid medium, e.g. air. After consulting with a Motorola sales representative, we concluded that these types of sensors are better suited for other biomedical applications, such as pump controllers.

A more thorough search of force sensors was conducted, this time with more consideration given to the operation medium. This time, however, we perused Honeywell’s semiconductor catalog and we discovered two robust, horizontally fitting force sensors models called the FSS and FS01.

Finding a vendor for these sensors, however, proved to be difficult, reason being that the FSS and

FS01 were not readily in stock. After consultation with Carlton-Bates, a Honeywell distributor, we

6 were informed that the sensors needed to be ordered, and fabrication from the factory would take eight to ten weeks.

As the possibilities of finding a suitable, compact force sensor seemed to be diminishing, we were referred to a company known as CUI Incorporated by our main distributor, Digi-Key. We

contacted Vice President James Seller, who

graciously provided us with four free samples

of their IESF force sensor line. We settled

upon the IESF-R-5 force sensor, since it was

capable of recording up to 4 kilograms of force

(which was within our designed range) as well

Figure 2: Resistance v. Force relationship. as economical. The compact device outputs a varying resistance that is proportional to the force applied to the back button at the end of the device.

A schematic of the force sensor is included in (Figure 1).

3.2 Project Circuitry

The SF sensor is a passive device. The basis of the sensor is a carbon-impregnated elastomer,

covered with a rubber membrane that deflects under

pressure. As force is applied to the elastomer the Figure 1: Sensor schematic. carbon molecules are squeezed into closer proximity and the result is lower resistance (CUI, online). As such, it acts as a variable resistor where the resistance decreases as force is applied. A graph of the output resistance versus force applied is provided in (Figure 2) as well as a table of operational parameters (Figure 3).

7 Preliminary tests of the sensors showed that the baseline resistance at zero kilograms of force was upwards of 1-MΩ. We realized that should a voltage drop be introduced across the sensor, there would be a negligible amount of current passing through during baseline resistance. When we

designed our circuit, we exploited this fact by placing the IESF-R-5

sensor right before our lead-out voltage (VI), such that it would be

zero at baseline and increase as pressure is introduced. The

schematic of the circuit is shown in Figure 4 -circuit. In series with

the IESF-R-5 is an ultra-bright green LED, allowing us to test its

functionality at any given time. The voltage output lead was then

 9(R  R ) connected to the National Instruments VI box for V  G 1 1 R  R Figure 4: Circuit schematic. G 1 data collection.

Node-voltage analysis of the circuit reveals that the output voltage (V1) is equal to

(1)

In Equation 1, it is evident that as RG approaches

infinite, the output voltage V1 effectively drops to

zero. From a design standpoint, it was up to us to

decide the value of the resistor R1, which would be

Figure 3: Operational parameters and specifications. crucial to determining the overall sensitivity of the circuit. As Figure 2 attests to, the relationship between resistance and force is not linear. Rather, it exhibits an exponential decay that is certainly more sensitive to the initial increments of force rather than the latter. That said, we could choose a value of R1 that would place this sensitivity on a particular range of forces that we felt would be of most importance to our design. That is, while the

8 max force rating is 4 kilograms, we found that this level of force was a bit excessive during normal wheelchair propulsion. From our experiments, we found appropriate output voltages when R1 was of a lower value. Subsequently, a 1.2-kΩ resistor was chosen for R1.

Calibration curves of force versus voltage were then constructed for the eight pressure sensors. We had intended to average the eight curves together for a single equation for back calculations of force, but we found the R2 to be too nominal. We Figure 5: Sensor placement on glove. found that the base index and left palm pressure sensors were data outliers and we decided to constructed separate calibration curves for these two, while averaging the other six sensors’ responses together. The three calibration curves are included in Appendix 1.

3.3 Design Fabrication

Before fabrication force glove could begin, we had to decide upon the placement of the IESF-

R-5 force sensors. We carried out a series of ink analysis tests, in which we coated the hand rim with ink and had a group member undergo normal wheelchair propulsion. The individual’s hand was then photographed and we arranged the sensors to where we felt were strong contact points between the hand rim and the user. As Figure 5 attests to, we decided upon placing sensors within the mid-portions of the index, middle and ring fingers, as well as the thumb. Additional sensors were then placed at the base of the index and at the left, middle and right lower portions of the palm.

Fabrication of the glove began by inverting it and creating small incisions for the contact points. The black heads of the pressure sensors were then slipped through, held in place by electrical tape. In order to minimize user discomfort, the leads were then wired along the seams of the glove.

The eight leads were then encased in plastic wire wrap and attached to a breadboard for data collection.

9 3.4 Procedure and Data Collection

The LabView VI was then set to run with eight channels of input from the DAQ board, exporting the data for further analysis in MATLAB. For verification of LabView and MATLAB, please refer to Appendices 2 and 3, respectively. We conducted two sets of data collection, consisting of three trials each. In the first set, we placed the wheelchair upon cinder blocks and recorded force values during normal wheelchair propulsion. For the second set, however, we wanted to mimic wheelchair propulsion during adverse conditions, e.g., propulsion against some kind of resistive element. Thus, we acquired the second set of data with the breaking mechanism partly engaged.

Results

4.1 Data

10 There were three sets of data collected for wheelchair propulsion without resistance and three

sets of data with resistance. Figure 6 contains the summary tables for the overall force data collected

with no resistance on the wheelchair hand rim as a function of time. Trials one and two show that the

sensors for the index finger, middle palm and thumb were the only sources of pressure changes. The

Trial 1 /------\ | Hand Location | Average Force (N) | Time-Average Force (N/s) | Total Time of | | | | |Applied Pressure (sec)| |------| | Ring Finger | NaN NaN 0.00 | | Middle Finger | NaN NaN 0.00 | | Index Finger | 4.40 2.89 1.52 | | Index Base | NaN NaN 0.00 | | Left Palm | NaN NaN 0.00 | | Middle Palm | 9.99 3.31 3.02 | | Right Palm | NaN NaN 0.00 | | Thumb | 10.92 4.51 2.42 | \------/

Trial 2 /------\ | Hand Location | Average Force (N) | Time-Average Force (N/s) | Total Time of | | | | |Applied Pressure (sec)| |------| | Ring Finger | NaN NaN 0.00 | | Middle Finger | NaN NaN 0.00 | | Index Finger | 1.10 2.11 0.52 | | Index Base | NaN NaN 0.00 | | Left Palm | NaN NaN 0.00 | | Middle Palm | 9.31 2.74 3.40 | | Right Palm | NaN NaN 0.00 | | Thumb | 8.26 2.35 3.52 | \------/

Trial 3 /------\ | Hand Location | Average Force (N) | Time-Average Force (N/s) | Total Time of | | | | |Applied Pressure (sec)| |------| | Ring Finger | 0.42 2.11 0.20 | | Middle Finger | NaN NaN 0.00 | | Index Finger | 0.73 3.32 0.22 | | Index Base | NaN NaN 0.00 | | Left Palm | NaN NaN 0.00 | | Middle Palm | 4.74 1.36 3.48 | | Right Palm | 12.55 3.39 3.70 | | Thumb | 3.92 1.22 3.22 | \------/ Figure 6: MATLAB output of force data summary for “No Resistance.” NOTE: ‘NaN’ indicates that no force data was collected.

thumb and middle palm contribute to the majority of the pressure applied, with an approximate range

of 8 to 10 N of average force. The time averaged data indicates that these areas of the hand also

11 remain in contact with the hand rim for a longer period of time, an approximate range of 2.5 to 3.5 seconds. Trial three shows, however, that the right side of the palm and ring finger region contributed to the grip of the hand rim. The right palm registered the highest average force of about 12.5 N for a period of 3.2 seconds. This suggests that pressure applied by other regions of the hand sometimes lead to a decrease in registered force application from the areas. Trial three is indicative of this because as more pressure was produced by the palm, less was emphasized by the fingers.

Figure 7 contains the summary tables for the overall force data collected with resistance. These trials show that with resistance applied to the wheel, additional regions of the hand are used to grip the hand rim, such as the ring finger and right palm regions. Trials one and three produced similar results with respect to the area of the hand that produced pressure. Significantly different data was collected with the thumb again registering the greatest amount of pressure applied to the hand rim, approximately 7 to 15 N. The middle palm was the next highest, with a range of 5 to 12 N. In Trial 2

12 Trial 1 /------\ | Hand Location | Average Force (N) | Time-Average Force (N/s) | Total Time of | | | | |Applied Pressure (sec)| |------| | Ring Finger | 0.46 3.84 0.12 | | Middle Finger | NaN NaN 0.00 | | Index Finger | 2.54 6.36 0.40 | | Index Base | NaN NaN 0.00 | | Left Palm | NaN NaN 0.00 | | Middle Palm | 5.10 4.39 1.16 | | Right Palm | NaN NaN 0.00 | | Thumb | 10.35 6.47 1.60 | \------/

Trial 2 /------\ | Hand Location | Average Force (N) | Time-Average Force (N/s) | Total Time of | | | | |Applied Pressure (sec)| |------| | Ring Finger | 0.26 4.30 0.06 | | Middle Finger | NaN NaN 0.00 | | Index Finger | NaN NaN 0.00 | | Index Base | NaN NaN 0.00 | | Left Palm | NaN NaN 0.00 | | Middle Palm | 6.83 4.17 1.64 | | Right Palm | 9.37 6.99 1.34 | | Thumb | 7.13 4.10 1.74 | \------/

Trial 3 /------\ | Hand Location | Average Force (N) | Time-Average Force (N/s) | Total Time of | | | | |Applied Pressure (sec)| |------| | Ring Finger | 2.29 1.59 1.44 | | Middle Finger | NaN NaN 0.00 | | Index Finger | 0.29 1.62 0.18 | | Index Base | NaN NaN 0.00 | | Left Palm | NaN NaN 0.00 | | Middle Palm | 12.92 7.34 1.76 | | Right Palm | NaN NaN 0.00 | | Thumb | 15.33 8.81 1.74 | \------/ Figure 7: MATLAB output of force data summary for “Resistance.” NOTE: ‘NaN’ indicates that no force data was collected.

the right palm registered approximately 9 N, while the rest of the data closely reflected Trials

one and three. The amount of time the mentioned areas of the hand were in contact with the hand rim

was on average 1.5 seconds, and the time averaged forces ranged from 1.5 N to 8 N.

13 4.2 Analysis

Appendix 4 contains graphs of the relative rise and fall of the forces produced, which are graphical representations of the data described above. The data in Figures 6 and 7 corroborates with our hypothesis that the palm region would on average be exposed to more force than any other parts of the hand. Additionally, the thumb region was consistently an area of force application. The data aligns with the professional opinion of Dr. Weikert. He suggested that the thumb and palm area experience the most pressure because of their placement on the hand. Their degrees of freedom allow the hand to exert large amounts of pressure with the contraction of the thenar muscle group. The graphs and tables also point out that the thumb consistently applied pressure for relatively longer periods of time than other regions of the hand.

Based on the graphical outputs in Appendix 4, stage one, the grip stage, saw a significant pressure increase in the thumb and palm area. It was the latter stages that saw a pressure application from the other areas of the hand, such as the fingers, and middle index area, the interossei muscle group. This data makes sense because the initial pressure applied is in a downward direction that allows the wheelchair user to hold on to the hand rim. This gripping mechanism incorporates a stress, a pushing force, and then shear, an angular force. Thus the pressure changes are usually greater in each region upon first contact on the hand rim then the next change in pressure is not as great. The reason for this is because the force sensor does not detect the angular pressures that are applied to it once the propulsion stage begins. The sensors are designed to detect only pressures perpendicular to its surface. Any force applied at an angle will not be detected, and this could be the cause for loss of data during the acquisition portion.

14 4.3 Miscellaneous Considerations

Based on the hand rim of the wheelchair that was used, we

considered these data to be reasonable. All eight channels are able to

sense pressure changes, but not all registered changes as shown in the

tables (from Figures 6 and 7). The reason for this is due to the design

of the hand rim that was used. The hand rim itself, as depicted in

Figure 8, was a very thin rim as it comes standard on most hospital Figure 8: Experimental wheelchair. wheelchairs. This thin design reduces the chances of making contact

on the pressure sensors; meaning, only a few sensors would be activated when the wheelchair was in

use. If a hand rim based on a wider design was used in the trials, then the pressure data would show

force data from all eight channels. This explains why the bar charts and line graphs show fewer than

eight channel outputs of data.

15 Safety Issues

Our glove-based device is for diagnostic purposes only. Since this glove is intended to be used for educational or research applications there are more safety concerns than during normal operations.

As a diagnostic tool, this device will be used to improve existing designs, and therefore researchers will try to test the wheelchair’s maximum threshold resulting in possible safety concerns.

Using the DesignSafe software our group was able to identify all the safety concerns associated with the use of the glove. We used a task-based method to ensure that all safety concerns will be met.

Most of the safety concerns assessed by DesignSafe were due to not observing the specified limitations and trying to test threshold information. Only one hazard has been identified due to poor design, which is in the process of being corrected for the final model used for industrial purposes. Please see a list of all the safety concerns in Appendix 5.

16 Marketing and Economics

6.1 Economics Analysis

This device will be used as a diagnostic tool for Beneficial Designs and therefore is not intended for sale to the consumer. With this in mind, this glove is designed for Beneficial Designs and can be sold to consumers if they agree to do so. As a diagnostic tool, our group conducted extensive research to create a cost-efficient model that would satisfy the necessary goals. A glove-based design was implemented with force sensors imbedded within the glove. A LabView acquisition system was used to collect the output voltages, after which a MATLAB program was used to analyze the data as well as convert the voltages into force information.

Item Quantity Unit Cost Total Cost Adam Leathers Racing1 $40

glove $40.00 CUI Force Sensor 8 $8.55 $68.40 Serpac M4 Enclosure 1 $5.83 $5.83 Miscellaneous 1 $50.00 $50.00 Total Production Cost $164.23 Table 1: Itemized list for total production cost

6.2 Market Analysis

17 There is a vast consumer market for manual wheelchairs. It is estimated that in 2001 there were approximately 2.2 million wheelchair users (United States Census Bureau, 2001). With the use of our glove sensors, a new wheelchair can be developed that can potentially reduce the amount of

Force Glove Production Cost stain applied to

shoulders and

$40.00 wrists. This new $50.00 Adam Leathers Racing glove type of wheelchair CUI Force Sensor Serpac M4 Enclosure will be able to $5.83 Miscellaneous obtain a $68.40 significant portion

Figure 9: Production cost distribution. of the manual wheelchair annual sales. This shows that there is a high demand to develop a new wheel or hand rim for wheelchairs. In order to create these improvements, a diagnostic tool, such as our glove, can be used to make the necessary adjustments. Since a specialty wheel costs approximately $500 we feel that a strong company could afford to purchase our diagnostic glove for approximately $165.00 per glove prior to patents and intellectual property assignment (Figure 9). This is a very low one-time expenditure that can lead to profits that were previously unavailable. This glove can be sold to large manufacturers as well as small companies that need an innovative way to become competitive in this market. Since the basic wheelchair design has not changed significantly in the last 50 years, it is believed that with the use of this glove a new manual wheelchair can be created that will lead this industry in a new direction, focused on reducing physical strain for quadriplegics. Since Beneficial

Designs is a relatively small corporation, it would be difficult for them to solicit business from a number of customers. Therefore, it is reasonable for them to target manufacturers as well as research

18 institutions that are within the wheelchair field. Since this has strong beneficial aspects for all consumers, we are going to focus on major companies to get this product to be used in mainstream manufacturing markets.

Conclusion

Quantitative analysis has proven to be very effective in determining the key locations on the hand that exert forces on the hand rim and vice-versa. The data gathered from our initial trials have shown that the ring finger, middle palm, and index finger undergo extensive force in relation to the other locations on the hand. More specifically, our analysis has shown that these three locations are more in contact on average versus any other sensor location on the glove.

There are a few ideas that merit special consideration for improving the overall glove design. It could be possible that the implemented wiring scheme, which led to heavy grouping of the sensor leads, could have adversely affected the user’s natural gripping posture. To this effect, normal wheelchair propulsion would be altered since the user would have to compensate for such a bulge or irregularity in gripping formation. In addition, we would like to utilize a thinner glove, placing the sensors in a more form-fitting design, such as neoprene, that would more accurately record pressure values. Of concern was the fact that there was too much cloth in between the user’s hand and the pressure sensor, effectively dampening the force readings.

Despite these shortcomings, we believe that quantitative analysis still has the potential in gathering pinpoint force data, which could be used to determine the biomechanical causes of CTS and upper extremity injuries. From our current standpoint, we believe that this prototype diagnostic tool

19 could be of great use to Beneficial Designs and their pursuit for designing a more ergonomic MWU hand rim.

Recommendations

Engineering projects do not always turn out the way they were initially planned. This was the case for our project mainly due to a lack of resources as well as time. In order to improve our project several minor recommendations could be used to strengthen our data and provide a suitable glove for

Beneficial Designs. The first modification is to purchase a large number of force sensors to cover the entire glove. This will allow us to correctly identify which locations are main pressure points as well as allow the flexibility for different users, since everyone grips the wheelchair hand rim differently.

Another limitation was that the sensors that were purchased could only detect stress forces. In the future, with regards to normal wheelchair operations, it is important to have sensors that could effectively detect stress forces in conjunction with shear stress. Due to the constraint of our wires and the necessity for a computer with LabView, we were unable to test our prototype on a free-motion wheelchair using a treadmill. Instead we had to simulate our results on a stationary wheelchair with the brakes partially engaged to mimic a resistive force an MWU would encounter, such as an incline.

In order to strengthen our results, we will use the force glove system on a treadmill to simulate normal wheelchair operation. This will allow us to continuously collect reproducible measurements, which

20 would aide us in determining the cause of CTS and upper extremity injuries. With these changes we feel that a strong diagnostic tool can be created that will help Beneficial Designs create a new wheelchair to reduce injury.

Works Cited

Armstrong, T.J., (1987) Ergonomics consideration in hand and wrist tendonitis. Journal of Hand Surgeries. [Am.] 12, 830-837.

Boninger, Michael L, MD, Aaron L. Souza, MS, Rory A. Cooper, PhD, Shirley G. Fitzgerald, PhD, Alicia M. Koontz, MS, ATP, and Brian T. Fay, MS. “Propulsion Patterns and Pushrim Biomechanics in Manual Wheelchair Propulsion.” Arch Phys Med Rehabil. 2002: 718-723.

Burnham, Robert. 1994. Acute Median Nerve Dysfunction From Wheelchair Propulsion: The Development of a Model and Study of the Effect of Hand Protection. Arch Phys Med Rehabil, May, 513.

Cobb, Tyson. 1995. Externally Forces to the Palm Increase Carpal Tunnel Pressure. The Journal of Hand Surgery, 4 June, 181.

Davis, Jaime L. 1998. Three-dimensional kinematics of the shoulder complex during wheelchair propulsion: A technical report. Journal of Rehabilitation Research and Development, January, 61

Galloway, Robert Ph.D. Personal Interview. October, December, January.

IESF-R-5XX Specifications. CUI Incorporated. Date Accessed: March 21st, 2004.

Independence Through Enhancement of Medicare and Medicaid Coalition. 23 March 2004 (http://www.shhh.org/html/fact_sheet_on_assistive_techno.HTM). 23 April 2004

21 Kuhn, John M.D. Personal Interview. 29 January 2004.

Rao, S.S. (1996) Three-dimensional kinematics of wheelchair propulsion. IEEE Trans. Rehabilitation Engineering. 4, 152-160.

Shimada, Sean D., Rory A. Cooper, Michael L. Boninger, Alicia M. Koontz, and Thomas A. Corfman. “Comparison of Three Different Models to Represent the Wrist During Wheelchair Propulsion.” IEEE Transactions on Neural Systems and Rehabilitation Engineering. 2001: 274-282.

Veeger, DirkJan. 1998. Wrist motion in handrim wheelchair propulsion. Journal of Rehabilitation Research and Development, 35 July, 305-313.

Veeger, H.E.J. and van der Woude, L.H.V. (1994) Force generation in manual wheelchair propulsion., Washington, D.C.

Weikert, Doug. M.D. Personal Interview. 20 February 2004.

APPENDIX 1: Calibration Curves

22 Voltage v. Force (SIX)

4.5 4 3.5

3 2 ) y = 0.1481x - 0.2979x g 2.5 k

( 2

R = 0.9245 e 2 c r o 1.5 F 1 0.5 0 -0.5 0 1 2 3 4 5 6 7 Voltage (V)

Voltage v. Force (Index Base)

4.5 4 3.5 2 3 y = 0.2178x - 0.1259x

) 2 g 2.5 R = 0.9102 k (

e 2 c r o 1.5 F 1 0.5 0 -0.5 0 1 2 3 4 5 Voltage (V)

23 Voltage v. Force (Left Palm)

4.5 4 3.5 3 y = 0.2648x2 - 0.6419x ) 2.5 g 2 k

( R = 0.95

2 e

c 1.5 r o

F 1 0.5 0 -0.5 0 1 2 3 4 5 6 -1 Voltage (V)

APPENDIX 2: LabView explanation

The software portion included LabView modules for data acquisition and MATLAB code for data analysis. The main LabView module and real time interface are shown here in Figure 1 and Figure 2, respectively, and it is configured to collect signals from the National Instruments DAQ board, which is defined as “Device 1” in the diagram. The “channel” input was configured as “0:7” for the eight BNC connectors stemming from the circuit into the DAQ board. The “AI Config” module takes inputs from the channels and samples data based on the buffer size, sampling rate, and scans per second. The “AI Start” component was set for continuous acquisition until an error occurred or the stop button was pressed on the interface. In our experiments, the buffer size was set at 16384, and the sampling rate and scans per second were set to 50. Once data collection began, the “AI Read” function passed the voltages changes into “My Data Processing” to compose real time graphs as data was collected. This was done so that the changes in voltage data could be confirmed as the trial was being executed.

24 Figure 1: LabView module for data acquisition.

Figure 2: LabView real-time 25 output. APPENDIX 3: MATLAB Code and explanation

[Please refer to following attached pages for the explanation below.] This voltages were saved in a “*.dat” file; this was converted to an excel file and an extra column was added to the voltages listed. This last column in the files was the time array that was used by MATLAB. The first thing the MATLAB code did after reading in the excel file was set a filter for each channel input. The initial reasoning for doing this was to minimize noise and display the peaks for the voltage changes. However, the sensors proved to be quite efficient and produced very little noise. This forced us to set the filters at a very low voltage, and this component of the code was essentially unused. The next step converted the voltage data to force data based on the calibration method described earlier. At this point the matrix ‘fhand’ contains the filtered and converted force data from the data acquisition. The next portion of the code plots individual graphs of the filtered and unfiltered channel inputs. Because of the unused filters, this output was made into an option since the filtered and unfiltered graphs would look relatively the same. We left the code in the file in case it was needed for future experiments where the sensors accumulated noise. The next portion of the code found the start and stop positions of the forces changes for each channel. This was then saved as a separate matrix, ‘marker,’ to calculate the time differentials of pressure application for each channel. The next part of the code calculated the average forces and uses the time periods from ‘marker’ to calculate average timed forces. The remainder graphical outputs contained the overall pressure changes as a function of time and the calculated average forces from each channel. The overall pressure changes were created as line plots while the pressure averages were formed as bar charts. Lastly, a tabulated form of the graphical results is outputted to the main screen for data analysis.

26 APPENDIX 4: Graphical Output

NO Resistance

-Trial 1-

27 28 -Trial 2-

29 30 -Trial 3-

31 Resistance

-Trial 1-

32 33 -Trial 2-

34 35 -Trial 3-

36 37 APPENDIX 5: DesignSafe

38