The Roscan Thermal 3D Body Scanning System: Medical Applicability and Beneﬁts for Unobtrusive Sensing and Objective Diagnosis

sensors

Article The RoScan Thermal 3D Body Scanning System: Medical Applicability and Beneﬁts for Unobtrusive Sensing and Objective Diagnosis

Adam Chromy 1,2,* and Ludek Zalud 1,2 1 CEITEC—Central European Institute of Technology, Brno University of Technology, Purkynova 656/123, 612 00 Brno, Czech Republic; [email protected] 2 Faculty of Electrical Engineering and Communications, Brno University of Technology, Technicka 3082/12, 616 00 Brno, Czech Republic * Correspondence: [email protected]

Received: 19 October 2020; Accepted: 19 November 2020; Published: 20 November 2020

Abstract: The RoScan is a novel, high-accuracy multispectral surface scanning system producing colored 3D models that include a thermal layer. (1) Background: at present, medicine still exhibits a lack of objective diagnostic methods. As many diseases involve thermal changes, thermography may appear to be a convenient technique for the given purpose; however, there are three limiting problems: exact localization, resolution vs. range, and impossibility of quantification. (2) Methods: the basic principles and benefits of the system are described. The procedures rely on a robotic manipulator with multiple sensors to create a multispectral 3D model. Importantly, the structure is robust, scene-independent, and features quantifiable measurement uncertainty; thus, all of the above problems of medical thermography are resolved. (3) Results: the benefits were demonstrated by several pilot case studies: medicament efficacy assessment in dermatology, objective recovery progress assessment in traumatology, applied force quantification in forensic sciences, exact localization of the cause of pain in physiotherapy, objective assessment of atopic dermatitis, and soft tissue volumetric measurements. (4) Conclusion: the RoScan addresses medical quantification, which embodies a frequent problem in several medical sectors, and can deliver new, objective information to improve the quality of healthcare and to eliminate false diagnoses.

Keywords: multimodal imaging; multispectral imaging; 3D thermography; high-accuracy 3D scanning; robotic 3D scanning

1. Introduction In general terms, the effectiveness of treatment, and thus also the recovery time and the amount of resources spent, stands on two prominent pillars: correct diagnosis and appropriate therapy [1]. At present, the latter pillar is comparatively well-characterized thanks to the availability of a wide range of knowledge sources generated through simplified global communication. Recommended therapies are often standardized, with the “best practice” methods systematically presented in research articles. Moreover, and importantly in the given context, almost any diagnosis can be paired with a suitable treatment approach [2]. The pace of progress in the former pillar, however, is markedly different. Although the correct diagnosis is at least as important as proper therapy, many of the current, commonly used diagnostic procedures still rely on healthcare professionals’ experience, intuition, and memory. The correctness of the decision then depends on the specific abilities of the medical specialist [3]. The diagnosis can be very close to the patient’s real condition, but it can also be completely wrong. The variance of the results then embodies a major problem within intuitive medicine.

Sensors 2020, 20, 6656; doi:10.3390/s20226656 www.mdpi.com/journal/sensors Sensors 2020, 20, 6656 2 of 22 Sensors 2020, 20, x FOR PEER REVIEW 2 of 24

To prevent mistakes in determining the correct diagnosis,diagnosis, to decrease the required time, and to reduce the dependence of successful disease classifi classificationcation on the doctor’s expertise and and experience, supporting diagnostic medical devices are introduced increasingly more often.often. Most hospitals are presently equipped with ge general-purposeneral-purpose imaging methods, methods, such such as as X-ray, X-ray, MRI, MRI, and and CT; CT; these these tools, tools, however, may be very expensive and thus less accessible to smaller medical centres. In In such such cases, smaller, cheaper, andand moremore specializedspecialized instrumentsinstruments areare employed.employed. Almost all types of injuries, multiple diseases, and pathological changes share a significantsignificant symptom: the the local local temperature temperature change. This This thermal thermal abnormality abnormality is is caused caused by by increased increased blood blood flow flow and stronger cellular metabolism in the affected affected area; the two aspects then induce a local temperature temperature differencedifference generated by the given negative health-relatedhealth-related eeffectffect [[4].4]. Such local thermal deviations can be detected and visualized via thermal cameras operatingoperating within the long-wave infrared spectrum (LWIR) [[5].5]. As the resolution of thermal imagers is about 0.04 °C◦C[ [6],6], even the smallest changes are detectable, and the disease may be terminated before developing significantlysignificantly enough to produce observable symptoms like redness or swelling (Figure1 1).).

Figure 1. TheThe symptoms symptoms have have not not progressed progressed far enough to be observable ( leftleft),), but but the the inflammation inﬂammation is already clearly visible in the thermal imageimage ((right).).

Medical thermography thermography is is generally generally referred referred to toas asDigital Digital Medical Medical Thermal Thermal Imaging Imaging (DMTI); (DMTI); the techniquethe technique currently currently finds finds use use in inseveral several relevant relevant applications applications (e.g., (e.g., inflamed inflamed tissue tissue analysis analysis [7,8], [7,8], cancer detection [9], [9], blood perfusion [10], [10], and huma humann stress research [11]), [11]), although it has not been developed susufficientlyfficiently to date. The main factors limiting the usability and popularity of this powerful imaging approach consist of three centralcentral drawbacksdrawbacks [[12–16]:12–16]: 1. ExactExact localizationlocalization problem—due problem—due to theto lackthe oflack easily of recognizableeasily recognizable and clearly and bounded clearly landmarksbounded landmarksin the thermal in the image thermal (which image contains (which mostly contains smooth mostly gradients smooth instead gradients of sharp instead edges), of sharp it is edges),not possible it is not to possible determine to determine the exact positionthe exact of position a thermal of a deviation thermal deviation on the skin. on the This skin. issue This is issueillustrated is illustrated in Figure in2, namely,Figure 2, the namely, indicated the spots indicated with aspots locally with lower a locally temperature; lower temperature; however, we however,are presently we are unable presently to accurately unable assignto accurately the individual assign pointsthe individual to exact placespoints onto exact the body, places and on it theis then body, impossible and it is then to distinguish impossible whether to distinguish such thermal whether footprints such thermal relate to,footprints for example, relate an to, area for example,of eczema an or area a birthmark of eczema in or the a vicinity.birthmark in the vicinity. 2. ResolutionResolution vs. range problem—the lowlow resolution resolution of of thermal thermal imagers imagers forces forces the the user user to searchto search for fora compromise a compromise between between two two options: options: (a) (a) awide-range a wide-range view, view, where where the the entire entire part part of of the the body body is isrepresented, represented, but but the the resolution resolution is is too too low low to to show show enough enough detailsdetails (Figure(Figure3 3a);a); and and (b) (b) aa detaileddetailed viewview providing providing sufficient sufficient resolution resolution but but contai containingning a a very very narrow narrow segment segment of of the the body, body, from whichwhich the the context context and and exact exact position position are are difficult difficult to understand (Figure 3 3b).b). AsAs thethe thermogramthermogram analysisanalysis consists predominantlypredominantly of of comparing comparing the the temperature temperature in bodyin body regions, regions, the clinician the clinician must mustsee a relativelysee a relatively large area large of thearea body; of the simultaneously, body; simultaneously, the high resolution the high hasresolution to be preserved has to be to preserved to deliver the required details. Overall, the resolution vs. range problems is one of the most limiting drawbacks restricting the development of DMTI.

Sensors 2020, 20, 6656 3 of 22

deliver the required details. Overall, the resolution vs. range problems is one of the most limiting drawbacks restricting the development of DMTI. 3. Qualitative only, no quantification—although 2D thermography basically constitutes a quantification tool (since it accurately measures the temperature in each pixel), in terms of the medical quantification of disease or injury it is a qualitative method only [13,17]. By using it, we can recognize the physiological and non-physiological regions of the body but remain unable to quantify them. According to [13], this issue stems from the following aspects: (1) changes in the thermal image are gradual and continuous, and thus the image contains very few clear landmarks (sharp edges or clear points), which are required for specific definition of the region of interest Sensors(ROI). 2020, 20 The, x FOR exactly PEER REVIEW bounded ROI embodies a necessary precondition to enable quantification3 of 24 within the impacted region. (2) The DMTI output is created by the lossy transformation of the 3. Qualitativereal 3D world only, into ano 2D quantification—althoug projection; thus, the imageh 2D is distorted.thermography This scenario basically does constitutes not allow usa quantificationto implement certaintool (since steps, it accurately for example, measures measuring the temperature the surface of in the each aff ectedpixel), area: in terms we can of the see medicalonly the quantification projection, which of disease is significantly or injury influenced it is a qualitative by the actualmethod setup only of [13,17]. the camera By using in relation it, we canto the recognize scanned the body. physiological (3) In order and to non-physiological obtain two comparable regions thermograms, of the body but we remain have tounable exactly to quantifypreserve them. the scene, According including to [13], the exact this issue position stems of thefrom subject the following (as even tiny aspects: changes (1) changes of the position in the thermalmay exert image a substantial are gradual impact and on thecontinuous, results). Suchand athus condition the image is hardly contains achievable very in few currently clear landmarksused setups. (sharp edges or clear points), which are required for specific definition of the region of interest (ROI). The exactly bounded ROI embodies a necessary precondition to enable The RoScan multispectral 3D scanning project, which is described in the following chapters, quantification within the impacted region. (2) The DMTI output is created by the lossy addresses and eliminates all of the three DMTI problems outlined above. Thus, the novel system transformation of the real 3D world into a 2D projection; thus, the image is distorted. This upgrades DMTI to wide-range high-accuracy 3D thermography, which can be readily used as an scenario does not allow us to implement certain steps, for example, measuring the surface of the objective quantitative tool in various medical subdomains [10]. affected area: we can see only the projection, which is significantly influenced by the actual setup At present, the actual medical quantification embodies a topic broadly discussed throughout the of the camera in relation to the scanned body. (3) In order to obtain two comparable healthcare sector. Insufficiently accurate, inadequately sensitive, or excessively subjective diagnostic thermograms, we have to exactly preserve the scene, including the exact position of the subject methods are perceived as very restrictive in several medical domains, including dermatology, (as even tiny changes of the position may exert a substantial impact on the results). Such a traumatology, physiotherapy, and forensic sciences. condition is hardly achievable in currently used setups.

Figure 2. The thermalthermal footprintsfootprints cannot cannot be be clearly clearly assigned assigned to to particular particular locations locations on theon body.the body. The lackThe lackof sharp of sharp edges edges and and other other spot spot landmarks landmarks makes make exacts exact localization localization in in a a single single 2D2D thermalthermal image impossible unless registration is performed with such a supplementary supplementary layer that yields yields clear clear features. features.

Sensors 2020, 20, 6656 4 of 22 Sensors 2020, 20, x FOR PEER REVIEW 4 of 24

(a) (b)

Figure 3.3. TheThe impossibility impossibility of of simultaneously reaching reaching both both a wide range and a high resolution restrains the use of thermalthermal imaging in medicine:medicine: ( (a)) in a wide-range view, the whole body part is visible, butbut thethe detailsdetails of of the the temperature temperature distribution distribution are are not; not; (b )(b in) in a narrow-range a narrow-range view, view, the the details details are visible,are visible, but thebut region the region cannot cannot be localized, be localized, and context and context characterizing characterizing the whole the a wholeffected affected area is missing. area is missing. In dermatology, coarse and subjective scoring systems remain in wide use as the primary means of evaluationThe RoScan because multispectral an objective 3D and scanning reliable project,in vivo quantificationwhich is described method in applicablethe following to a particularchapters, regionaddresses of interest and eliminates has not beenall of developed the three toDMTI date. pr Dueoblems to this outlined deficiency, above. a diagnosisThus, the basednovel onsystem such subjectiveupgrades DMTI data is to then wide-range unable to high-accuracy reveal tiny abnormalities 3D thermography, to detect which the disease can be in itsreadily early used stage. as an objectiveTraumatology quantitative and tool forensic in various sciences medical would subdomains greatly welcome [10]. an objective method for quantifying and monitoringAt present, diversethe actual eff medicalects or symptoms, quantification such em asbodies bruises a andtopic their broadly severity; discussed however, throughout this type the of methodhealthcare is stillsector. missing Insufficiently even in the accurate, two fields. inadequately sensitive, or excessively subjective diagnostic methodsIn physiotherapy, are perceived as one very of restrictive the common in several symptoms medical is adomains, change including in the body dermatology, volume. Thesetraumatology, manifestations physiotherapy, have to be and measured forensic objectivelysciences. and sensitively in order to reveal an emerging diseaseIn dermatology, already at the coarse initial stage.and subjective Although scoring the changes systems are remain currently in wide measurable use as the ina primary comparatively means accurateof evaluation manner, because we are an unable objective to distinguish and reliable between in vivo a physiologicalquantification (e.g., method muscle applicable growth) to and a aparticular non-physiological region of interest (e.g., swelling) has not been change. develope Moreover,d to date. it is Due also to di ffithiscult deficiency, to assess a the diagnosis impact ofbased the treatmenton such subjective procedures data because is then the unable existing to reveal treatment tiny evaluation abnormalities exploits to detect subjective the disease health in surveys its early or low-resolutionstage. scoring systems with poor inter-observer correlation. TheseTraumatology problems and (and forensic many sciences others) are would tackled greatl byy the welcome RoScan an multispectral objective method 3D scanning for quantifying system. Inand this monitoring study, the diverse basic principles effects or and symptoms, benefits such of the as approach bruises and are their characterized severity; andhowever, demonstrated this type onof severalmethod medicalis still missing applications even in where the two pilot fields. case studies were performed with positive results. In physiotherapy, one of the common symptoms is a change in the body volume. These 2. Materials and Methods manifestations have to be measured objectively and sensitively in order to reveal an emerging disease alreadyThis at paperthe initial discusses stage. novel Although approaches the changes to and are applications currently measurable within 3D in thermal a comparatively scanning. Theaccurate presented manner, scanning we are system, unable developedto distinguish completely between ata Brnophysiological University (e.g., of Technology,muscle growth) combines and a anon-physiological high operational (e.g., flexibility swelling) with change. high absolute Moreover, accuracy it is ofalso the difficult resulting to 3Dassess model. the Suchimpact a fusionof the oftreatment capabilities procedures is still because uncommon the ex inisting current treatment commercial evaluation 3D scanners, exploits assubjective these usually health surveys embody or a compromiselow-resolution between scoring the systems scanning with accuracy poor inter-observer and flexibility correlation. [18]; in medicine, however, both of these parametersThese problems are needed. (and A flexiblemany others) and accurate are tackled 3D surface by the RoScan scanning multispectral system is thus 3D very scanning likely system. to find wideIn this use study, in various the basic medical principles domains, and benefits especially of ifthe it yieldsapproach extensive are characterized 3D models and demonstrated maintains the highon several accuracy medical across applications all the spectral where layers, pilot thus case providingstudies were options performed not regularly with positive available results. within the state of the art. The RoScan multispectral 3D system consists of a robotic manipulator and several sensors, grouped together at the endpoint into a sensing head (Figure4a). To allow medical use, we combine a laser scanner, a color camera (wavelength range between 390 and 750 nm), and a thermal imager

Sensors 2020, 20, x FOR PEER REVIEW 5 of 24

2. Materials and Methods This paper discusses novel approaches to and applications within 3D thermal scanning. The presented scanning system, developed completely at Brno University of Technology, combines a high operational flexibility with high absolute accuracy of the resulting 3D model. Such a fusion of capabilities is still uncommon in current commercial 3D scanners, as these usually embody a compromise between the scanning accuracy and flexibility [18]; in medicine, however, both of these parameters are needed. A flexible and accurate 3D surface scanning system is thus very likely to find wide use in various medical domains, especially if it yields extensive 3D models and maintains the high accuracy across all the spectral layers, thus providing options not regularly available within the state of the art. The RoScan multispectral 3D system consists of a robotic manipulator and several sensors, grouped together at the endpoint into a sensing head (Figure 4a). To allow medical use, we combine a laser scanner, a color camera (wavelength range between 390 and 750 nm), and a thermal imager (wavelength range 8 to 14 µm); however, the system is modular, meaning that the cameras can be replaced by other sensors providing data in various spectral ranges. The manipulator moves the sensors along the body being scanned (Figure 4d), and the software processes, in real time, the distance data measured by the laser scanner into the form of a 3D surface model (Figure 4b). In addition to the laser scanner necessary for acquiring the 3D model itself, the robot’s endpoint is equipped with a color camera (Figure 4c), which delivers color information relating to each point of the 3D mesh; a thermal imager is also included (but will be discussed only later in the text). To this point, the RoScan’s output may have seemed to be almost the same as that of commercially available 3D scanners; such an impression, however, will change quickly if we emphasize the fact that the output model can be large and complex without any influence on the accuracy. Compared to other current 3D scanning instruments, the robotic manipulator ensures very accurate and independent localization of the sensors; the accuracy of the resulting model then does not affect the object’s appearance or shape. The method is robust, and the accuracy constraints are Sensorsidentical2020 regardless, 20, 6656 of the scanned scene. Smooth homogeneous structures or repeated patterns5 ofare 22 processed flawlessly (this being a substantial difference from the problematic procedure that characterizes, for example, methods based on stereophotogrammetric principles [19]), and the 3D (wavelengthmodel does not range exhibit 8 to 14a µcumulativem); however, deviation the system (unlike is modular, the products meaning of, thatfor example, the cameras handheld can be replacedscanners by[20] other or other sensors methods providing based data on Iter in variousative closest spectral point ranges. (ICP) principle [21–23]).

Figure 4. The RoScan system: (a) the main hardware components, namely, the robotic manipulator and head with several sensors; (b) the software yields a 3D mesh in real time; (c) the sensing head comprises a 2D laser scanner, a thermal imager, and an RGB camera; (d) the sensing head moves along the body, capturing thousands of distance proﬁles needed to create a 3D spatial model and taking several images from the thermal and visible spectra; the spectra are then mapped onto the surface.

The manipulator moves the sensors along the body being scanned (Figure4d), and the software processes, in real time, the distance data measured by the laser scanner into the form of a 3D surface model (Figure4b). In addition to the laser scanner necessary for acquiring the 3D model itself, the robot’s endpoint is equipped with a color camera (Figure4c), which delivers color information relating to each point of the 3D mesh; a thermal imager is also included (but will be discussed only later in the text). To this point, the RoScan’s output may have seemed to be almost the same as that of commercially available 3D scanners; such an impression, however, will change quickly if we emphasize the fact that the output model can be large and complex without any influence on the accuracy. Compared to other current 3D scanning instruments, the robotic manipulator ensures very accurate and independent localization of the sensors; the accuracy of the resulting model then does not affect the object’s appearance or shape. The method is robust, and the accuracy constraints are identical regardless of the scanned scene. Smooth homogeneous structures or repeated patterns are processed flawlessly (this being a substantial difference from the problematic procedure that characterizes, for example, methods based on stereophotogrammetric principles [19]), and the 3D model does not exhibit a cumulative deviation (unlike the products of, for example, handheld scanners [20] or other methods based on Iterative closest point (ICP) principle [21–23]). An additional information layer, unavailable in current 3D scanners, is generated by the thermal imager. The advantageously composed structure utilizes the robotic manipulator to capture the temperature images and 3D spatial data at different positions: The images are taken at a significantly greater proximity to compensate for the low resolution of thermal cameras and to achieve the same resolution in the thermogram, the other layers, and the entire 3D model. Consequently, a true, Sensors 2020, 20, 6656 6 of 22

non-interpolated temperature value is assignable to each model point, even in extensive and highly detailed models. This capability then resolves and eliminates one of the problems that accompany the use of DMTI (see Sections1 and 2.4, issue No. 2, resolution vs. range problem). In the RoScan, the output is a multispectral 3D computer model providing a color shape coupled with the thermal distribution. The combination of 3D data, thermal data, and a color image addresses also the DMTI drawbacks Nos. 1 and 3, as further explained in Section 2.4. This paper brieﬂy exposes the abilities, functions, and beneﬁts of the entire RoScan system, focusing on its applicability; multiple details on the implementation are then outlined in other papers written by the authors of this article. The studies in question are dedicated to the actual 3D scanning [24–26], the mapping of 2D images onto the surface of the 3D model [27], and the laser scanner and camera calibration [28,29].

2.1. Capturing 3D Models with the RoScan A multispectral 3D model is captured via the robotic manipulator moving the sensing head around the scanned object and along the predefined trajectory. The standard approach consists in defining different trajectories for each body part to be scanned; such predefined trajectories are simply selected according to the body part where a disease is expected. During the execution of a trajectory, the laser scanner captures the distance profiles, while the cameras collect the thermal and color images. The color images and distance profiles are obtained at similar proximities because both sensors exhibit an almost identical resolution rate. The resolution of the thermal imager is nevertheless significantly lower, making us capture (compared to the corresponding process performed with the color camera) markedly more thermal images from a significantly shorter Sensors 2020, 20, x FOR PEER REVIEW 7 of 24 distance. Thus, we finally obtain the same resolution in the color and the thermal layers (Figure5).

(a) (b)

FigureFigure 5. 5. TheThe procedures procedures of of reaching reaching the the sa sameme resolution resolution rates rates in in all all of of the the spectral spectral layers layers despite despite the the nativenative resolution resolution of of the the se sensorsnsors differ differ significantly: significantly: ( (aa)) hundreds hundreds of of thermal thermal images images are are captured captured at at a a veryvery short distance,distance, yielding yielding a higha high resolution resolution in a in single a single image; image; the range, the rang however,e, however, is narrow. is Multiplenarrow. Multipleimages are images taken are and taken then mergedand then by merged mapping by onto ma thepping surface onto to the form surface a single, to form high-resolution a single, high- layer resolutioncovering the layer whole coveri model;ng the (b whole) the native model; resolution (b) the native of the resolution color camera of the is markedlycolor camera higher, is markedly meaning higher,that only meaning several that images only taken several at a images greater taken distance at a are greater needed distance to reach are the needed final resolution to reach the identical final resolutionwith that achievedidentical inwith scenario that achieved (a). in scenario (a).

All of the collected distance profiles are aligned to the same system of coordinates by using homogeneous transformations; subsequently, the 3D shaded mesh is generated [25]. In the next step, the 2D images from the cameras are mapped to the surface of the 3D model [27]. At this stage, the ray-tracing algorithm [30] is employed to examine the visibility of each point of the mesh from the camera. If a single point is visible in multiple images, the resulting temperature is defined as the average of the values, as follows: the color in such points is computed as linear interpolation between the colors from these images, weighted by the angle relative to the 3D surface normal because the lightness of the color is influenced by the light reflection angle. The output model is delivered in a particular file format, which can be opened in the RoScan Analyzer software tool, described in the next section.

2.2. Tool to Allow 3D Model View and Analysis Multispectral 3D surface models created with the RoScan can be displayed via the RoScan Analyzer. In addition to the common features, such as rotating, scaling, and adjusting the modes of view, the software instrument also facilitates measuring many spatial parameters of the selected region of interest (ROI) and allows exporting the 3D data in several standard file formats (PTS, XYZ, and PLY); the data are then processable with any third-party 3D modeling program. Each model can be viewed in the following modes (Figure 6): • Color—the basic view of the clinician. In the model, this layer is used mainly for localization, either by natural landmarks of the skin (such as pigmented spots), via pen-marked markers, or through the use of region boundaries; such landmarks can be highlighted to be visible also in other layers (Figure 7). Alternatively, when highlighting abnormal places in the visualized thermal layer, we can adopt the opposite approach, which relies on localizing the spots in a color representation.

Sensors 2020, 20, 6656 7 of 22

All of the collected distance profiles are aligned to the same system of coordinates by using homogeneous transformations; subsequently, the 3D shaded mesh is generated [25]. In the next step, the 2D images from the cameras are mapped to the surface of the 3D model [27]. At this stage, the ray-tracing algorithm [30] is employed to examine the visibility of each point of the mesh from the camera. If a single point is visible in multiple images, the resulting temperature is defined as the average of the values, as follows: the color in such points is computed as linear interpolation between the colors from these images, weighted by the angle relative to the 3D surface normal because the lightness of the color is influenced by the light reflection angle. The output model is delivered in a particular file format, which can be opened in the RoScan Analyzer software tool, described in the next section.

Color—the basic view of the clinician. In the model, this layer is used mainly for localization, • either by natural landmarks of the skin (such as pigmented spots), via pen-marked markers, or through the use of region boundaries; such landmarks can be highlighted to be visible also in other layers (Figure7). Alternatively, when highlighting abnormal places in the visualized thermal layer, we can adopt the opposite approach, which relies on localizing the spots in a color representation. Temperature—the layer with false colors that correspond to the temperature of the skin; here, • the exact value of a particular point can be examined by clicking. To reach the appropriate contrast between the physiological and the non-physiological areas, we can adjust the coloring palette by moving a slider. Roughness—this layer shows false colors that correspond to the roughness of the scanned • surface. This output utilizes the measuring principle of the applied laser scanner, which exploits triangulation [31]. As a side result of this measuring method, the divergence of the reﬂected beam is acquired; such divergence correlates with the roughness of the scanned surface. Surface only—in certain cases, the coloring and shading of the model may hide some details or • produce an optical illusion. Representing the surface in a single color allows us to see even the tiny details that normally often remain uncovered. When spatial details are to be examined, it is often convenient to show the 3D surface only, with the coloring eliminated.

On each layer, the user can select points of interest (POIs), facilitating spatial measurements with the model. In the selected points or regions of interest (ROIs) deﬁned thereby, the following parameters can be assessed:

Distance [mm]—separates several selected points. The distances are measured either in the direct • mode or via the shortest path along the surface (these options are applicable in determining, e.g., the circumference of the ROI and the distance between the focus and the landmark). Angle [degree]—the angle between every three subsequent POIs, including, for example, those • that allow an analysis of vertebrae positions. Surface area [mm2]—relating to the whole model or the ROI only, for example, the area of burn, • or extent of lesion. Volume [mm3]—concerning the entire model or a part thereof. Such a portion can be deﬁned by • the cutting plane or deﬂected cutting surface. Useful for, e.g., measurements of the limb volume. Sensors 2020, 20, x FOR PEER REVIEW 8 of 24

• Temperature—the layer with false colors that correspond to the temperature of the skin; here, Sensorsthe2020 exact, 20, 6656 value of a particular point can be examined by clicking. To reach the appropriate8 of 22 contrast between the physiological and the non-physiological areas, we can adjust the coloring palette by moving a slider. Color—of a selected POI or the average color of the ROI. It is a more objective alternative to the • Roughness—this layer shows false colors that correspond to the roughness of the scanned currently used scoring systems for evaluating lesion redness. surface. This output utilizes the measuring principle of the applied laser scanner, which exploits Roughness—deﬁned by the beam divergence width (in the sensor pixels), which is the • triangulation [31]. As a side result of this measuring method, the divergence of the reflected dimensionless index correlating with the surface roughness at a particular point. We can beam is acquired; such divergence correlates with the roughness of the scanned surface. either examine such roughness or calculate the average roughness of the ROI. • Surface only—in certain cases, the coloring and shading of the model may hide some details or Temperature [ C]—of a highlighted point or the average temperature of the ROI. This parameter • produce an optical◦ illusion. Representing the surface in a single color allows us to see even the hastinythe details highest that information normally often value, remain since uncovered. it correlates When with inﬂammationspatial details inare the to area.be examined, it is often convenient to show the 3D surface only, with the coloring eliminated.

Sensors 2020, 20, x FOR PEER REVIEW 9 of 24

• Color—of a selected POI or the average color of the ROI. It is a more objective alternative to the currently used scoring systems for evaluating lesion redness. • Roughness—defined by the beam divergence width (in the sensor pixels), which is the dimensionless index correlating with the surface roughness at a particular point. We can either Figureexamine 6.6. suchTheThe outputroughness output model model or calculatof of the RoScan e the average systemsystem roughness consists of of several the ROI. layers:layers: ( (aa)) a 3D spatialspatial • representationTemperature without[°C]—ofwithout coloring;coloring; a highlighted ( b(b)) a a surface surface point with orwith the true true av colors;erage colors; ( ctemperature )(c false-color-mapped) false-color-mapped of the ROI. thermal thermal This data; parameter data; (d) false-color-mappedhas(d) false-color-mapped the highest information roughness roughness data.value, data. since it correlates with inflammation in the area.

On each layer, the user can select points of interest (POIs), facilitating spatial measurements with the model. In the selected points or regions of interest (ROIs) defined thereby, the following parameters can be assessed: • Distance [mm]—separates several selected points. The distances are measured either in the direct mode or via the shortest path along the surface (these options are applicable in determining, e.g., the circumference of the ROI and the distance between the focus and the landmark). • Angle [degree]—the angle between every three subsequent POIs, including, for example, those that allow an analysis of vertebrae positions. • Surface area [mm2]—relating(a) to the whole model or the ROI only, for(b )example, the area of burn, or extent of lesion. Figure 7. TheThe navigation navigation between between several several layers layers carried carried out out by byusing using points points of interest of interest (POIs): (POIs): (a) • Volume [mm3]—concerning the entire model or a part thereof. Such a portion can be defined by (selectinga) selecting the thearea area with with suspicious suspicious features features in inthe the color color layer; layer; (b (b) )examining examining the the temperature at the cutting plane or deflected cutting surface. Useful for, e.g., measurements of the limb volume. highlighted pointspoints inin thethe thermalthermal layer. layer. In In this this case, case, we we did did not not observe observe any any thermal thermal changes changes relating relating to theto the eﬀ effectect indicated indicated in thein the image image (a); (a the); the observation observation corresponds corresponds to reality,to reality, as theas the injury injury was was created created by aby make-up a make-up artist. artist.

Such a multimodal 3D computer model finds use in, for example, inflammation monitoring. The process then comprises the following stages: first, the thermal layer is studied, and possible thermal abnormalities are highlighted in the 3D model. Second, we switch to the color mode, which facilitates exact localization of the inflammation foci and allows us to set up the region of interest (ROI). Afterwards, we can quantify the inflammation activity (usually in the form of the mean temperature within the ROI). At the last stage, the swelling-induced volume variation is measured accurately. The inflammation monitoring and quantification constitute only one of the RoScan’s potential target applications. To illustrate the claim, we can note in this context that while inflammation leads to a positive thermal data gradient, ischemia results in a local decrease; thus, the RoScan may be employed in the monitoring of diabetic necrotic tissues.

2.3. Key Parameters of the RoScan The central advantage of the discussed concept rests in the applied combination of spatial, color, and thermal data within a single 3D model; such an approach delivers new research opportunities, which would otherwise remain virtually impossible to implement. More concretely, the RoScan has a major potential in the objective evaluation of spatial changes of the body; this medical task includes, above all, the monitoring of oedemas, muscle growth and atrophy, inflammation, and necrosis. The color layer guarantees the precise selection of a ROI (via natural landmarks or markers drawn on the subject’s skin). The thermal layer helps to find inflamed areas or determine if a spatial change is caused by physiological or non-physiological processes. Finally, 3D scanning yields an undistorted view, thus offering an effective alternative to standard 2D imaging technologies, where the distortion prevents objective measurements from being performed. The essential capabilities and properties of the RoScan system are summarized in Table 1.

Sensors 2020, 20, 6656 9 of 22

Such a multimodal 3D computer model finds use in, for example, inflammation monitoring. The process then comprises the following stages: first, the thermal layer is studied, and possible thermal abnormalities are highlighted in the 3D model. Second, we switch to the color mode, which facilitates exact localization of the inflammation foci and allows us to set up the region of interest (ROI). Afterwards, we can quantify the inflammation activity (usually in the form of the mean temperature within the ROI). At the last stage, the swelling-induced volume variation is measured accurately. The inflammation monitoring and quantification constitute only one of the RoScan’s potential target applications. To illustrate the claim, we can note in this context that while inflammation leads to a positive thermal data gradient, ischemia results in a local decrease; thus, the RoScan may be employed in the monitoring of diabetic necrotic tissues.

2.3. Key Parameters of the RoScan The central advantage of the discussed concept rests in the applied combination of spatial, color, and thermal data within a single 3D model; such an approach delivers new research opportunities, which would otherwise remain virtually impossible to implement. More concretely, the RoScan has a major potential in the objective evaluation of spatial changes of the body; this medical task includes, above all, the monitoring of oedemas, muscle growth and atrophy, inflammation, and necrosis. The color layer guarantees the precise selection of a ROI (via natural landmarks or markers drawn on the subject’s skin). The thermal layer helps to find inflamed areas or determine if a spatial change is caused by physiological or non-physiological processes. Finally, 3D scanning yields an undistorted view, thus offering an effective alternative to standard 2D imaging technologies, where the distortion prevents objective measurements from being performed. The essential capabilities and properties of the RoScan system are summarized in Table1.

Table 1. The key parameters.

Parameter Value Positional accuracy of model points (3 σ). 0.12 mm Scanning speed 10 mm/s Scanning trajectory ﬂexible, programmable Measured values position, color, temperature, roughness Operational cost approximately 0.1 EUR per model Price of a device approximately 30,000 EUR (or less) Data storage format internal Export formats for other software PLY, PTS, XYZ

The price of the device indicated in the table above relates to the research setup characterized herein; that is, a configuration which utilizes a robotic manipulator. Advantageously, this component delivers superior flexibility and versatility, allowing the entire device to find use in many different applications, even outside of medicine. If deployed in practice, however, the system is usually not required to perform an overly wide range of activities, and thus its six-degrees-of-freedom flexibility can become overkill. The manipulator can then be replaced by a cheaper (but still very accurate) positioning system with fewer degrees of freedom (Rails, Cartesian robots, etc.), maintaining the key parameters. As the price of the manipulator amounts to 19,000 EUR out of the total of 30,000 EUR, the acquisition costs could then drop markedly. Although the purchase costs are significantly lower than those in the current imaging methods, the most extensive savings are achieved in regular operation, where the RoScan generates almost zero additional costs, regardless of the frequency of use. This is a major benefit compared to, for example, MRI, in which each scan is worth 200–400 EUR [32,33]. Sensors 2020, 20, 6656 10 of 22

2.4. Eliminating the Drawbacks of DMTI As described in the Introduction, thermography is considered a useful method, even though its benefits are limited by three fundamental drawbacks [12–16]. These issues are addressed by the RoScan, as follows: (1) The exact localization problem arises from a lack of clear point objects in a thermogram. Despite being able to detect areas with non-physiological behavior in a thermogram, we cannot determine exactly where they are (Figure2 and Section1, issue No. 1, the exact localization problem). For such a purpose, mutual registration of the thermal and the color layers in the RoScan appears to be advantageous, as the potentially problematic spots are marked in the thermal layer (Figure8a) and then localized in the color one (Figure8b), which comprises a significantly higher number of landmarksSensors 2020, 20 allowing, x FOR PEER suffi REVIEWcient orientation (e.g., pigmented spots, scars, eczema, and veins). 11 of 24

(a) (b)

Figure 8.8.Resolving Resolving the the exact exact localization localization problem: problem: (a) the ( potentiallya) the potentially problematic problematic spots are highlighted spots are inhighlighted the thermal in layerthe thermal of the 3D layer model; of the (b) 3D the model; selected (b points) the selected are localized points in are the colorlocalized layer. in the color layer. (2) The resolution vs. range problem is generated by the low resolution of thermal imagers. Since the thermal(2) changesThe resolution inside vs. the range human problem body areis generated usually small by the and low can resolution hide in the of physiological thermal imagers. temperature Since gradientsthe thermal [8 ],changes it is necessary inside tothe provide human databody having are usually very high small resolution and can [hide13]. Thisin the objective physiological can be achievedtemperature by focusinggradients the [8], thermal it is necessary imager on to the provide nearest data possible having distance; very thehigh obtained resolution range [13]. of view,This however,objective willcan bebe narrowachieved and by will focusing neither the contain thermal enough imager information on the nearest to perform possible the localization distance; northe coverobtained the range whole of area view, aﬀected however, (Figure will3 and be Sectionnarrow1 and, issue will No. neither 2, the contain resolution enough vs. range information problem). to Theperform RoScan the resolveslocalization the problemnor cover by the capturing whole area multiple affected images (Figure at a 3 near and distance Section and1, issue from No. a known 2, the position;resolution these vs. range images problem). are then The mapped RoScan on resolves the 3D surface,the problem thus beingby capturing in fact mergedmultiple into images a single, at a large,near distance high-resolution and from image a known (Figure position;9). these images are then mapped on the 3D surface, thus being in fact merged into a single, large, high-resolution image (Figure 9).

Figure 9. The resolution vs. range problem: combining multiple images with the RoScan delivers a wide range and a high resolution simultaneously.

(3) The most prominent drawback lies in that classic DMTI is a mere qualitative tool, enabling us to distinguish between the physiological and non-physiological states of the body but lacking the ability to quantify them [13,17]. This deficiency is eliminated through the data fusion ensured by the RoScan, where the advantageous combination of sensors delivers color 3D surface scans with hi-

Sensors 2020, 20, x FOR PEER REVIEW 11 of 24

(a) (b)

Figure 8. Resolving the exact localization problem: (a) the potentially problematic spots are highlighted in the thermal layer of the 3D model; (b) the selected points are localized in the color layer.

(2) The resolution vs. range problem is generated by the low resolution of thermal imagers. Since the thermal changes inside the human body are usually small and can hide in the physiological temperature gradients [8], it is necessary to provide data having very high resolution [13]. This objective can be achieved by focusing the thermal imager on the nearest possible distance; the obtained range of view, however, will be narrow and will neither contain enough information to perform the localization nor cover the whole area affected (Figure 3 and Section 1, issue No. 2, the resolution vs. range problem). The RoScan resolves the problem by capturing multiple images at a nearSensors distance2020, 20, 6656and from a known position; these images are then mapped on the 3D surface,11 thus of 22 being in fact merged into a single, large, high-resolution image (Figure 9).

FigureFigure 9. 9. TheThe resolution resolution vs. vs. range range problem: problem: combining combining multiple multiple images images with with the the RoScan RoScan delivers delivers a a widewide range range and and a a high high resolution resolution simultaneously.

(3) The most prominent drawback lies in that classic DMTI is a mere qualitative tool, enabling (3) The most prominent drawback lies in that classic DMTI is a mere qualitative tool, enabling us to distinguish between the physiological and non-physiological states of the body but lacking the us to distinguish between the physiological and non-physiological states of the body but lacking the ability to quantify them [13,17]. This deficiency is eliminated through the data fusion ensured by ability to quantify them [13,17]. This deficiency is eliminated through the data fusion ensured by the the RoScan, where the advantageous combination of sensors delivers color 3D surface scans with RoScan, where the advantageous combination of sensors delivers color 3D surface scans with hi- hi-resolution thermal information (Figure6). Thanks to the color layer, the ROI can be selected precisely (see Section 2.4, issue 1, the localization problem); such a step is unfeasible with DMTI. The 3D model features the ability to capture undistorted measurements and provides the data in a form independent from the point of view adopted in the scanning. Finally, the thermal layer exposes non-physiological areas and yields a quantifiable index reflecting the severity of the disease. Resolving these limitations opens the door to suitable thermal imaging applications within the medical domain. Later in the text, a set of pilot case studies is presented, showing the potential of quantification carried out via 3D thermal imaging.

3. Results Several pilot experiments were performed within multiple medical specialties to validate the benefits of the RoScan. As this is the first application of 3D thermography utilizing the RoScan in the discussed field (and, in some instances, even thermography itself), we preferred case studies to broader analyses with wider populations of subjects. The outcomes of these experiments pointed to considerable usability of the device in medical diagnostics and demonstrated its advantages as related to the state of the art in a particular subdomain. We focused on the following activities and fields:

Assessing medicament eﬃcacy in dermatology (Section 3.1). • Assessing the objective progress of recovery in traumatology (Section 3.2). • Quantifying applied force in forensic sciences (Section 3.3). • Localizing exactly the cause of pain in physiotherapy (Section 3.4). • Volumetric measurements of soft tissues (Section 3.5). • Objective assessment of atopic dermatitis (Section 3.6). • The results obtained from the experiments, complemented with the methodologies used, are summarized in the relevant subchapters of this section. In order to provide a comprehensive overview of possible applications, we also included the studies already published; these studies already cite our original ﬁndings partially released through other articles, which contain more details about the experiment.

3.1. Assessing Medicament Eﬃcacy in Dermatology The case study was performed on a subject suﬀering from allergy to hazelnut substances. The experiment started when red, itchy lesions appeared on one of the subject’s feet. The Sensors 2020, 20, 6656 12 of 22 lesion-containing regions were highlighted with markers directly on the skin and then divided into two areas: K, where the Protopic 0.1% ointment was applied, and V, where a shea butter and coconut oil lotion was utilized. The patient was repeatedly scanned with the RoScan, and the average temperature gradient in both areas was evaluated by selecting the boundaries visible on the color layer ofSensors the 3D2020 model., 20, x FOR The PEER results REVIEW are shown in Figure 10. 13 of 24

(a) (b)

FigureFigure 10.10. The development of the temperature gradient index in both areas during 31 h afterafter thethe applicationapplication ofof thethe ointments:ointments: (a) the overall progress of temperaturestemperatures in both areas; ((b)) aa detaildetail ofof thethe ﬁrstfirst 100100 minmin afterafter thethe applicationapplication ofof thethe drugdrug [[34].34].

TheThe growthgrowth ofof thethe temperaturetemperature gradientgradient inin thethe sectorsector treatedtreated withwith ProtopicProtopic 0.1%0.1% culminatedculminated earlierearlier thanthan in thethe sheashea butterbutter andand coconutcoconut oil zone,zone, showing Protopic 0.1% to be the more eeffectiveffective medicamentmedicament inin thisthis particularparticular case.case. ItIt isis importantimportant toto stressstress thatthat thethe seriesseries ofof experimentsexperiments wouldwould notnot havehave beenbeen executableexecutable withwith aa regularregular state-of-the-artstate-of-the-art method.method. At present, the clinicalclinical tests are usually performedperformed on two groups of subjects,subjects, ofof which which one one is treatedis treated with with the testedthe tested medicament, medicament, while while the other the is other not;the is overallnot; the progress overall ofprogress each subject’s of each symptoms subject’s (including,symptoms for(including, example, fo redness,r example, swelling, redness, and swelling, itching) is and evaluated itching) by is a coarseevaluated scoring by a system coarse (usually scoring notsystem more (usually than three not grades). more than Such three an approach grades). thenSuch may an approach lead to wrong then conclusionsmay lead to ifwrong the selected conclusions group doesif the not selected constitute grou ap representativedoes not constitute sample a ofrepresentative the population sample (i.e., we of dothenot population know how (i.e., the we disease do not would know develop how the without disease treatment would indevelop each patient), without and treatment its sensitivity in each is ratherpatient), low. and its sensitivity is rather low. TheThe RoScan allows allows comparing comparing the the effe effctsects exerted exerted by by two two different different drugs drugs on the on thesame same subject subject and andeven even the thesame same lesion, lesion, thus thus greatly greatly increasing increasing the the reliability reliability of of the the study study because because an identicalidentical spontaneousspontaneous developmentdevelopment isis markedlymarkedly moremore likelylikely toto be foundfound inin twotwo partsparts ofof oneone lesionlesion thanthan inin thethe lesionslesions ofof twotwo didifferentfferent patients.patients. ThisThis tasktask isis virtuallyvirtually impossibleimpossible to implement via the current methods; moreover, thethe system’ssystem’s sensitivitysensitivity andand selectivityselectivity enablesenables thethe useruser to evaluate the progress of the treatment inin great detail,detail, asas shownshown inin FigureFigure 10 10b.b. InIn thethe areaarea K, K, the the patient patient began began to to feel feel burning burning skin skin 2 min2 min after after the the application application of theof the ointments; ointments; the the sensation sensation culminated culminated at at 7 7 min min following following the the medical medical intervention, intervention, andand atat 1212 minmin thethe burningburning andand itchingitching in the area disappeared completely. In the area V, thethe itchingitching diddid notnot changechange duringduring thethe time.time. The patient’s report correlates with the responseresponse toto thethe developingdeveloping inflammationinflammation duringduring the firstfirst minutesminutes afterafter thethe applicationapplication of the drug:drug: initially, the process appeared to bebe moremore intensiveintensive inin thethe areaarea K;K; here,here, however,however, thethe inflammationinflammation also also regresses regresses at at a a faster faster pace. pace. AnotherAnother majormajor advantage advantage of theof the RoScan RoScan consists consists in that in more that quantitative more quantitative parameters parameters are available are atavailable the same at time,the same while time, the currentwhile the methods current usually method focuss usually on a singlefocus on item a single only; thus,item inonly; addition thus, toin addition to the lesion temperature, the data relating to the color, reflectance, and roughness of the lesion surface and area are accessible. As the thermal distribution of a patient’s surface temperature is affected by many environmental factors and also the subject’s pre-measurement activities, the measured index was established as a difference between the area’s average temperature and reference temperature, normalized by the value at the time when the dermatological medicaments were applied. Assessing the thermal difference between the reference point and the ROI eliminates environmental influences (it has been observed that the subjects’ previous activity, together with environmental effects, leads to an identical increase within the whole area). Normalization had been chosen due to the unequal distance of both

Sensors 2020, 20, 6656 13 of 22 the lesion temperature, the data relating to the color, reflectance, and roughness of the lesion surface and area are accessible. As the thermal distribution of a patient’s surface temperature is affected by many environmental factors and also the subject’s pre-measurement activities, the measured index was established as a difference between the area’s average temperature and reference temperature, normalized by the value at the time when the dermatological medicaments were applied. Assessing the thermal difference betweenSensors 2020 the, 20,reference x FOR PEER point REVIEW and the ROI eliminates environmental influences (it has been observed14 of 24 that the subjects’ previous activity, together with environmental effects, leads to an identical increase withinareas from the whole the edge area). of Normalizationthe body, i.e., hada problem been chosen that causes due to differences the unequal in distance the absolute of both values areas fromof a thetemperature. edge of the This body, approach i.e., a problem normal thatizes causes both divaluesfferences to the in thesame absolute scaled values index ofand a temperature. makes the areas This approachcomparable normalizes between botheach valuesother. Further to the same details scaled concerning index and this makes experiment the areas are comparable provided in between source each[34]. other. Further details concerning this experiment are provided in source [34].

3.2. Assessing Objective Progress of Recovery in Traumatology The experiment was performed with a subject ha havingving a bruised toe following a hit by a heavy object. Figure Figure 1111 contains images images of the bruised toe immediately after the injury and then several hours later, withwith thethe aaffectedﬀected areaarea stillstill clearlyclearly highlightedhighlighted inin thethe 3D3D thermogram.thermogram.

(a) (b)

Figure 11. The injured toe captured 2 h ( a) and 74 h ( b) after after the accident. The The volume volume and the average temperature ofof thethe thumbthumb appearappear toto bebe reducedreduced [[35].35].

In order order to to eliminate eliminate temperature temperature fluctuations fluctuations generated generated by byexternal external factors, factors, which which are non- are non-negligiblenegligible especially especially in the in thelower lower limbs, limbs, we we choos choosee the the difference difference between between the the average average temperatures temperatures of both toes as an objectiveobjective index, since we assume that the impact of external influencesinfluences will be comparable inin bothboth toes, toes, and and their their thermal thermal di ffdifferenceerence will will then then be almostbe almost zero zero in the in physiological the physiological state. state.Figure 12 represents the development of the average temperature of the bruised toe; the actual processFigure is expressed 12 represents in relation the development to changes in of the the toe’s aver volume.age temperature The temperature of the bruised decrease toe; during the actual the timeprocess exhibits is expressed an almost in relation exponential to changes trend: in the the swelling toe’s volume. recedes The slowly temperature at the beginning decrease during but abates the moretime exhibits quickly throughan almost the exponential second and trend: third daysthe sw ofelling recovery. recedes slowly at the beginning but abates more quickly through the second and third days of recovery. Importantly, the subject’s feelings of pain subsided already 74 h after the injury (Figure 11b); however, compared to the healthy toe, there was a still visible and measurable temperature difference (0.8 °C). Additional details on the experiment are outlined in source [35]. This scenario shows that the RoScan facilitates early detection of inflamed regions before they start to give pain or develop significantly; the system is thus capable of reducing the treatment and recovery time and cost. Unlike the current visual evaluation-based approach, the novel technique yields more accurate and objective data.

Sensors 2020, 20, 6656 14 of 22 Sensors 2020, 20, x FOR PEER REVIEW 15 of 24

Figure 12. The progress curve relating to the average temperature in the region of the bruised toe, Figurecompleted 12. The with progress an indication curve of relating the volume to the [35 average]. temperature in the region of the bruised toe, completed with an indication of the volume [35]. Importantly, the subject’s feelings of pain subsided already 74 h after the injury (Figure 11b); however,The relevant compared experiment to the healthy was toe,perf thereormed was with a still five visible subjects and suffer measurableing from temperature the same problem difference at (0.8various◦C). degrees Additional of intensity. details on In the all experimentof the cases, are the outlined thermal in difference source [35 progressed]. similarly, with the maximumThis scenario value corresponding shows that the to RoScanthe bruise facilitates severity. early detection of inflamed regions before they start to give pain or develop significantly; the system is thus capable of reducing the treatment and 3.3.recovery Quantifying time and Applied cost. Force Unlike in Forensic the current Sciences visual evaluation-based approach, the novel technique yieldsAnother more accurate use for and3D objectivethermography data. is in the evaluation of the force applied in violent crimes. AlthoughThe relevant the RoScan experiment will probably was performed be unable to with dete fivermine subjects the specific suffering value from of the force same or problem the exact at varioustime of degreesthe actual of intensity.act, it may In allbeneficially of the cases, allow the thermalcomparison difference of two progressed forces by similarly, examining with their the consequences.maximum value corresponding to the bruise severity. In the simulation, the subject was punched with an imaginary attacker’s fist in the biceps, once 3.3. Quantifying Applied Force in Forensic Sciences and then twice, at two pre-marked places; the first was modeled by using a heavy pack of sand having a small,Another hard userubber for ball 3D thermographyattached to the isbottom. in the evaluationThe pack was of theinvariably force applied released in from violent the crimes. same Althoughheight, thus the hitting RoScan the will biceps probably with be unablean identical to determine force in the each specific cycle. value During of the the force 6 h or following the exact time the experiment,of the actual both act, it areas may beneficiallywere repeatedly allow scanned, comparison and ofthe two temperature forces by examining differences their between consequences. the pre- markedIn thelocations simulation, and the the pre-marked subject was reference punched poin witht anon imaginarythe shoulder attacker’s were evaluated fist in the (Figure biceps, 13). once and then twice, at two pre-marked places; the first was modeled by using a heavy pack of sand having a small, hard rubber ball attached to the bottom. The pack was invariably released from the same height, thus hitting the biceps with an identical force in each cycle. During the 6 h following the experiment, both areas were repeatedly scanned, and the temperature differences between the pre-marked locations and the pre-marked reference point on the shoulder were evaluated (Figure 13). The temperature in both of the affected areas increased sharply at the exact time of the punch and then continued to grow during the first 50 min after the act. Interestingly, the measured temperature correlated with the number of the first hits, and the traces of the assault remained visible on the thermogram for more than 3 h following the punching. In this context, it should be noted that the “experimental punches” were very soft, and the subject had already stopped feeling pain 10 min later; such a progress stresses the high sensitivity of the method, even though in a real attack the signs would likely have been noticeable for markedly longer.

Sensors 2020, 20, 6656 15 of 22 Sensors 2020, 20, x FOR PEER REVIEW 16 of 24

Figure 13. TheThe thermal thermal differences differences relating relating to to the the initial initial conditions conditions and and their their progress progress during during the the first first 66 h after the simulated assault; th thee subject was punched at t = 0:00.0:00. Conversely, we need to admit that the hits were aimed at a temperature-homogeneous area of the The temperature in both of the affected areas increased sharply at the exact time of the punch body, in which a uniform change caused by external influences can be expected. This type of results and then continued to grow during the first 50 min after the act. Interestingly, the measured may not be generally achievable throughout the whole body; however, given that violent cases are temperature correlated with the number of the first hits, and the traces of the assault remained visible often accompanied by a lack of evidence, such a record, too, has a useful potential. on the thermogram for more than 3 h following the punching. In this context, it should be noted that the3.4. “experimental Localizing Exactly punches” the Cause were ofPain very in soft, Physiotherapy and the subject had already stopped feeling pain 10 min later; such a progress stresses the high sensitivity of the method, even though in a real attack the signs wouldAn likely important have been premise noticeable for selecting for markedly the right longer. treatment in physiotherapy consists in ensuring the correctConversely, diagnosis we of theneed cause to admit of the that pain the [36 hits]; however,were aimed this at task a temperature-homogeneous is often very difficult to perform, area of thegiven body, that in at timeswhich we a haveuniform to rely change on only caused the feedback by external from influences a patient who can cannotbe expected. distinguish This betweentype of resultsdifferent may types not ofbe pain generally or accurately achievable identify throughout the relevant the whole location body; [37 however,]. given that violent cases are oftenIn practice, accompanied DMTI is by used a lack where of evidence, suitable, helping such a record, the examiners too, has to a reveal useful the potential. source of inflammation and to determine the character of the problem (e.g., whether the cause rests in an articular block or 3.4.muscular Localizing defects). Exactly The the technique Cause of Pain is nevertheless in Physiotherapy incapable of accurately localizing the center of pain and preserving an objective record similar to that produced by the RoScan. An important premise for selecting the right treatment in physiotherapy consists in ensuring the Such records can be employed as follows: First, the ability to capture the current state of the correct diagnosis of the cause of the pain [36]; however, this task is often very difficult to perform, patient objectively facilitates further comparison during his or her next clinical consultation; otherwise, given that at times we have to rely on only the feedback from a patient who cannot distinguish if the changes are not large enough to be observable, the patient’s condition may be wrongly considered between different types of pain or accurately identify the relevant location [37]. inalterable. The records also enable us to reliably assess the severity of the disease because the method In practice, DMTI is used where suitable, helping the examiners to reveal the source of is not as subjective as the others. Second, the thermal quantification (if possible) leads towards an inflammation and to determine the character of the problem (e.g., whether the cause rests in an assessment of the treatment suitability or medication effectiveness [35]. articular block or muscular defects). The technique is nevertheless incapable of accurately localizing In our study, several patients were repeatedly scanned with the RoScan during the recovery the center of pain and preserving an objective record similar to that produced by the RoScan. process following an injury or surgery. A comparison of the inter-visit 3D scans provided a proof of Such records can be employed as follows: First, the ability to capture the current state of the positive progress and facilitated the initial choice and definition of further treatment. In this context, patient objectively facilitates further comparison during his or her next clinical consultation; for example, we can refer to the evidence of inflammation in a subject suffering from entezopathy of otherwise, if the changes are not large enough to be observable, the patient’s condition may be the deltoid muscle, revealed during the first examination (Figure 14a). The inflammation had recessed wrongly considered inalterable. The records also enable us to reliably assess the severity of the markedly before the second examination (Figure 14b), and improved blood flow in the patella area, disease because the method is not as subjective as the others. Second, the thermal quantification (if accompanying the post-surgical recovery between the first (Figure 14c) and the second examinations possible) leads towards an assessment of the treatment suitability or medication effectiveness [35]. (Figure 14d), is also evident. Positive progress was determined from the 3D models already after two In our study, several patients were repeatedly scanned with the RoScan during the recovery process following an injury or surgery. A comparison of the inter-visit 3D scans provided a proof of positive progress and facilitated the initial choice and definition of further treatment. In this context,

Sensors 2020, 20, x FOR PEER REVIEW 17 of 24 for example, we can refer to the evidence of inflammation in a subject suffering from entezopathy of the deltoid muscle, revealed during the first examination (Figure 14a). The inflammation had recessed markedly before the second examination (Figure 14b), and improved blood flow in the patella area, accompanying the post-surgical recovery between the first (Figure 14c) and the second examinations (Figure 14d), is also evident. Positive progress was determined from the 3D models already after two days, when the condition was not yet evaluable by means of the common methods (which mostly exploit visual observation and manual examination). The benefit of the RoScan in physiotherapy rests mainly in the exact localization capability; however, the device can also function as a quantification tool in cases where it is possible to suppress Sensors 2020, 20, 6656 16 of 22 the external fluctuations that affect body temperature. Such a situation is presented in Figure 14a,b; the images show an area with a homogeneous temperature distribution where the external influences candays, be whensuppressed the condition via comparison was not with yet evaluable a reference by point means nearby. of the In common item 14ac, methods such normalization (which mostly is difficultexploit visual to achieve, observation and the and user manual will thus examination). benefit from the above-mentioned localization only.

Figure 14. An upper limblimb aaffectedffected byby entezopathy entezopathy of of the the deltoid deltoid muscle muscle at at (a )( thea) the first first visit visit and and (b) the(b) thesecond second visit; visit; the healing the healing of the of inflammation the inflammation is clearly is observable.clearly observable. The recovery The followingrecovery following a surgery ofa surgeryruptured of knee ruptured ligaments knee atligaments (c) the first at visit(c) the and first (d )visit the secondand (d) visit; the second improved visit; blood improved circulation blood is circulationvisible. Note: is visible. the regular Note: hot the spots regular in ( ahot,b) spots are caused in (a,b by) are cellulite. caused by cellulite.

3.5. VolumetricThe benefit Measurements of the RoScan of Soft in Tissues physiotherapy rests mainly in the exact localization capability; however, the device can also function as a quantification tool in cases where it is possible to suppress the externalWithin this fluctuations experiment, that the affect current body volumetric temperature. techniques Such a situation were compared is presented with inthe Figure RoScan 14 a,b;via measuringthe images the show volumes an area of with reference a homogeneous objects (phant temperatureoms). When distribution performing where against the external the current influences gold standard,can be suppressed i.e., water via displacement comparison withvolumetry, a reference the pointRoScan nearby. delivered In item volumetric 14ac, such measurements normalization at is higherdifficult accuracy to achieve, and and repeatability, the user will as thusshown benefit in Figure from 15. the above-mentioned localization only.

3.5. Volumetric Measurements of Soft Tissues Within this experiment, the current volumetric techniques were compared with the RoScan via measuring the volumes of reference objects (phantoms). When performing against the current gold standard, i.e., water displacement volumetry, the RoScan delivered volumetric measurements at higher accuracy and repeatability, as shown in Figure 15. In terms of its parameters and capabilities, the RoScan embodies a faster, safer, and more accurate approach with very low operational costs; the system is especially valuable in specific cases where advanced ROI examination and high accuracy are needed. Such special volumetric cycles normally require the standard MRI or CT; however, the RoScan, if employed, reaches more accurate results at significantly lower costs [38]. As the RoScan is generally the most accurate method, it will allow measuring of even very small changes in the volume, which can enable us to discover pathological changes in a timely manner, well before they have developed significantly. Additional details on the experiment can be found in source [12], and the current volumetric methods are compared between one another and with the RoScan in [39] and [38,40], respectively.

Sensors 2020, 20, 6656 17 of 22 Sensors 2020, 20, x FOR PEER REVIEW 18 of 24

Figure 15. The RoScan compared with the circumferential techniques and the water displacement Figuremethod 15. implemented The RoScan incompared a reference with object the havingcircumferent a knownial techniques volume: the and RoScan the water displayed displacement the most methodaccurate implemented and repeatable in a performance reference object [12]. having a known volume: the RoScan displayed the most accurate and repeatable performance [12]. 3.6. Objective Assessment of Atopic Dermatitis In terms of its parameters and capabilities, the RoScan embodies a faster, safer, and more There are many diagnostic methods for assessing atopic dermatitis (AD), varying from scoring accurate approach with very low operational costs; the system is especially valuable in specific cases systems (very subjective, insensitive, and providing low resolution) to sophisticated and expensive where advanced ROI examination and high accuracy are needed. Such special volumetric cycles options [41]. normally require the standard MRI or CT; however, the RoScan, if employed, reaches more accurate In this context, DMTI is considered a relatively objective (and thus a potentially reliable) method, results at significantly lower costs [38]. as it measures increased blood flow, which manifests itself almost identically in all patients [42]. As the RoScan is generally the most accurate method, it will allow measuring of even very small The technique, however, is rarely used due to the low resolution and complicated standardization of changes in the volume, which can enable us to discover pathological changes in a timely manner, the data that facilitate comparison between the results (the DMTI exact localization problem). well before they have developed significantly. As mentioned above, the RoScan’s structure eliminates this difficulty, and thus it finds use in Additional details on the experiment can be found in source [12], and the current volumetric AD lesion assessment. The capability was demonstrated in the experiment, whereby an AD patient methods are compared between one another and with the RoScan in [39] and [38,40], respectively. was examined via DMTI; nevertheless, the thermal data could not be causally linked to the real lesions. The relationship was then established only by using the RoScan and its ability to select the 3.6. Objective Assessment of Atopic Dermatitis nonphysiological locations and to navigate in the color layer (Figure8). The thermal information wasThere then properlyare many related diagnostic to the methods AD lesions, for assessing and the atopic progress dermatitis was quantified (AD), varying to meet from the scoring goal of systemsthe experiment. (very subjective, insensitive, and providing low resolution) to sophisticated and expensive options [41]. 4. DiscussionIn this context, DMTI is considered a relatively objective (and thus a potentially reliable) method, as it measuresThe RoScan increased project, blood completely flow, which developed manifests at Brno itself University almost identically of Technology, in all patients embodies [42]. a novel The technique,concept of however, using a robotic is rarely manipulator used due to combinedthe low resolution with several and sensorscomplicated producing standardization data in di offferent the dataspectra. that facilitate For medical comparison applications, between it appears the results to be (the advantageous DMTI exact to localization combine a problem). laser scanner, a color cameraAs mentioned (wavelength above, range the between RoScan’s 390 structure and 750 nm), eliminates and a thermalthis difficulty, imager and (wavelength thus it finds range use from in AD8 to lesion 14 µm); assessment. such a configuration The capability is beneficialwas demonstr becauseated thein the thermal experiment, changes whereby commonly an AD accompany patient wasthe non-physiologicalexamined via DMTI; states. nevertheless, This innovative the thermal method da forta could multispectral not be causally 3D scanning linked brings to the several real lesions.advantages The relationship compared to was the state-of-the-artthen established approaches. only by using the RoScan and its ability to select the nonphysiologicalThanks to the locations projection and of to the navigate spectral in data the color onto thelayer surface (Figure of the8). The 3D modelthermal via information an independent was thenlocalization properly system, related the to process the AD of lesions, registering and thethe 2D progress images was is robust quantified and independent to meet the of goal the scanned of the experiment.object. The procedure is applicable to any smooth and homogeneous surface or texture, thus being especially valuable in merging several thermal images into a large one if there is a lack of reliably detectable registration landmarks.

Sensors 2020, 20, 6656 18 of 22

Another major benefit of independent localization lies in cross-spectral image registration (color vs. thermal). With this method, one can reach the same resolution in both the thermal and the color layers, even if the native resolutions of the sensor differ significantly. Moreover, both layers are properly aligned to the 3D model and to each other; importantly, the uncertainty is exactly specified and equal in the entire dataset. Advancement beyond the possibilities of the state-of-the-art is recognizable also within 3D scanning; unlike many of the current, commercially available 3D methods, the RoScan allows each point of the captured model to exhibit an equal, absolute spatial accuracy independent of the scanned scene and lacking cumulative deviations. Additionally, the accuracy is precisely quantifiable from the parameters of the manipulator and sensor, bringing a major advantage in precise measurements, where the uncertainty must be specified in an exact manner. Most of the current 3D scanners are not capable of such operations [43,44]. The RoScan offers a considerable potential in those medical applications where objective quantification plays the key role. The system quantifies inflammatory processes inside the human body; several pilot case studies described in this paper showed that using the RoScan to quantify inflammation is feasible when the inflamed region remains close to the skin, has sufficient “strength” to influence the skin temperature, and ideally appears in an area where the temperature distribution is as homogeneous as possible, enabling us to find an index with the potential to eliminate the environmental factors. The RoScan-based assessment of medicament efficacy in dermatology showed the possibility of comparing the impacts of two different drugs on the same subject and even the same lesion, an objective virtually unachievable via the current methods. Standard clinical tests are usually carried out on two groups of subjects, of which one is treated with the tested medicament, while the other is not; such an approach then may lead to wrong conclusions if the selected group does not constitute a representative sample of the population, and its sensitivity is rather low (meaning that only major progress is distinguishable). A traumatology-related study discussing the assessment of objective progress in recovery pointed out that the RoScan enables early detection of inflamed regions before they start to give pain or develop significantly, stressing also the facts that the system is thus capable of reducing the treatment and recovery time and cost. Moreover, the technique embodies a valuable tool to yield an objective record of the current state of a disease; producing such records is a common problem of present medicine, which intensively relies on the personal memory of the clinician. Another research subdomain where the RoScan reaches beyond the state of the art lies in investigating the quantification of applied force: the system proved to be highly sensitive, with the traces of a physical assault remaining visible on the thermogram for more than 3 h after the violent act, even though the actual pain caused by the punching disappeared in only 10 min. The measured temperature correlated with the number of punches received. In physiotherapy, the RoScan delivered objective evaluation of the treatment effects, documenting objectively blood circulation improvement after an invasive surgery. Finally, the tool has been verified as suitable for soft tissue volumetry, bringing superior accuracy and repeatability. An additional benefit is the possibility to select the edema only, stemming from the ability of the thermal layer to distinguish the edema from a healthy tissue. Neither of the achievements is obtainable via the current volumetric methods. At this point, however, it is necessary to emphasize that 3D thermography remains a mere tool requiring expert assessment by a doctor. The RoScan does not aim to offer automatic detection and localization of inflammation; rather, the device is designed to embody a source of objective data that will allow clinicians to decide and quantify better and more correctly than if such data are lacking. Areas with higher temperatures are not always inflamed; often the effect expresses an anatomical constitution (e.g., when the arteries lie closer to the skin). A significant impact is generated also by Sensors 2020, 20, 6656 19 of 22 external influences, including the ambient temperature, the patient’s previous activity, and current footwear or clothing. Although the effect of these elements can be limited in many cases by choosing a suitable relative index, this approach is not applicable generally, meaning that the 3D thermogram will still have to be interpreted and evaluated by a medical doctor, who nevertheless may not be as experienced as a physician diagnosing without the thermogram.

5. Conclusions Generally, precise 3D thermography tools, such as the RoScan, can yield major benefits in various healthcare domains. The relevant instruments facilitate early detection of inflamed areas, the reason being that their sensitivities surpass those of the commonly used methods, which are mostly based on visual observations [45–47]. The precise 3D thermography as performed by the RoScan addresses three main limitations of Digital Medical Thermal Imaging (DMTI), discussed in the present scientific literature: the exact localization problem, the resolution vs. range problem, and quantification impossibility. In this context, DMTI has been characterized as having a substantial potential if the drawbacks were resolved. The RoScan system eliminates all of these issues, thus opening new possibilities for thermal imaging in medicine. Several pilot case studies confirmed the benefits introduced by the RoScan into diverse subdomains of healthcare, where the advanced concept overcomes the limits inherent in the other state of-the-art methods. This scanning approach appears to be particularly beneficial within medical efficacy assessment, dermatology, traumatology, and physiotherapy, where it facilitates objective monitoring of the recovery progress. In all of these areas, the novel multispectral 3D method can be advantageously adopted as an objective quantification tool to fill the void in the set of options available for the given purpose. More concretely, the RoScan is especially useful when the common symptoms have not yet fully developed; however, multispectral 3D scanning in general finds application also when the symptoms are already visible. Importantly, the capability of preserving the exact state of the patient’s body (to allows comparison during the next visit at the clinician) leads to objective evaluation of the progress of the disease. The positive results of the initial case studies presented herein pointed out the areas where the opportunity to enhance the state of the art is potentially high. In future projects, we will focus predominantly on these domains, and we will attempt to transfer the proposed technology into everyday medical practice. Although it will probably be necessary to modify the system for the given applications, this task will not be problematic, especially due to the fact that the entire system was fully developed at our department and is thus completely controlled by the authors of the paper. Major portions of the research will be dedicated to the maximum simplification and automation of the scanning process because if the system is to be practically usable, it must offer easy operability without expert knowledge and significant time intensity. The system is convenient for not only medicine but also technology and science: thermal 3D scanning procedures are usable in, for example, electrical engineering to ensure thermal analyses centered on the efficiency of electrical components [48]; civil engineering to monitor thermal loss in materials [49]; mechanical engineering to investigate mechanical strain [50]; and rescue robotics or experimental biology to evaluate the indispensable life functions [51–53]. Moreover, if we also take into account the fact all algorithms are universal and thus make it possible to easily replace the color and thermal camera with another sensor of a different spectral range, the usability of the RoScan system will expand even more significantly.

Author Contributions: Concept—L.Z. and A.C.; methodology—A.C.; software—A.C.; validation—A.C. and L.Z.; resources—L.Z.; data curation—A.C.; writing/original draft structuring—A.C.; writing—review and editing—A.C.; visualization—A.C.; supervision—L.Z.; project administration—A.C.; funding acquisition—L.Z. All authors have read and agreed to the published version of the manuscript. Sensors 2020, 20, 6656 20 of 22

Funding: This research was funded by the grant “Advancing Smart Optical Imaging and Sensing for Health (ASTONISH)”, H2020-EU.2.1.1.7.—ECSEL programme, topic ECSEL-04-2015—Smart Health, grant agreement ID: 692470. The research has been financially supported by the Ministry of Education, Youth and Sports of the Czech Republic under the project CEITEC 2020 (LQ1601). The completion of this paper was made possible by the grant No. FEKT-S-20-6205—“Research in Automation, Cybernetics and Artificial Intelligence within Industry 4.0” financially supported by the Internal science fund of Brno University of Technology. Acknowledgments: The authors would like to thank all the patients who volunteered to participate in the experiments which involved 3D thermal scanning of their body parts. The experiments were performed in compliance with the Declaration of Helsinki, i.e., the participants agreed to the form and content of the experiment appointments. Conflicts of Interest: The authors declare no conflict of interest. The funding institutions had no role in designing the study; in collecting, analyzing, and interpreting the data; in writing the manuscript; or in the decision to publish the results.

References

1. Papadakis, M.A.; McPhee, S.J.; Rabow, M.W. CURRENT Medical Diagnosis and Treatment 2019; McGraw Hill Professional: New York, NY, USA, 2019; ISBN 978-1-260-11744-8. 2. New, A. Evidence-based medicine. A new approach to teaching the practice of medicine. Evidence-Based Medicine Working Group. JAMA J. Am. Med. Assoc. 1992, 268, 2420–2425. [CrossRef] 3. Sackett, D.L. Evidence-based medicine. Semin. Perinatol. 1997, 21, 3–5. [CrossRef] 4. Chang, T.-C.; Hsiao, Y.-L.; Liao, S.-L. Application of digital infrared thermal imaging in determining inflammatory state and follow-up effect of methylprednisolone pulse therapy in patients with Graves’ ophthalmopathy. Graefes Arch. Clin. Exp. Ophthalmol. 2008, 246, 45–49. [CrossRef] 5. Gabrlik, P.; la Cour-Harbo, A.; Kalvodova, P.; Zalud, L.; Janata, P. Calibration and accuracy assessment in a direct georeferencing system for UAS photogrammetry. Int. J. Remote Sens. 2018, 39, 4931–4959. [CrossRef] 6. Coffey, V.C. Multispectral Imaging Moves into the Mainstream. Opt. Photonics News 2012, 23, 18–24. [CrossRef] 7. Hilton-Jones, D. Diagnosis and Treatment of Inflammatory Muscle Diseases. J. Neurol. Neurosurg. Psychiatry 2003, 74, ii25–ii31. [CrossRef] 8. Ring, E.F.J.; Ammer, K. Infrared thermal imaging in medicine. Physiol. Meas. 2012, 33, R33. [CrossRef] 9. Lu, G.; Fei, B. Medical hyperspectral imaging: A review. J. Biomed. Opt. 2014, 19, 010901. [CrossRef] 10. Aretusi, G.; Fontanella, L.; Ippoliti, L.; Merla, A. Supervised Classification of Thermal High-Resolution IR Images for the Diagnosis of Raynaud’s Phenomenon. In New Perspectives in Statistical Modeling and Data Analysis; Ingrassia, S., Rocci, R., Vichi, M., Eds.; Springer: Berlin/Heidelberg, Germany, 2011; pp. 419–427. 11. Engert, V.; Merla, A.; Grant, J.A.; Cardone, D.; Tusche, A.; Singer, T. Exploring the Use of Thermal Infrared Imaging in Human Stress Research. PLoS ONE 2014, 9, e90782. [CrossRef] 12. Chromy, A. High Resolution Multispectral 3D Scanning and Its Medical Applications. Ph.D. Thesis, Department of Control and Instrumentation, Faculty of Electrical Engineering and Communication, Brno University of Technology, Brno, Czech Republic, 2017. 13. Vardasca, R.; Simoes, R. Current Issues in Medical Thermography. In Topics in Medical Image Processing and Computational Vision; Lecture Notes in Computational Vision and Biomechanics; Springer: Dordrecht, The Netherlands, 2013; pp. 223–237. ISBN 978-94-007-0725-2. 14. Gul, K.M.; Ahmadi, N.; Wang, Z.; Jamieson, C.; Nasir, K.; Metcalfe, R.; Hecht, H.S.; Hartley, C.J.; Naghavi, M. Digital thermal monitoring of vascular function: A novel tool to improve cardiovascular risk assessment. Vasc. Med. Lond. Engl. 2009, 14, 143–148. [CrossRef] 15. Bruehl, S.; Lubenow, T.R.; Nath, H.; Ivankovich, O. Validation of thermography in the diagnosis of reflex sympathetic dystrophy. Clin. J. Pain 1996, 12, 316–325. [CrossRef][PubMed] 16. Di Carlo, A. Thermography and the possibilities for its applications in clinical and experimental dermatology. Clin. Dermatol. 1995, 13, 329–336. [CrossRef] 17. Ju, X.; Nebel, J.-C.; Siebert, J.P. 3D thermography imaging standardization technique for inflammation diagnosis. In Infrared Components and Their Applications; Gong, H., Cai, Y., Chatard, J.-P., Eds.; International Society for Optics and Photonics: Beijing, China, 2005; Volume 5640, p. 266. 18. Curless, B. From Range Scans to 3D Models. SIGGRAPH Comput. Graph. 1999, 33, 38–41. [CrossRef] 19. Brown, L.G. A Survey of Image Registration Techniques. ACM Comput. Surv. 1992, 24, 325–376. [CrossRef] Sensors 2020, 20, 6656 21 of 22

20. Strobl, K.H.; Mair, E.; Hirzinger, G. Image-based pose estimation for 3-D modeling in rapid, hand-held motion. In Proceedings of the 2011 IEEE International Conference on Robotics and Automation (ICRA), Shanghai, China, 9–13 May 2011; pp. 2593–2600. 21. Alshawa, M. ICL: Iterative closest line—A novel point cloud registration algorithm based on linear features. Ekscentar 2007, 10. 22. Poppinga, J.; Vaskevicius, N.; Birk, A.; Pathak, K. Fast plane detection and polygonalization in noisy 3D range images. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2008, Nice, France, 22–26 September 2008; pp. 3378–3383. 23. Bellocchio, F.; Borghese, N.A.; Ferrari, S.; Piuri, V. 3D Surface Reconstruction; Springer: New York, NY, USA, 2013; ISBN 978-1-4614-5631-5. 24. Chromy, A.; Zalud, L. Novel 3D Modelling System Capturing Objects with Sub-Millimetre Resolution. Adv. Electr. Electron. Eng. 2014, 12, 476–487. [CrossRef] 25. Chromy, A.; Zalud, L. Robotic 3D scanner as an alternative to standard modalities of medical imaging. SpringerPlus 2014, 3, 13. [CrossRef] 26. Chromy, A.; Kocmanova, P.; Zalud, L. Creating Three-Dimensional Computer Models Using Robotic Manipulator and Laser Scanners. In Proceedings of the 12th IFAC Conference on Programmable Devices and Embedded Systems (2013), Velke Karlovice, Czech Republic, 25–27 September 2013; Elsevier B.V.: Velke Karlovice, Czech Republic, 2013; pp. 268–273. 27. Chromy, A.; Klima, O. A 3D Scan Model and Thermal Image Data Fusion Algorithms for 3D Thermography in Medicine. J. Healthc. Eng. 2017, 2017, 5134021. [CrossRef] 28. Kocmanova, P.; Zalud, L.; Chromy, A. 3D proximity laser scanner calibration. In Proceedings of the 2013 18th International Conference on Methods and Models in Automation and Robotics (MMAR), Miedzyzdroje, Poland, 26–29 August 2013; pp. 742–747. 29. Chromy, A. Mutual Calibration of Sensors for Multispectral 3D Scanning of Surface. In Proceedings of the 9th International Congress on Ultra Modern Telecommunications and Control Systems, Munich, Germany, 6–8 November 2017; IEEE: Munich, Germany, 2017; pp. 199–204. 30. Suffern, K. Ray Tracing from the Ground Up; CRC Press: Boca Raton, FL, USA, 2016; ISBN 978-1-4987-7470-3. 31. Smith, K.B.; Zheng, Y.F. Accuracy Analysis of Point Laser Triangulation Probes Using Simulation. J. Manuf. Sci. Eng. 1998, 120, 736–745. [CrossRef] 32. Magnetic Resonance Imaging. Available online: http://www.mri-portal.com/clanky/magneticka_rezonance. php (accessed on 22 April 2018). 33. Novelline, R.A.; Squire, L.F. Squire’s Fundamentals of Radiology, 6th ed.; Harvard University Press: Cambridge, MA, USA, 2004; ISBN 978-0-674-01279-0. 34. Chromy,A. Multispectral 3D Surface Scanning System RoScan and its Application in Inflammation Monitoring and Quantification. In Proceedings of the 11th International Joint Conference on Biomedical Engineering Systems and Technologies: BIOIMAGING, Madeira, Portugal, 19–21 January 2018; Volume 2, pp. 106–113. 35. Chromy,A.; Zalud, L. Three-dimensional Thermal Imaging in Medicine. In Proceedings of the 3DBODY.TECH 2017—8th International Conference and Exhibition on 3D Body Scanning and Processing Technologies, Montreal, QC, Canada, 11–12 October 2017; Hometrica Consulting—Dr. Nicola D’Apuzzo: Montreal, QC, Canada, 2017; pp. 232–238. 36. Véle, F. Funkˇcní diagnostika—Pˇredpoklad úspˇechufyzioterapeuta. Rehabil. Fyzikální Lékaˇrství 2012, 19, 155–158. 37. Vele, F. Vysetreni Hybnych Funkci z Pohledu Neurofyziologie: Prirucka pro Terapeuty Pracujici v Neurorehabilitaci; Triton: Praha, Czech Republic, 2012; ISBN 978-80-7387-608-1. 38. Chromy, A. Application of High-Resolution 3D Scanning in Medical Volumetry. Int. J. Electron. Telecommun. 2016, 62, 23–31. [CrossRef] 39. Chromy, A.; Zalud, L.; Dobsak, P.; Suskevic, I.; Mrkvicova, V. Limb volume measurements: Comparison of accuracy and decisive parameters of the most used present methods. SpringerPlus 2015, 4, 707. [CrossRef] [PubMed] 40. Chromy, A. High-Accuracy Volumetric Measurements of Soft Tissues using Robotic 3D Scanner. IFAC-Pap 2015, 48, 318–323. [CrossRef] Sensors 2020, 20, 6656 22 of 22

41. Sprikkelman, A.B.; Tupker, R.A.; Burgerhof, H.; Schouten, J.P.; Brand, P.L.; Heymans, H.S.; van Aalderen, W.M. Severity scoring of atopic dermatitis: A comparison of three scoring systems. Allergy 1997, 52, 944–949. [CrossRef][PubMed] 42. Johansen, J.D.; Frosch, P.J.; Lepoittevin, J.-P. Contact Dermatitis; Springer Science & Business Media: Berlin/Heidelberg, Germany, 2010; ISBN 978-3-642-03827-3. 43. Polo, M.-E.; Felicísimo, Á.M. Analysis of Uncertainty and Repeatability of a Low-Cost 3D Laser Scanner. Sensors 2012, 12, 9046–9054. [CrossRef][PubMed] 44. Beraldin, J.-A. Basic theory on surface measurement uncertainty of 3D imaging systems. Three-Dimens. Imaging Metrol. 2009, 7239.[CrossRef] 45. Lipsky, B.A.; Berendt, A.R.; Deery, H.G.; Embil, J.M.; Joseph, W.S.; Karchmer, A.W.; LeFrock, J.L.; Lew, D.P.; Mader, J.T.; Norden, C.; et al. Diagnosis and Treatment of Diabetic Foot Infections. Clin. Infect. Dis. 2004, 39, 885–910. [CrossRef] 46. Krysko, D.V.; Vanden Berghe, T.; D’Herde, K.; Vandenabeele, P. Apoptosis and necrosis: Detection, discrimination and phagocytosis. Methods 2008, 44, 205–221. [CrossRef] 47. Merla, A.; Romani, G.L. Biomedical Applications of Functional Infrared Imaging. In Proceedings of the 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference, Shanghai, China, 17–18 January 2006; pp. 690–693. 48. Merryman, S.A.; Nelms, R.M. Diagnostic technique for power systems utilizing infrared thermal imaging. IEEE Trans. Ind. Electron. 1995, 42, 615–628. [CrossRef] 49. Taileb, A.; Dekkiche, H. Infrared Imaging as a Means of Analyzing and Improving Energy Eﬃciency of Building Envelopes: The case of a LEED Gold Building. Procedia Eng. 2015, 118, 639–646. [CrossRef] 50. Munz, D.; Yang, Y.Y. Stress Singularities at the Interface in Bonded Dissimilar Materials under Mechanical and Thermal Loading. J. Appl. Mech. 1992, 59, 857–861. [CrossRef] 51. Nejdl, L.; Kudr, J.; Cihalova, K.; Chudobova, D.; Zurek, M.; Zalud, L.; Kopecny, L.; Burian, F.; Ruttkay-Nedecky, B.; Krizkova, S.; et al. Remote-controlled robotic platform ORPHEUS as a new tool for detection of bacteria in the environment. Electrophoresis 2014, 35, 2333–2345. [CrossRef][PubMed] 52. Nejdl, L.; Kudr, J.; Ruttkay-Nedecky, B.; Heger, Z.; Zima, L.; Zalud, L.; Krizkova, S.; Adam, V.; Vaculovicova, M.; Kizek, R. Remote-Controlled Robotic Platform for Electrochemical Determination of Water Contaminated by Heavy Metal Ions. Int. J. Electrochem. Sci. 2015, 10, 3635–3643. 53. Zalud, L.; Kopecny, L.; Burian, F.; Florian, T. CASSANDRA—Heterogeneous reconnaissance robotic system for dangerous environments. In Proceedings of the 2011 IEEE/SICE International Symposium on System Integration, SII 2011, Kyoto, Japan, 20–22 December 2011; pp. 1275–1280.

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional aﬃliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).