A system and method of classifying a region of interest of a subject

Abstract

A system and method of categorizing a tissue by at least one processor may include obtaining a sequence of thermal images depicting a preheated, or precooled area of the tissue. The sequence may represent thermal changes in the depicted area over time. Embodiments may compute a plurality of time-based pixel vectors in the sequence of images, and for each pixel vector, calculate one or more respective thermal feature values, characteristic of the thermal changes over time in the corresponding location. Embodiments may determine one or more cold dots (CDs) in the depicted area, based on the thermal feature values of candidate pixel vectors. For each CD, embodiments may apply a classifier on the thermal feature values of one or more member candidate pixel vectors, to predict a probability that the relevant CD is anomalous.

Description

A SYSTEM AND METHOD OF CLASSIFYING A REGION OF INTEREST OFA SUBJECTCROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of priority of U.S. Provisional Patent Application No. 63 / 528,574, filed July 24, 2023, the contents of which are all incorporated herein by reference in their entirety.FIELD OF THE INVENTION

[0002] The present invention relates generally to the field of assistive diagnostics. More specifically, the present invention relates to systems and methods of categorizing a Region Of Interest (ROI) of a tissue.BACKGROUND OF THE INVENTION

[0003] Current methods for identifying and classifying tumors in tissues often rely on visual inspections or biopsies, which can be time-consuming, invasive, and may not provide accurate results. There is a need for a more efficient and accurate way of identifying and classifying tumors in tissues.SUMMARY OF THE INVENTION

[0004] The present invention is directed to a novel method for identifying and classifying tumors in tissues using thermal imaging (e.g., infrared scanning) technology. The invention involves receiving a thermal image (e.g., infrared scan) of a Region Of Interest (ROI) in a tissue and analyzing this scan to identify and / or classify a tumor in the tissue.

[0005] Infrared scanning technology has been used in medical applications for many years and has proven to be effective in detecting changes in tissue characteristics, such as temperature, blood flow, and chemical composition.

[0006] The present invention builds upon thermal imaging technology by providing a new approach for identifying and classifying tumors. Embodiments of the invention may include preheating, or precooling a tissue of interest, and subsequently analyzing changes in the unique thermal signature of the tissue over time, to gain insight or indication regarding a type and / or extent of a suspected anomaly. Such indication (denoted element 410 in Fig. 4)of an anomaly may pertain to an abnormal structure, shape, or composition of a tissue, and may include for example an indication of extent and / or type of a malignant tissue or tumor, an indication of extent and / or type of a benign tissue or tumor, an indication of extent of an inflammation (e.g., an inflamed tissue), an indication of a healthy tissue, and the like.

[0007] By analyzing the thermal scan of the tissue, the present invention can identify and classify tumors with a high degree of accuracy and reliability, making it a valuable tool for medical professionals in diagnosing and treating cancer. This technology has the potential to improve patient outcomes by enabling earlier and more accurate detection of tumors, leading to more effective and targeted treatments.

[0008] Embodiments of the invention may include a method of categorizing anomaly of a tissue by at least one processor. According to some embodiments, the at least one processor may obtain, from a thermal imaging device, a sequence of thermal images depicting a preheated, or precooled area of the tissue. The sequence may represent thermal changes in the depicted area over a predetermined measurement period. The at least one processor may compute a plurality of pixel vectors in the sequence of images, where each pixel vector may include member pixels that correspond to substantially the same location in the depicted area, along the measurement period. For each pixel vector, the at least one processor may calculate one or more respective thermal feature values, characteristic of the thermal changes over time in the corresponding location. The at least one processor may apply a filter on the plurality of pixel vectors, to obtain a subset of candidate pixel vectors, and determine one or more cold dots (CDs) in the depicted area, based on the thermal feature values of candidate pixel vectors, where each CD may include one or more member candidate pixel vectors. For each CD, the at least one processor may apply a first classifier on the thermal feature values of the one or more member candidate pixel vectors, to predict a risk score representing a probability that said CD is anomalous. The at least one processor may subsequently categorize anomaly of the depicted area based on the risk scores of the one or more CDs.

[0009] For example, the at least one processor may categorize the depicted area as anomalous by receiving a definition of an ROI within the depicted area, suspected to be anomalous; identifying one or more first CDs in the ROI; and categorizing the ROI as anomalous, based on the risk scores of the one or more first CDs.

[0010] Additionally, or alternatively, the at least one processor may identify one or more second CDs in a peripheral region of the ROI; calculate one or more spatial feature values,characteristic of spatial distribution of the one or more first CDs and the one or more second CDs; and categorize the ROI as anomalous, further based on the spatial feature values.

[0011] According to some embodiments, categorizing the ROI may include applying a pretrained, Machine Learning (ML) based, second classifier on (i) the risk scores, and (ii) the one or more spatial feature values, to predict an anomaly category of the ROI. The anomaly category may include, for example, a type of a malignant tumor, a type of a benign tumor, an inflamed tissue, and a healthy tissue.

[0012] According to some embodiments, categorizing the ROI may further include applying the pretrained, ML based, second classifier on (i) the risk scores, and (ii) the one or more spatial feature values, to predict a confidence level value. The confidence level value may include, for example a malignancy confidence level value, representing probability that the ROI is malignant, a benign confidence level value, representing probability that the ROI is benign, an inflammation confidence level value, representing probability that the ROI is inflamed, and a health confidence value, representing probability that the ROI is healthy.

[0013] According to some embodiments, the at least one processor may pretrain the second, ML based classifier by: receiving an annotated training thermal image sequence, said annotation may include: (i) a definition of an ROI, and (ii) an anomaly category of the ROI; determining one or more CDs in the training thermal image sequence; calculating risk scores of the one or more CDs in the training thermal image sequence; calculating spatial feature values characteristic of spatial distribution of the one or more CDs in the training thermal image sequence; and using the annotation as supervisory data, to train the second classifier, so as to predict the anomaly category of the defined ROI in the training thermal image sequence, based on (a) the calculated risk scores, and (b) the calculated spatial feature values.

[0014] Additionally, or alternatively, the at least one processor may categorize the ROI by receiving a visible-light image of the depicted area; analyzing the visible-light image to extract at least one morphological feature; and applying a pretrained, machine learning (ML) based, second classifier on (i) the risk scores, (ii) the one or more spatial feature values, and (iii) the at least one morphological feature, to predict an anomaly category of the ROI. In such embodiments, the anomaly category may include, for example, a type of a malignant tumor, a type of a benign tumor, an inflamed tissue and a healthy tissue.

[0015] According to some embodiments, obtaining the subset of candidate pixel vectors may include analyzing a specific pixel vector of the plurality of pixel vectors, to calculate amaximal heating value of the corresponding location; and rejecting, or filtering-out the specific pixel vector when the calculated maximal heating value does not surpass a predefined threshold.

[0016] According to some embodiments the tissue may be a cutaneous, or sub-cutaneous tissue of a furry animal, or a hairy person. In such embodiments, obtaining the subset of candidate pixel vectors may further include analyzing a thermal feature value of the specific pixel vector, to determine consistency of the thermal feature value with heat dissipation by fur; and rejecting, or filtering-out the specific pixel vector based on this consistency.

[0017] Additionally, or alternatively, obtaining the subset of candidate pixel vectors may include applying an ML-based, third classification model on a thermal feature value of the specific pixel vector, to predict a probability that the specific pixel vector depicts a fur of the animal; and rejecting the specific pixel vector based on said probability.

[0018] According to some embodiments, the at least one processor may determine a CD in the depicted area by determining a window of pixel vectors surrounding a candidate pixel vector; calculating at least one statistical metric value, representing difference in temperature between (i) the candidate pixel vector and (ii) pixel vectors at the border of said window, along the measurement period; and identifying the candidate pixel vector as pertaining to a CD, when the calculated at least one statistical metric value surpasses a predetermined threshold, for at least a predetermined duration.

[0019] According to some embodiments, the at least one processor may calculate a thermal feature value of a specific pixel vector by applying a polynomial fitting algorithm on the specific pixel vector, to fit values of member pixels into a polynomial function; and selecting the thermal feature value as a coefficient of a fourth-degree variable of said polynomial function.

[0020] Additionally, or alternatively, calculating a thermal feature value of a specific pixel vector may include calculating one or more Fourier series coefficients of the ROI pixel vector; and selecting the thermal feature value from the one or more calculated Fourier series coefficients.

[0021] Additionally, or alternatively, the at least one processor may employ a scanning light source, to obtain a scan data element representing a topography of the depicted area; and compensate at least one thermal feature value based on the topography, as represented by the scan data element.

[0022] Additionally, or alternatively, the at least one processor may apply an ML-based ROI calculation module on at least one thermal image of the depicted area, to automatically define (i) a first ROI (denoted herein as 40R), suspected as anomalous, and (ii) another ROI (denoted herein as 42R), beyond the first ROI, that is presumed healthy. As elaborated herein, the at least one processor may extract one or more features representing ROI 40R and ROI 42R. These features may include, for example spatial features, thermal features and morphological features, as explained herein. The at least one processor may subsequently categorize ROI 40R as anomalous based on the features of ROI 40R third ROI 42R.

[0023] Embodiments of the invention may include a system for categorizing anomaly of a tissue. Embodiments of the system may include a thermal imaging device, configured to obtain a sequence of thermal images depicting a preheated, or precooled area of the tissue, said sequence representing thermal changes in the depicted area over a predetermined measurement period; a non-transitory memory device, wherein modules of instruction code are stored, and at least one processor associated with the memory device, and configured to execute the modules of instruction code.

[0024] Upon execution of said modules of instruction code, the at least one processor may be configured to compute a plurality of pixel vectors in the sequence of images, wherein each pixel vector may include member pixels that correspond to substantially the same location in the depicted area, along the measurement period; for each pixel vector, calculate one or more respective thermal feature values, characteristic of the thermal changes over time in the corresponding location; apply a filter on the plurality of pixel vectors, to obtain a subset of candidate pixel vectors; determine one or more cold dots (CDs) in the depicted area, based on the thermal feature values of candidate pixel vectors, wherein each CD may include one or more member candidate pixel vectors; for each CD, apply a first classifier on the thermal feature values of the one or more member candidate pixel vectors, to predict a risk score representing a probability that said CD is anomalous; and categorize anomaly of the depicted area based on the risk scores of the one or more CDs.

[0025] Additionally, or alternatively, embodiments of the system may include a thermal element, associated with the at least one processor. The at least one processor may be further configured to control the thermal element so as to preheat, or precool the depicted area.

[0026] Additionally, or alternatively, embodiments of the system may include a scanning light source, associated with the at least one processor. The at least one processor may befurther configured to control the scanning light source to produce a scan the depicted area; obtain a scan data element, representing a topography of the depicted area; and compensate at least one thermal feature value, based on the topography, as represented by the scan data element.

[0027] Additionally, or alternatively, embodiments of the system may include a visible light imaging device. The at least one processor may be configured to control the visible light imaging device to obtain a visible light image of the depicted area, and categorize the ROI by: analyzing the visible-light image to extract at least one morphological feature; and applying a pretrained, ML based, second classifier on (i) the risk scores, (ii) the one or more spatial feature values, and (iii) the at least one morphological feature, to predict an anomaly category of the ROI.BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The subject matter regarded as the invention is particularly pointed out and distinctly claimed in the concluding portion of the specification. The invention, however, both as to organization and method of operation, together with objects, features, and advantages thereof, may best be understood by reference to the following detailed description when read with the accompanying drawings in which:

[0029] Fig. 1 is a block diagram, depicting a computing device which may be included in a system for categorizing a Region Of Interest (ROI) of a tissue, according to some embodiments of the invention;

[0030] Fig. 2 is a block diagram, depicting a generalized view of a system for categorizing an ROI of a tissue, according to some embodiments of the invention;

[0031] Fig. 3 is a graph depicting typical evolution of temperature at a specific location on a preheated, cutaneous tissue;

[0032] Fig. 4 is a block diagram, depicting an analysis module which may be included in a system for categorizing an ROI of a tissue, according to some embodiments of the invention;

[0033] Figs. 5A, 5B, and 5C are graphs showing examples of thermal features, which may be calculated, and analyzed by embodiments of the invention, to categorize an ROI of the tissue, according to some embodiments of the invention;

[0034] Figs. 6A and 6B are examples of thermal (e.g., IR) images of cutaneous tissues, according to some embodiments of the invention;

[0035] Fig. 7A is an image depicting an example of mass ROIs that may be analyzed according to some embodiments of the invention;

[0036] Fig. 7B is a thermal (e.g., IR) image of cutaneous tissues surrounding the mass ROIs of Fig. 7 A, according to some embodiments of the invention;

[0037] Fig. 7C is a graph describing analysis of the thermal image of Fig. 7B, to categorize the mass ROIs of Fig. 7A, according to some embodiments of the invention;

[0038] Figs. 8A-8D are images depicting an example of analysis of a visible light image, to obtain cutaneous morphological features, according to some embodiments of the invention;

[0039] Fig. 9 A is an image depicting thermal distribution on a flat surface in a thermal image produced by a thermal imaging device, according to some embodiments of the invention;

[0040] Fig. 9B is a schematic image showing a pattern of light that may be produced by scanning light source, according to some embodiments of the invention;

[0041] Fig. 10 is a graph depicting typical evolution of temperature at a specific preheated location of fur, according to some embodiments of the invention;

[0042] Fig. 11 is a block diagram, depicting a skin mask module which may be included in a system for classifying or categorizing an ROI of a tissue, according to some embodiments of the invention; and

[0043] Fig. 12 is a flow diagram, depicting a method of categorizing an ROI of a tissue, according to some embodiments of the invention.

[0044] It will be appreciated that for simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements for clarity. Further, where considered appropriate, reference numerals may be repeated among the figures to indicate corresponding or analogous elements.DETAILED DESCRIPTION OF THE PRESENT INVENTION

[0045] One skilled in the art will realize the invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The foregoing embodiments are therefore to be considered in all respects illustrative rather than limiting of the invention described herein. Scope of the invention is thus indicated by the appended claims, rather than by the foregoing description, and all changes that come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

[0046] In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be understood by those skilled in the art that the present invention may be practiced without these specific details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the present invention. Some features or elements described with respect to one embodiment may be combined with features or elements described with respect to other embodiments. For the sake of clarity, discussion of same or similar features or elements may not be repeated.

[0047] Although embodiments of the invention are not limited in this regard, discussions utilizing terms such as, for example, “processing,” “computing,” “calculating,” “determining,” “establishing”, “analyzing”, “checking”, or the like, may refer to operation(s) and / or process(es) of a computer, a computing platform, a computing system, or other electronic computing device, that manipulates and / or transforms data represented as physical (e.g., electronic) quantities within the computer’s registers and / or memories into other data similarly represented as physical quantities within the computer’s registers and / or memories or other information non-transitory storage medium that may store instructions to perform operations and / or processes.

[0048] Although embodiments of the invention are not limited in this regard, the terms “plurality” and “a plurality” as used herein may include, for example, “multiple” or “two or more”. The terms “plurality” or “a plurality” may be used throughout the specification to describe two or more components, devices, elements, units, parameters, or the like. The term “set” when used herein may include one or more items.

[0049] Unless explicitly stated, the method embodiments described herein are not constrained to a particular order or sequence. Additionally, some of the described method embodiments or elements thereof can occur or be performed simultaneously, at the same point in time, or concurrently.

[0050] Reference is now made to Fig. 1, which is a block diagram depicting a computing device, which may be included within an embodiment of a system for categorizing an ROI of a tissue, according to some embodiments.

[0051] Computing device 1 may include a processor or controller 2 that may be, for example, a central processing unit (CPU) processor, a chip or any suitable computing or computational device, an operating system 3, a memory 4, executable code 5, a storagesystem 6, input devices 7 and output devices 8. Processor 2 (or one or more controllers or processors, possibly across multiple units or devices) may be configured to carry out methods described herein, and / or to execute or act as the various modules, units, etc. More than one computing device 1 may be included in, and one or more computing devices 1 may act as the components of, a system according to embodiments of the invention.

[0052] Operating system 3 may be or may include any code segment (e.g., one similar to executable code 5 described herein) designed and / or configured to perform tasks involving coordination, scheduling, arbitration, supervising, controlling or otherwise managing operation of computing device 1, for example, scheduling execution of software programs or tasks or enabling software programs or other modules or units to communicate. Operating system 3 may be a commercial operating system. It will be noted that an operating system 3 may be an optional component, e.g., in some embodiments, a system may include a computing device that does not require or include an operating system 3.

[0053] Memory 4 may be or may include, for example, a Random- Access Memory (RAM), a read only memory (ROM), a Dynamic RAM (DRAM), a Synchronous DRAM (SDRAM), a double data rate (DDR) memory chip, a Flash memory, a volatile memory, a nonvolatile memory, a cache memory, a buffer, a short term memory unit, a long term memory unit, or other suitable memory units or storage units. Memory 4 may be or may include a plurality of possibly different memory units. Memory 4 may be a computer or processor non-transitory readable medium, or a computer non-transitory storage medium, e.g., a RAM. In one embodiment, a non-transitory storage medium such as memory 4, a hard disk drive, another storage device, etc. may store instructions or code which when executed by a processor may cause the processor to carry out methods as described herein.

[0054] Executable code 5 may be any executable code, e.g., an application, a program, a process, task, or script. Executable code 5 may be executed by processor or controller 2 possibly under control of operating system 3. For example, executable code 5 may be an application that may categorize an ROI of a tissue as further described herein. Although, for the sake of clarity, a single item of executable code 5 is shown in Fig. 1, a system according to some embodiments of the invention may include a plurality of executable code segments similar to executable code 5 that may be loaded into memory 4 and cause processor 2 to carry out methods described herein.

[0055] Storage system 6 may be or may include, for example, a flash memory as known in the art, a memory that is internal to, or embedded in, a micro controller or chip as known in the art, a hard disk drive, a CD-Recordable (CD-R) drive, a Blu-ray disk (BD), a universal serial bus (USB) device or other suitable removable and / or fixed storage unit. Data pertaining to a depicted, or scanned tissue may be stored in storage system 6 and may be loaded from storage system 6 into memory 4 where it may be processed by processor or controller 2. In some embodiments, some of the components shown in Fig. 1 may be omitted. For example, memory 4 may be a non-volatile memory having the storage capacity of storage system 6. Accordingly, although shown as a separate component, storage system 6 may be embedded or included in memory 4.

[0056] Input devices 7 may be or may include any suitable input devices, components, or systems, e.g., a detachable keyboard or keypad, a mouse and the like. Output devices 8 may include one or more (possibly detachable) displays or monitors, speakers and / or any other suitable output devices. Any applicable input / output (RO) devices may be connected to Computing device 1 as shown by blocks 7 and 8. For example, a wired or wireless network interface card (NIC), a universal serial bus (USB) device or external hard drive may be included in input devices 7 and / or output devices 8. It will be recognized that any suitable number of input devices 7 and output device 8 may be operatively connected to Computing device 1 as shown by blocks 7 and 8.

[0057] A system according to some embodiments of the invention may include components such as, but not limited to, a plurality of central processing units (CPU) or any other suitable multi-purpose or specific processors or controllers (e.g., similar to element 2), a plurality of input units, a plurality of output units, a plurality of memory units, and a plurality of storage units.

[0058] The term neural network (NN) or artificial neural network (ANN), e.g., a neural network implementing a machine learning (ML) or artificial intelligence (Al) function, may be used herein to refer to an information processing paradigm that may include nodes, referred to as neurons, organized into layers, with links between the neurons. The links may transfer signals between neurons and may be associated with weights. A NN may be configured or trained for a specific task, e.g., pattern recognition or classification. Training a NN for the specific task may involve adjusting these weights based on examples. Each neuron of an intermediate or last layer may receive an input signal, e.g., a weighted sum ofoutput signals from other neurons, and may process the input signal using a linear or nonlinear function (e.g., an activation function). The results of the input and intermediate layers may be transferred to other neurons and the results of the output layer may be provided as the output of the NN. Typically, the neurons and links within a NN are represented by mathematical constructs, such as activation functions and matrices of data elements and weights. At least one processor (e.g., processor 2 of Fig. 1) such as one or more CPUs or graphics processing units (GPUs), or a dedicated hardware device may perform the relevant calculations.

[0059] Reference is now made to Fig. 2, which depicts a generalized view of a system 10 for categorizing an ROI of a tissue, according to some embodiments.

[0060] As shown in Fig. 2, system 10 may include at least three main modules, denoted herein as scanning device 200, image analysis module 100, and user interface (UI) module 400.

[0061] As elaborated herein, scanning device 200 may be adapted to scan a portion, or an area of a tissue of interest, for the purpose of identifying anomalous tissue in the scanned area. Image analysis module 100 may be configured to collaborate with scanning device 200 so as to identify such anomalies, and user interface (UI) module 400 may be adapted to provide notifications, regarding the scanned area, and receive data from a user, for example to apply specific configurations to system 10, and improve a training of one or more modules of system 10.

[0062] Arrows in Fig. 2 may represent flow of one or more data elements to and from system 10 and / or among modules or elements of system 10. Some arrows have been omitted in Fig. 2 for the purpose of clarity.

[0063] According to some embodiments of the invention, each of modules 100, 200, and 400 may be implemented as a software module, a hardware module, or any combination thereof. For example, each of modules 100, 200, and 400 may be, or may include a computing device such as element 1 of Fig. 1, having respective controllers or processors 101, 201, and 401. In such embodiments, controllers or processors 101, 201, and 401 (e.g., such as controller / processor 1 of Fig. 1) may be adapted to collaborate in order to execute one or more modules of executable code (e.g., element 5 of Fig. 1), so as to categorize an ROI of a tissue 70, as further described herein.

[0064] Additionally, or alternatively, modules 100, 200, and 400 may be integrated into a single platform, having common resources such as a common computing device 1, and one or more common controllers or processors 101, 201, and 401, in order to categorize an ROI of a scanned tissue 70, as further described herein.

[0065] It may be appreciated by a person skilled in the art that other such variations in configuration may also be possible. Therefore, reference to processors / controllers 101, 201 and 401 may be used herein interchangeably.

[0066] Additionally, or alternatively, at least some of the computations of processors / controllers 101, 201 and 401 may be done on a distributed, cloud-based computing platform. For example, as elaborated herein, image analysis 100 may include one or more modules (e.g., elements 140, 150, 160 of Fig. 4) that may include models of Artificial Intelligence (Al), or Machine Learning (ML). In such embodiments, the computational functionality of these modules may be implemented online, by one or more cloud-based computational servers, as known in the art.

[0067] As shown in Fig. 2, scanning device 200 may include a thermal element 210, associated with at least one processor 201. Processor 201 may be further configured to control thermal element 210 so as to preheat, or precool the scanned area, prior to, or during scanning.

[0068] For example, thermal element 210 may include a heat source, configured to radiate heat onto inspected tissue 70, thereby preheating tissue 70. Additionally, or alternatively, thermal element 210 may include a cryogenic element, configured to precool the inspected tissue 70 as known in the art. System 10 may scan the inspected tissue 70 during this preheating or precooling phase, and / or following the preheating or precooling phase, to identify anomalies in inspected tissue 70, as elaborated herein.

[0069] As shown in Fig. 2, scanning device 200 (or “scanner 200”) may include a thermal imaging device 220, such as an Infrared (IR) scanning device, configured to obtain a time- wise sequence of thermal (e.g., IR) images 220’. Sequence 220’ may depict an area of tissue 70 that may be preheated, or precooled by thermal element 210. In other words, sequence 220’ may represent thermal changes in the depicted area 70’ over a predetermined measurement period. Scanning device 200 may subsequently transmit, or forward sequence 220’ to image analysis module 100 for further processing, as elaborated herein.

[0070] Reference is also made to Fig. 3, which is a graph depicting typical evolution of temperature at a specific location on a preheated, cutaneous tissue 70.

[0071] Reference is also made to Fig. 4, which is a block diagram, depicting image analysis module 100 (or “analysis module 100”), which may be included in system 10 for categorizing an ROI of a tissue 70, according to some embodiments of the invention.

[0072] As shown in Fig. 4, analysis module 100 may include a pixel vector module 190, configured to compute a plurality of pixel vectors 190PV in the sequence of thermal (e.g., IR) images 220’. Each pixel vector 190PV may include pixels of sequence 220’ that correspond to substantially the same location in the depicted area 70’ of tissue 70, along the measurement period. In other words, pixel vector 190PV may include a plurality of member pixels values, e.g., pixels that had been acquired at different times in the measurement period.

[0073] According to some embodiments, scanner 200 may be held such that a specific pixel may depict substantially the same location in tissue 70 over time. Additionally, or alternatively, pixel vector module 190 may apply a motion compensation function on scan 220’ so as to omit motion artifacts in sequence 220’, thereby producing pixel vectors 190PV that represent stable monitoring of temperature evolution at a specific location of tissue 70.

[0074] As shown in the example of Fig. 3, pixel vector module 190 may calculate temperature values of a specific location based on values of brightness of a specific pixel of sequence 220’, as known in the art. Pixel vector module 190 may subsequently produce pixel vectors 190PV as a time-wise vector of measured temperature values.

[0075] Pixel vectors 190PV may include representation of one or more temperature phases. As shown in the example of Fig. 3, pixel vector 190PV may include a first phase, in which tissue 70 may be preheated or precooled, e.g., between Heating Point (HP) and decay point (DP). This phase is denoted herein as a “heating phase” but may also refer to a period in which tissue 70 is precooled. Additionally, or alternatively, pixel vector 190PV may include a decay phase, in which tissue 70 may restore its original temperature. For example, the decay phase may begin at a point where preheating or precooling has ceased (DP), and last for a measurement period of a predefined duration.

[0076] According to some embodiments, analysis module 100 may include a thermal feature extraction module 140, configured to calculate, for one or more (e.g., each) pixel vector 190PV one or more respective values of thermal features 140TF. For each pixel vector190PV, the values of thermal features 140TF may be characteristic of thermal changes over time in the corresponding location, as depicted in the example of Fig. 3.

[0077] Reference is also made to Figs. 5A, 5B, and 5C which are graphs showing examples of thermal features 140TF, which may be calculated, and analyzed by embodiments of the invention, to categorize an ROI of the tissue 70, according to some embodiments of the invention.

[0078] According to some embodiments, thermal features module 140 may apply an initial normalization, or calibration function on at least one specific pixel vector 190PV as elaborated herein, to obtain a calibrated version 190PV’ of pixel vector 190PV. For example, calibrated pixel vector values 190PV’ may be calculated according to equation Eq. 1 below: Calib_temp = (Meas_temp - TempHP) / (TempDP - TempHP), where:Meas_temp represents the temporal, measured temperature pixel vector 190PV,Calib temp represents the calibrated, or normalized version 190PV’ of pixel vector 190PV, TempHP represents the measured temperature at the HP point (e.g., onset of cooling or heating phase), andTempDP represents the measured temperature at the DP point (e.g., onset of decay phase).

[0079] It may be appreciated that system 10 may use measured temperature pixel vector 190PV and / or the normalized version 190PV’ of pixel vector 190PV to extract thermal features, and ascertain an anomaly category or classification 160AC, as elaborated herein.

[0080] According to some embodiments, thermal features module 140 may be configured to calculate at least one thermal feature 140TF that pertains to steady-state temperature TempHP, e.g., temperature prior to, or at the onset of heating or cooling of the inspected tissue.

[0081] Additionally, or alternatively, thermal features module 140 may be configured to calculate at least one thermal feature 140TF that pertains to a linear approximation of a section of the decay phase.

[0082] In other words, thermal features module 140 may analyze a specific pixel vector 190PV (or the calibrated version 190PV’ of the specific pixel vector 190PV), to calculate thermal feature 140TF as a parametric, linear approximation of pixel vector 190PV.

[0083] As shown in the example of Fig. 5A, the linear approximation may be described as a line which connects between two predefined points on pixel vector 190PV (190PV’), according to the equation Eq. 2 below:Eq . 2Y = a + b * X, where:‘X’ denotes time in the decay phase, represented as a frame index of sequence 220’, ‘ Y’ is a temperature value, derived from the brightness value of pixel vector 190PV, and ‘a’ and ‘b’ are the coefficients of the linear approximation.

[0084] According to some embodiments, thermal feature value 140TF may be selected as a coefficient of the linear approximation. In the example of Fig. 5A and Eq. 2, thermal feature value 140TF may be selected as the ‘a’ and / or ‘b’ coefficients of the linear approximation.

[0085] Additionally, or alternatively, thermal features module 140 may be configured to calculate at least one thermal feature 140TF that pertains to a polynomial representation of the decay phase.

[0086] For example, and as depicted in Fig. 5B, thermal features module 140 may apply a polynomial fitting algorithm on the specific pixel vector 190PV, or on the calibrated version 190PV’ of the specific pixel vector 190PV, to fit values of member pixels (e.g., pixels acquired at different times) into a polynomial function. In the example of Fig. 5B, the polynomial fitting function may be described as a function which fits between X and Y, according to the equation Eq. 3 below:Eq. 3Y = a + b * (1-X)A4, where:‘X’ denotes time in the decay phase, represented as a frame index of sequence 220’,‘Y’ is a temperature value, derived from the brightness value of pixel vector 190PV / 190PV’, and‘a’ and ‘b’ are the coefficients of the fitted polynomial.

[0087] In the example of Fig. 5B, the decay phase from which thermal features module 140 extracts coefficients ‘a’ and ‘b’ is defined between DP+5 (e.g., 5 frames after the DP point) and DP+255 (e.g., 255 frames after the DP point).

[0088] Additionally, or alternatively, thermal feature value 140TF may be selected as a coefficient of the fitted polynomial. In the example of Fig. 5B and Eq. 3, thermal feature value 140TF may be selected as ‘b’, e.g., a coefficient of a fourth-degree variable of the polynomial function. It may be appreciated that other polynomial functions (e.g., other than Eq. 3) may be used to perform the polynomial fitting, and other, respective polynomial coefficients may be selected as values of thermal features 140TF.

[0089] Additionally, or alternatively, thermal features module 140 may be configured to calculate at least one thermal feature 140TF that represents a rate or extent of heating (or cooling) during the heating phase.

[0090] For example, as depicted in Fig. 5C, thermal features module 140 may analyze pixel vector 190PV (or calibrated version 190PV’ of the specific pixel vector 190PV), to calculate a heating (or cooling) value thermal feature 140TF, representing an extent by which the tissue was heated (or cooled) at the corresponding location, e.g., between HP and (DP+5, 5 frames after DP). Additionally, or alternatively, thermal features module 140 may calculate a heat rate thermal feature 140TF, representing a rate of heating of the tissue at the corresponding location.

[0091] Additionally, or alternatively, thermal features module 140 may be configured to calculate at least one thermal feature 140TF that pertains to a Fourier transform representation of the decay phase.

[0092] For example, thermal features module 140 may analyze pixel vector 190PV (or calibrated version 190PV’ of the specific pixel vector 190PV), to calculate one or more Fourier series coefficients of pixel vector 190PV. Thermal features module 140 may then select the thermal feature value 140TF as a Fourier series coefficient from the one or more calculated Fourier series coefficients.

[0093] As shown in Fig. 4, image analysis module 100 may utilize thermal feature value 140TF in one or more ways to ascertain an anomaly class or category 160AC of a region of interest (ROI).

[0094] For example, and elaborated further herein, image analysis module 100 may use one or more individual thermal feature values 140TF, pertaining to respective, specific pixel vectors 190PV, as input for a classifier 160, to predict the anomaly category 160AC. Additionally, or alternatively, image analysis module 100 may use one or more individual thermal feature values 140TF (e.g., steady state temperature) to determine one or more spatial features 120SF (e.g., representing distribution of steady state temperature), and present the one or more spatial features 120SF as input for classifier 160, to predict anomaly category 160AC. In yet another example, image analysis module 100 may use one or more individual thermal feature values 140TF (e.g., heating value or heating rate) to identify one or more regions denoted herein as cold dots (CD) 130CD, and introduce geometric (e.g., shape, size) and / or spatial (e.g., spatial distribution) characteristics of cold dots 130CD asinput for classifier 160, to predict anomaly category 160 AC. Other such options and combinations for using thermal feature values 140TF are also possible.

[0095] According to some embodiments, image analysis module 100 may include a spatial filter module, denoted in Fig. 4 as skin mask module 110. As elaborated herein (e.g., in relation to Fig. 11), skin mask module 110 may be configured to analyze thermal features 140TF of the plurality of pixel vectors 190PV, to produce a skin mask 110SM. Skin mask 110SM may be, or may define a portion of the depicted area 70’ of tissue 70, that may include pixel vectors 190PV that can be analyzed in order to ascertain anomaly of tissue 70.

[0096] For example, tissue 70 may be a cutaneous tissue of a scalp, or a furry animal. In such embodiments, skin mask 110SM may be, or may include a portion of thermal image(s) 220’ that represents a portion of tissue 70 where the skin is properly exposed before thermal imaging element 220. Skin mask 110SM may thus define an area which image analysis module 100 may analyze to reliably categorize an anomaly (e.g., malignancy, or inflammation) of tissue 70.

[0097] According to some embodiments, image analysis module 100 my apply skin mask 110SM as a spatial filter on the plurality of pixel vectors, to obtain a subset of candidate pixel vectors 110CPV.

[0098] It has been experimentally observed by the inventors, that regions of excess heat dissipation may appear in the depicted area 70’ of tissue 70 during the decay period. In embodiments where thermal elements 210 are configured to preheat tissue 70, such excess heat dissipation results in apparent dots having a lower temperature in relation to their immediate surroundings, thus dubbed “cold dots” (CDs).

[0099] Reference is also made to Figs. 6A and 6B, which are examples of thermal images of cutaneous tissues. Some of the CDs in Figs. 6A and 6B are highlighted by white arrows.

[0100] As shown in Figs. 6A and 6B, CDs have been shown to appear in one of two main patterns: a focal pattern, as depicted in Fig. 6A, and a diffused pattern as depicted in Fig. 6B. Each of these patterns may indicate a specific category or type of an anomaly, such as a type of malignancy of tissue 70.

[0101] For example, Fig. 6A depicts a focal, or radial accumulation of cold dots in a region of a mass, defined by the inner circle ROI 40R of Fig. 6A. In contrast, few or no CDs are pronounced in a healthy region, marked beyond the external circle ROI 42R. It has been experimentally observed by the inventors that such characteristics may be indicative of aMast cell tumor. Among the studied, thermally visible lesions, 45% (20 / 42) of mast cell tumors have presented this feature. Among all other studied, thermally visible lesions, 6.8% (19 / 280) have presented this feature. It may be appreciated that this prevalence is statistically significant, having a p-value smaller than 0.01, Fisher exact test.

[0102] In another example, Fig. 6B depicts a diffused or scattered pattern of CDs in both the mass, defined by the inner circle ROI 40R of Fig. 6B, and in the healthy region, defined beyond, or outside the external circle ROI 42R of Fig. 6B. It has been experimentally observed by the inventors that such characteristics may be indicative of a Lipoma tumor. Among the studied, thermally visible lesions, 46% [54 / 118] of Lipoma tumors have presented this feature. Among all other studied, thermally visible lesions, 20% [42 / 206] have presented this feature. It may be appreciated that this prevalence is statistically significant, having a p-value smaller than 0.01, Fisher exact test.

[0103] As shown in Fig. 4, image analysis module 100 may include a CD identifier module 130, configured to determine, or identify one or more cold dots (CDs) 130CD in the depicted area 70’ of tissue 70. Each CD 130CD may include one or more (e.g., a plurality) of member candidate pixel vectors 110CPV.

[0104] According to some embodiments, CD identifier module 130 may identify CDs 130CD based on one or more thermal feature 140TF values of candidate pixel vectors 110CPV.

[0105] For example, CD identifier module 130 may define a sliding window, of a predefined dimension (e.g., 13x13 pixels), and may use the sliding window to scan (e.g., raster scan) the depicted area 70’ (e.g., the candidate pixel vectors 110CPV of depicted area 70’).

[0106] In each position of the sliding window, CD identifier module 130 may examine a candidate pixel vectors 110CPV located at that position, to determine whether the relevant candidate pixel vectors 110CPV pertains to a CD.

[0107] CD identifier module 130 may calculate at least one statistical metric value 130MV representing a difference in temperature between (i) the candidate pixel vector 110CPV at the center position of the sliding window, and (ii) other candidate pixel vectors 110CPV in the window (e.g., candidate pixel vectors 110CPV at the border of the sliding window), along the measurement period. In other words, statistical metric value 130MV may represent a statistical difference in rate, or in extent, by which a temperature of a specificexamined candidate pixel vector 110CPV changes along the measurement period, in relation to the temperature of other candidate pixel vectors 1 lOCPVs in its surroundings.

[0108] CD identifier module 130 may subsequently identify the studied candidate pixel vector 110CPV as pertaining to a CD, when the calculated at least one statistical metric value surpasses a predetermined threshold, for at least a predetermined duration.

[0109] For example, CD identifier module 130 may study a first candidate pixel vector 110CPV (e.g., in the center of the sliding window), in relation to one or more second candidate pixel vectors 110CPV (e.g., at the border of the sliding window). CD identifier module 130 may use statistical metric value 130MV to compare between the extent of cooling of the first 110CPV and the one or more second 1 lOCPVs.

[0110] For example, CD identifier module 130 may identify a condition in which extent of cooling of the first candidate pixel vector 110CPV during the decay phase, exceeds a mean (statistical metric value 130MV) extent of cooling of the one or more second candidate pixel vectors 110CPV. If this condition persists beyond a minimal duration (e.g., beyond a minimal number of frames in sequence 220’), then CD identifier module 130 may identify the first candidate pixel vector 110CPV as likely to pertain to, or be associated with a CD.

[0111] Additionally, or alternatively, CD identifier module 130 may apply additional logic to define specific regions in area 70’ as CDs. For example, CD identifier module 130 may define a CD as an aggregation of a minimal number of likely candidate pixel vectors 110CPV within a predefined region. For example, CD identifier module 130 may require at least N candidate pixel vectors 110CPV to be identified within a region of the sliding widow, so as to define these candidate pixel vectors 110CPV as including a CD.

[0112] As shown in Fig. 4, analysis module 100 may include a CD classification module 150 (or “CD classifier 150”, for short), that may be, or may include a machine-learning (ML) based classification model.

[0113] During a training phase, CD classifier 150 may be pretrained to predict a risk score 150RS, representing a probability that a CD of interest includes an anomaly (e.g., is malignant). For example, CD classifier 150 may receive a training dataset that represents at least one training CD. The training dataset may include one or more thermal feature 140TF values of candidate pixel vectors 110CPV that are members of the at least one training CD. The training dataset may be annotated, in a sense that the at least one training CD may be associated with annotation data 40AN, or label data (e.g., from an expert, via input device 7of Fig. 1). The annotation, or label data 40AN may, for example, represent a probability or ground-truth certainty that the at least one training CD pertains to anomalous tissue such as a malignant tissue, or to a healthy tissue. During the training phase, analysis module 100 may use the annotation data 40AN as supervisory information as known in the art, to train CD classifier 150, so as to predict risk score 150RS based on the thermal feature 140TF values of the at least one training CD. In other words, CD classifier 150 may be trained to classify CDs based on their probability of anomaly (e.g., inflammation or malignancy).

[0114] Additionally, or alternatively, CD classifier 150 may be trained further based on thermal feature 140TF values representing specific characteristics of depicted healthy tissues. For example, CD classifier 150 may include an ML-based normalization module 150NR. System 10 may (e.g., during a training stage), train normalization module 150NR on unsupervised scans, to predict thermal feature 140TF values of healthy tissue (e.g., in ROI 42R) of different depicted specimens, e.g., different animal types, breeds, ages, and / or colours.

[0115] In other words, normalization module 150NR may predict the behavior of normal tissue of a specific target animal under specific heating or cooling protocols. CD classifier 150 may use this predicted thermal feature 140TF values of healthy tissue as additional input. CD classifier 150 may thus classify CDs based on their probability of anomaly, while allowing normalization according to specimen- specific characteristics, such as animal types, breeds, ages, and / or colours.

[0116] During a subsequent inference, or applicative phase, analysis module 100 may apply classifier 150 on the thermal feature 140TF values of one or more member candidate pixel vectors 110CPV of one or more (e.g., each) identified CD 130CD. In other words, analysis module 100 may apply classifier 150 on each identified CD 130CD, to predict a risk score 150RS representing a probability that the identified CD 130CD is anomalous (e.g., malignant).

[0117] Additionally, or alternatively, analysis module 100 may include an ROI classification module 160 (or “classifier 160”, for short), which may be, or may include an ML-based classification model. Classifier 160 may be configured, or pretrained to categorize, or classify an anomaly (e.g., malignancy, benign tumor, inflammation, etc.) of depicted area 70’, based on the risk scores 150RS of one or more identified CDs 130CD.

[0118] As depicted in Fig. 2, and shown in the examples of Fig. 6A and 6B, analysis module 100 may receive, e.g., from an input device (e.g., input device 7 of Fig. 1) of UI module 400 a definition of at least one ROI 40R / 42R within depicted area 70’ of tissue 70.

[0119] Taking the example of Fig. 6A, ROI 40R and ROI 42R may be received by an expert, such as a physician, a veterinary doctor or care giver. ROI 40R (e.g., the area within the internal circle) may define or mark a region suspected to be anomalous or malignant (e.g., a malignant lesion or mass). ROI 42R (e.g., the area outside the outer circle) may define, or mark a region that is presumed to be healthy.

[0120] As elaborated herein, CD identifier module 130 may identify one or more first CDs in ROI 40R, e.g., within the region suspected to be anomalous (e.g., malignant). Additionally, or alternatively, CD identifier module 130 may identify one or more second CDs in ROI 42R, e.g., within the region expected to be healthy. Analysis module 100 may subsequently categorize the ROI 40R as anomalous (e.g., malignant), based on the risk scores of at least one of (i) the one or more first CDs, and (ii) the one or more second CDs.

[0121] Additionally, or alternatively, analysis module 100 may include a spatial features extraction module 120, configured to calculate one or more spatial feature 120SF values, characteristic of spatial distribution of the one or more first CDs and the one or more second CDs.

[0122] For example, spatial features 120SF may include a numerical value of density, representing compactness or spreading of CDs 130CD across 40R and / or 42R, a CD morphology value, numerically representing a size and / or a shape of the CDs 130CD in 40R and / or 42R, a first spatial distribution value, representing distribution of CDs 130CD with similar thermal features 140TF within 40R and / or 42R, a second spatial distribution value, representing distribution of CDs 130CD with similar risk scores, and the like.

[0123] Additionally, or alternatively, spatial features extraction module 120 may be configured to calculate one or more spatial feature 120SF value representing existence of an annular structure of pixel vectors 190PV, surrounding a mass ROI (40R) in the sequence of thermal image 220’.

[0124] Reference is now made to Figs. 7A - 7C, where Fig. 7A is an image depicting an example of mass ROIs 40R, Fig. 7B is a thermal (e.g., IR) image 220’ of cutaneous tissues surrounding the mass ROIs 40R of Fig. 7A, and Fig. 7C is a graph describing analysis of thermal image 220’, and / or analysis of pixel vectors 190PV of a sequence of thermal images220’, to categorize the mass ROIs 40R of Fig. 7A, according to some embodiments of the invention.

[0125] The inventors have observed that mass ROIs 40R typically include a cutaneous component, and a sub-cutaneous component. The cutaneous component would typically be visibly deformed, or abnormally structured, and would typically be cooler than its surrounding region, marked 41R in Fig. 7A and 7B. As shown in Fig. 7B, surrounding region 41R may form an annular thermal region, or thermal ring, characterized by a different (e.g., warmer) temperature. This annular region 41R is clearly demonstrated in two instances in Fig. 7B.

[0126] Additionally, the inventors have observed that different levels of manifestation of the annular thermal region 41R may indicate different categories of anomalies. For example, a “weak” thermal ring, e.g., one that is characterized by a mild thermal difference from its surroundings, may indicate inflammation in a cutaneous region. In another example, a “strong” thermal ring, e.g., one that is characterized by a more significant thermal difference from its surroundings, may indicate a malignant, subcutaneous region.

[0127] Accordingly, spatial feature 120SF value may be, or include a numerical value representing characteristics of an annular thermal region 41R surrounding ROI 40R (e.g., the indicated mass).

[0128] As shown in Fig. 7C, features extraction module 120 may calculate a profile of temperature in region 41R surrounding ROI 40R. The profile in Fig. 7C is depicted as a grey region, and may represent a distribution of temperature in increasing, concentric rings, surrounding the center of ROI 40R (e.g., the mass), in thermal image 220’. Features extraction module 120 may then calculate spatial feature 120SF as the area of the graph (e.g., the distributed temperature) between the ROI 40R radius defining the mass, and a predetermined radius of ROI 42R (e.g., region 41R between 40R and 42R).

[0129] Additionally, or alternatively, spatial feature 120SF may be a numerical value representing a difference or comparison between the area of thermal distribution within 41R described above, and the average temperature (denoted by a black line in Fig. 7C) plus a standard deviation value (denoted by a dashed line in Fig. 7C). The inventors have found that high values of such a spatial feature 120SF may indicate malignancy of the mass, whereas low values of such a spatial feature 120SF may indicate a condition of inflammation surrounding the mass.

[0130] According to some embodiments, ROI classification module 160 may be configured to classify, or categorize ROI as anomalous (e.g., malignant), further based on the spatial feature values 120SF.

[0131] In other words, system 10 may apply (e.g., during an inference stage) a pretrained ML based classifier 160 on input data to predict an anomaly category 160AC of ROI 40R / 40CR. The input data for pretrained ML based classifier 160 may include, for example (i) risk scores 150RS of one or more respective CDs 130CD in 40R and / or 42R, and / or (ii) one or more spatial feature values 120SF of one or more respective CDs 130CD in 40R and / or 42R. In such embodiments, anomaly category 160 AC may include, for example a type of a malignant tumor in ROI 40R, a type of a benign tumor in ROI 40R, an inflamed tissue in ROI 40R, and a healthy tissue in ROI 40R.

[0132] Additionally, or alternatively, system 10 may apply pretrained ML based classifier 160 (e.g., during the inference stage) on the input data (e.g., risk scores 150RS and / or spatial feature values 120SF) to predict a confidence level value 160CL. Confidence level value 160CL may include, for example a malignancy confidence level value, representing probability that ROI 40R / 40CR is malignant, a benign confidence level value, representing probability that ROI 40R / 40CR is benign, an inflammation confidence level value, representing probability that ROI 40R / 40CR is inflamed, and a health confidence value, representing probability that that ROI 40R / 40CR is healthy.

[0133] Pertaining to the example of Fig. 6A, analysis module 100 may analyze location of the one or more first CDs and one or more second CDs, to identify a focal pattern, or a focal distribution in, or around ROI 40R, to categorize ROI 40R as a malignant, Mast tumor.

[0134] Pertaining to the example of Fig. 6B, analysis module 100 may analyze location of the one or more first CDs and one or more second CDs, to identify a diffused pattern, or a diffused distribution of CDs in ROI 40R, as well as in ROI 42R, to categorize ROI 40R as a Lipoma tumor.

[0135] According to some embodiments, system 10 may employ a supervised training scheme to pretrain ML based classifier 160, so as to predict confidence level value 160CL and / or anomaly category 160AC of ROI 40R / 40CR.

[0136] For example, during a training stage, image analysis module 100 may receive (e.g., via input device 7, from UI module) an annotated training thermal image sequence 220’. The term “annotated” may be used in this context to indicate that at least one image220’ of the training thermal image 220’ sequence may include a respective label or annotation 40AN. Annotation 40AN may include a definition of an ROI 40R / 42R, and an anomaly category 160 AC of the received ROI 40R / 42R in the respective at least one image 220’.

[0137] Image analysis module 100 may determine one or more CDs 130CD in the training thermal image 220’ sequence, and may calculate risk scores 150RS of the one or more CDs 130CD in the training thermal image 220’ sequence, as elaborated herein. Additionally, or alternatively, image analysis module 100 may calculate spatial feature values 120SF characteristic of spatial distribution of the one or more CDs 130CD in the training thermal image 220’ sequence, as elaborated herein.

[0138] According to some embodiments, image analysis module 100 may subsequently use annotation 40 AN as supervisory data, to train classifier 160, so as to predict the anomaly category 160AC and / or confidence level 160CL of the defined ROI 40R in the training thermal image 220’ sequence, based on (a) the calculated risk scores 150RS, and (b) the calculated spatial feature values 120SF.

[0139] For example, classifier 160 may be implemented as a NN architecture, that may include a plurality of weighted, interconnected neural nodes. During the training stage, image analysis module 100 may calculate a loss value representing difference between (i) an anomaly category 160 AC value (or a confidence level 160CL value) as predicted by ML model 160, and (ii) a corresponding anomaly category 160 AC value (or a confidence level 160CL value) as annotated by annotation 40AN. Image analysis module 100 may train classifier 160 by employing a backpropagation algorithm, as known in the art, to adjust weights of the NN architecture of classifier 160, thereby minimizing the calculated loss value.

[0140] Additionally, or alternatively, image analysis module 100 may include an ROI calculation module 40, configured to obtain ROI 40R and / or ROI 42R (here denoted 40CR and 42CR, respectively) automatically, based on thermal features 140TF of thermal images(s) 220’.

[0141] For example, and as depicted in Fig. 6A, ROI calculation module 40 may mark a first ROI 40R (now 40CR) of a suspected mass as a circle surrounding a region of skin mask 110SM. This circle may define a region where a combination of values (e.g., a weightedsum) of one or more thermal features 140TF (e.g., extent of heating, Fourier coefficient, etc.) of member candidate pixel vectors 110CPV exceeds a predetermined threshold.

[0142] ROI calculation module 40 may proceed to mark a second ROI 42R (now 42CR), as a region beyond a second circle, concentric with the circle of ROI 40R (40CR). ROI 42R (42CR) may define a region where the combination of values of the one or more thermal features 140TF of member candidate pixel vectors 110CPV falls below a second, predetermined threshold, and thus assumed to be a healthy region within skin mask 110SM.

[0143] Additionally, or alternatively, ROI calculation module 40 may be, or may include an ML-based model, adapted to determine the regions of interest automatically. In other words, ROI calculation module 40 may produce a first boundary, defining ROI 40R (40CR), suspected as anomalous (e.g., including a tumor), and a second boundary, defining a region ROI 42R (42CR) beyond ROI 40R (40CR), that is presumed healthy. As elaborated herein, system 100 may subsequently extract features (e.g., spatial features 120SF, thermal features 140TF and / or morphological features 180MF) of ROIs 42R (42CR) and / or 40R (40CR), to identify, and / or categorizing anomaly of the depicted area.

[0144] According to some embodiments, ROI calculation module 40 may (e.g., during a training stage) receive a training dataset 40DS that may include a plurality of annotated thermal 220’ and / or visible light images 230’. The images (2207230’) may be annotated in a sense that they may be associated with labels, or markings (e.g., provided by a human expert) of ROIs 40R, encircling a suspected tumor, and / or 42R representing a peripheral region, assumed to be healthy.

[0145] As known in the art, system 10 may subsequently utilize a training scheme (e.g., a backward propagation scheme), to train ROI calculation module 40, while using training dataset 40DS as supervisory information.

[0146] In a subsequent, inference stage, ROI calculation module 40 may be configured to receive target thermal 220’ and / or visible light images 230’. Based on the training, ROI calculation module 40 may automatically classify, or determine ROIs 42R (42CR) and 40R (40CR) in target thermal 220’ and / or visible light images 230’.

[0147] It may be appreciated that the training stage of ROI calculation module 40 may precede a subsequent inference of pretrained ROI calculation module 40 on incoming images 2207230’. Additionally, or alternatively, the training and inference stages of ROIcalculation module 40 may be intermittent, allowing system 100 to refine the training of ROI calculation module 40 iteratively, over time.

[0148] As shown in Fig. 2, scanning device 200 may include, or may be associated with a visible light imaging device 230, also referred to herein as a camera device 230. Scanning device 200 may use camera 230 to obtain or receive a visible light image 230’ of area 70’ substantially concurrent with the obtaining of thermal image 220’ from thermal imaging device 220.

[0149] As shown in Fig. 4, image analysis module 100 may include a visual image processing module 180, configured to analyze visible-light image 230’ and extract therefrom one or more morphological features 180MF representing a morphology of the depicted area 70’ in visible light.

[0150] According to some embodiments, visual image processing module 180 may apply a registration algorithm as known in the art, to align visible-light images 230’, so as to mitigate effect of relative movement between imaging device 230 and the depicted area of interest.

[0151] According to some embodiments, ML model 160 may be pretrained, based on annotation 40 AN to predict anomaly category 160 AC value, and / or confidence level 160CL value further based on the one or more morphological features 180MF.

[0152] During an inference stage, image analysis module 100 may introduce the one or more morphological features 180MF as input for trained ML model 160, which may predict anomaly category 160 AC value, and / or confidence level 160CL value further based on the one or more morphological features 180MF.

[0153] In other words ML model 160 may predict an anomaly category 160AC of the ROI and / or confidence level 160CL value based on (i) the risk scores 150RS, (ii) the one or more spatial feature values 120SF, and / or (iii) the at least one morphological feature 180MF, to predict an anomaly category 160 AC of the ROI 40R (40CR), which may include a type of a malignant tumor, a type of a benign tumor, an inflamed tissue and a healthy tissue.

[0154] Reference is now made to Figs. 8A-8D which are images depicting an example of analysis of a visible light image 230’, to obtain cutaneous morphological features 180MF, according to some embodiments of the invention.

[0155] In the example of Figs. 8A-8D, image analysis module 100 may produce morphological features 180MF as layered information, indicative of presence, and / or type of a cutaneous anomaly category 160 AC.

[0156] A first such morphological feature 180MF may represent a colour, or a shape or pattern of colour in ROI 40R (40CR). Another morphological feature 180MF may represent presence, and / or a pattern of hair or fur follicles in the inspected ROI.

[0157] For example, visual image processing module 180 may employ a segmentation algorithm to define a region in ROI 40R having specific, predefined color spectrum characteristics, or such characteristics that are significantly different from its surroundings. As shown in Fig. 8A, the left part of the inspected tissue is brighter, and may have (e.g., when shown in color) a distinct reddish hue, in comparison to the left part of that tissue.

[0158] In another example, and as shown in Figs. 8B and 8C, visual image processing module 180 may employ an edge detection algorithm, to identify patterns, such as appearance, or density of follicles in the inspected region. In this example, the right part of the inspected tissue contains more follicles than the left part of that tissue. This may be indicative of a cutaneous, or sub cutaneous anomaly 160AC on the left part of the examined tissue.

[0159] As shown in Fig. 8D, visual image processing module 180 may utilize a predetermined ROI 40R / 40CR and / or a skin mask 110SM to calculate, or relate to (e.g., propagate to ML model 160) morphological features 180MF that are within ROI 40R / 40CR and / or skin mask 110SM.

[0160] The inventors have experimentally observed that certain textures or colours of the depicted region may correspond to different percentages of a mass being either cutaneous or subcutaneous. Accordingly, morphological features 180MF may further represent a texture, or color, or a spatial change in texture or colour, that represents the cutaneous or subcutaneous characteristics of a mass in ROI 40R. Additionally, or alternatively, morphological features 180MF may include a predicted percentage of a mass being either cutaneous or subcutaneous, herein denoted percentage 180MF-P.

[0161] Additionally, or alternatively, ROI classifier 160 may include two, separate classification models, denoted 160-C, and 160-SC. Each of classification models 160-C and 160-SC may be trained to categorize, or classify an anomaly (e.g., malignancy, benign tumor, inflammation, etc.) of depicted area 70’, based on the risk scores 150RS of one ormore identified CDs 130CD. Classification model 160-C (Cutaneous) may be trained to categorize, or classify the anomaly, under the assumption that the anomaly may be a cutaneous mass. Classification model 160-SC (Subcutaneous) may be trained to categorize, or classify the anomaly, under the assumption that the anomaly may be a subcutaneous mass. The outcome of both classification models 160-C and 160-SC may be averaged, weighted according to percentage 180MF-P.

[0162] In other words, when percentage 180MF-P indicates that a large percentage of a ROI 40R is suspected to be a cutaneous malignancy, a large weight may be applied to the outcome of classification model 160-C to produce anomaly category 160 AC and / or confidence level 160CL value.

[0163] In a complementary manner, when percentage 180MF-P indicates that a large percentage of a ROI 40R is suspected to be a subcutaneous malignancy, a large weight may be applied to the outcome of classification model 160-SC to produce anomaly category 160 AC and / or confidence level 160CL value.

[0164] Reference is now made to Fig. 9 A which is an image depicting thermal characteristics, e.g., heat distribution on a flat surface, in thermal image 220’ produced by a thermal imaging device 220. According to some embodiments of the invention, thermal features’ module 140 receive one or more such thermal image 220’, e.g., during a heating phase, and / or a decay phase. Thermal features’ module 140 may calibrate thermal features 140TF based on the received thermal image 220’, e.g., according to the thermal distribution. In other words, thermal features’ module 140 may calculate thermal features 140TF so as to compensate for uneven thermal distribution during the heating period and / or decay period.

[0165] As shown in Fig. 2, scanning device 200 may include a scanning light source 250, such as a laser light source 250 that may be associated with, or controlled by at least one processor 101 / 201. Processor 101 / 201 may control the scanning light source 250 to produce a scan of depicted area 70’, or project a light pattern on depicted area 70’.

[0166] For example, scanning light source 250 may produce a pattern of light, such as one or more illuminated dots on area 70’, to allow image analysis module 100 to easily register, or match between different images 2207230’ along the examination procedure, e.g., though the heating period and / or the decay period.

[0167] Reference is also made to Fig. 9B, which is a schematic image showing a pattern of light that may be produced by scanning light source 250, according to some embodimentsof the invention. The pattern of light may include, for example, a grid, a repetitive pattern of lines, and the like.

[0168] As shown in the example of Fig. 9B, the pattern of light may be affected by a topography of the scanned region (depicted area 70’). For example, a mass, or a “bump” (represented by a sphere) may affect the projected pattern of light in a first manner, and a depression in the scanned area may affect the projected pattern of light in another manner. Therefore, the projected pattern of light may include a representation of depicted area 70’.

[0169] As shown in Fig. 2, and as depicted in the example of Fig. 9B, imaging device(s) 230 / 240 of scanning device 200 may obtain a scan data element 250’ (e.g., an image) representing a topography of depicted area 70’. Thermal features module 140 may apply an image analysis algorithm on the scan data element 250’, to extract or calculate a topography data element 140TOP, representing topography of the depicted area 70’. For example, topography data element 140TOP may include a matrix, where each entry represents a relative “elevation” (e.g., vicinity to scanning device 200) of a respective location on area 70’.

[0170] It may be appreciated that points in depicted area 70’ may be heated unevenly due to their topography, where “elevated” points may be heated more than other points. According to some embodiments, thermal features’ module 140 may calibrate thermal features 140TF based on topography data element 140TOP, e.g., according to the topography of area 70’. In other words, thermal features’ module 140 may calculate thermal features 140TF so as to further compensate at least one thermal feature value for uneven heating of points in depicted area 70’, based on the topography, as represented by scan data element 250’.

[0171] As shown in Fig. 4, image analysis module 100 may include an anomaly analysis module 170, adapted to provide, e.g., via output device 8 of Fig. 1, one or more recommendations, or indications 410 based on the anomaly category 160AC and / or confidence level 160CL value.

[0172] For example, indications 410 may include an indication of anomaly category 160 AC, e.g., whether the anomaly is classified 416C as benign or malignant, and a type or diagnosis 418 thereof (e.g., lymphoma, Mast tumor, Sarcoma, etc.).

[0173] In another example, indications 410 may include a recommendation 412 for treatment of the ROI 40R. For example, anomaly analysis module 170 may determine thata biopsy may be needed based on confidence level 160CL (e.g., when confidence level 160CL value is within a predefined range).

[0174] In another example, indications 410 may include a recommendation 414 for an action, to improve the diagnosis process based on confidence level 160CL. Recommendation 414 may include, for example, a recommendation to improve shaving of the ROI 40R / 42R, prevention of movement, based on analysis of image(s) 2207230’, improvement of location of ROI 40R in image(s) 2207230’, and the like.

[0175] In another example, indications 410 may include thermal data 416T such as thermal properties of CDs in ROI 40R, a mass in ROI 40R and / or in its surroundings, a segmentation of a region according to thermal properties, and the like.

[0176] Reference is now made to Fig. 10 which is a graph depicting typical evolution of temperature at a specific preheated location of fur. By comparing the thermal evolution graph of Fig. 10 (fur) with that of Fig. 5A (skin) it is evident that the heat decay period (post DP) for fur is typically characterized by relatively rapid decay (e.g., sharper slope).

[0177] Embodiments of the invention may utilize this property, to filter-out signals obtained from furry regions in depicted area 70’.

[0178] Reference is also made to Fig. 11 which is a block diagram, depicting a skin mask module 110 which may be included in a system for categorizing an ROI of a tissue 70, according to some embodiments of the invention.

[0179] As shown in Fig. 11, skin mask module 110 may include a plurality (e.g., a cascade) of filters, each configured to omit, or filter-out pixel vectors 190PV according to respective criteria, to obtain candidate pixel vectors 110CPV.

[0180] A first such filter is denoted “spacer mask module 112” in Fig. 11. This filter may include, or may employ an image analysis algorithm, configured to detect a mask or frame that may be applied on depicted area 70’, to define a region surrounding ROI 40R / 42R. An example of such a mask is depicted in the example of Fig. 8 A. Skin mask module 110 may filter-out pixel vectors 190PV that correspond to, or depict parts of the detected mask, to allow image analysis module 100 to focus on the area of live tissue defined by the applied mask.

[0181] Another filter of pixel vectors 190PV is referred to herein as an “initial heating filter 114” (Filter 114, for short). Filter 114 may be configured to analyze specific pixel vectors 190PV of the plurality of pixel vectors, to calculate a maximal heating value ofcorresponding location in area 70’. Filter 114 may subsequently reject, or filter out specific pixel vectors 190PV when the calculated maximal heating value does not surpass a predefined threshold.

[0182] Another filter of pixel vectors 190PV is referred to herein as a “fur classifier” module 116. According to some embodiments, the examined tissue is a cutaneous, or subcutaneous tissue of a furry, or hairy animal or human subject. In such embodiments, fur classifier 116 may obtain the subset of CPVs 110CPV by analyzing a thermal feature value 140TF of a specific pixel vector 190PV, to determine consistency of the thermal feature value with heat dissipation by fur, e.g., according to sharpness of a decay slope (e.g., as shown in Fig. 10). Fur classifier 116 may reject, or filter-out the specific pixel vector 190PV based on this consistency, e.g., when the thermal parameters (e.g., slope of decay period) are characteristic with fur.

[0183] Additionally, or alternatively, fur classifier 116 may be, or may include an ML- based, classification model, pretrained to classify pixel vectors 190PV as depicting fur, or as depicting other tissue types (e.g., skin). In such embodiments, image analysis module 100 may apply the ML-based fur classification model 116 on one or more thermal feature values 140TF of the specific pixel vector, to predict a probability that the specific pixel vector depicts a fur of the animal. Filter 110 may subsequently reject the specific pixel vector 190PV based on this probability, e.g., when a probability of depicted fur exceeds a predefined threshold.

[0184] Reference is also made to Fig. 12, which is a flow diagram, depicting a method of categorizing anomaly of an ROI of a tissue 70, by at least one processor (e.g., 101 / 201 / 401 of Fig. 2) according to some embodiments of the invention.

[0185] As shown in step S1005, the at least one processor may obtain, from a thermal imaging device (e.g., 220 of Fig. 2), a sequence of thermal images 220’ depicting a preheated, or precooled area 70’ of the tissue70. The sequence of thermal images 220’ may represent thermal changes or evolution in the depicted area 70’ over a predetermined measurement period.

[0186] As shown in step S1010, the at least one processor may compute a plurality of pixel vectors (e.g., 190PV of Fig. 4) in the sequence of images 220’, where one or more (e.g., each) pixel vector 190PV may include member pixels that correspond to substantially the same location in the depicted area 70’, along the measurement period.

[0187] As shown in step S1015, the at least one processor may calculate, for each pixel vector 190PV, one or more respective thermal feature values (e.g., 140TF of Fig. 4). Thermal feature values 140TF may be characteristic of the thermal changes (e.g., preheating, precooling and / or decay) over time in the corresponding location in depicted area 70’.

[0188] As shown in steps S 1020 and S 1025, the at least one processor may apply a filter (e.g., 110 of Fig. 4 and / or Fig. 11) on the plurality of pixel vectors 190PV, to obtain a subset of candidate pixel vectors (e.g., 110CPV of Fig. 4). The at least one processor may subsequently determine one or more cold dots (e.g., 130CD of Fig. 4) in the depicted area, based on the thermal feature values 140TF of candidate pixel vectors 110CPV. Each CD may include one or more member candidate pixel vectors 190PV, as elaborated herein.

[0189] As shown in step S1030, for each CD 130CD the at least one processor may apply a classifier (e.g., 150 of Fig. 4) on the thermal feature values of the one or more member candidate pixel vectors 190PV. Classifier 150 may predict a risk score (e.g., 150RS of Fig. 4) representing a probability that the relevant CD 130CD is anomalous. Subsequently, and as shown in step S1035, the at least one processor may utilize a second classification model to categorize an anomaly of the depicted area 70’ based on the risk scores 150RS of the one or more CDs 130CD.

[0190] Unless explicitly stated, the method embodiments described herein are not constrained to a particular order or sequence. Furthermore, all formulas described herein are intended as examples only and other or different formulas may be used. Additionally, some of the described method embodiments or elements thereof may occur or be performed at the same point in time.

[0191] While certain features of the invention have been illustrated and described herein, many modifications, substitutions, changes, and equivalents may occur to those skilled in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.

[0192] Various embodiments have been presented. Each of these embodiments may of course include features from other embodiments presented, and embodiments not specifically described may include various features described herein.

Claims

CLAIMS1. A method of categorizing anomaly of a tissue by at least one processor, the method comprising: obtaining, from a thermal imaging device, a sequence of thermal images depicting a preheated, or precooled area of the tissue, said sequence representing thermal changes in the depicted area over a predetermined measurement period; computing a plurality of pixel vectors in the sequence of images, wherein each pixel vector comprises member pixels that correspond to substantially the same location in the depicted area, along the measurement period; for each pixel vector, calculating one or more respective thermal feature values, characteristic of the thermal changes over time in the corresponding location; applying a filter on the plurality of pixel vectors, to obtain a subset of candidate pixel vectors; determining one or more cold dots (CDs) in the depicted area, based on the thermal feature values of candidate pixel vectors, wherein each CD comprises one or more member candidate pixel vectors; for each CD, applying a first classifier on the thermal feature values of the one or more member candidate pixel vectors, to predict a risk score representing a probability that said CD is anomalous; and categorizing anomaly of the depicted area based on the risk scores of the one or more CDs.

2. The method of claim 1, wherein categorizing the depicted area as anomalous comprises: receiving a definition of a first region of interest (ROI) within the depicted area, suspected to be anomalous; identifying one or more first CDs in the first ROI; and categorizing the first ROI as anomalous, based on the risk scores of the one or more first CDs.

3. The method of according to any one of claims 1-2, further comprising: identifying one or more second CDs in a region peripheral of the first ROI;calculating one or more spatial feature values, characteristic of spatial distribution of the one or more first CDs and the one or more second CDs; and categorizing the first ROI as anomalous, further based on the spatial feature values.

4. The method according to any one of claims 1-3, wherein categorizing the first ROI comprises applying a pretrained, machine learning (ML) based, second classifier on (i) the risk scores, and (ii) the one or more spatial feature values, to predict an anomaly category of the first ROI, wherein said anomaly category is selected from a list of: a type of a malignant tumor, a type of a benign tumor, an inflamed tissue, and a healthy tissue.

5. The method according to any one of claims 1-4, wherein categorizing the first ROI further comprises applying the pretrained, ML based, second classifier on (i) the risk scores, and (ii) the one or more spatial feature values, to predict a confidence level value, wherein said confidence level value is selected from a list consisting of: (a) a malignancy confidence level value, representing probability that the first ROI is malignant, (b) a benign confidence level value, representing probability that the first ROI is benign, (c) an inflammation confidence level value, representing probability that the first ROI is inflamed, and (d) a health confidence value, representing probability that the first ROI is healthy.

6. The method according to any one of claims 1-5, further comprising pretraining the second, ML based classifier by: receiving an annotated training thermal image sequence, said annotation comprising: (i) a definition of a first ROI, and (ii) an anomaly category of the defined, first ROI; determining one or more CDs in the training thermal image sequence; calculating risk scores of the one or more CDs in the training thermal image sequence; calculating spatial feature values characteristic of spatial distribution of the one or more CDs in the training thermal image sequence; and using said annotation as supervisory data, to train the second classifier, so as to predict the anomaly category of the defined, first ROI in the training thermal image sequence, based on (a) the calculated risk scores, and (b) the calculated spatial feature values.

7. The method according to any one of claims 1-6, wherein categorizing the first ROI comprises: receiving a visible-light image of the depicted area; analyzing the visible-light image to extract at least one morphological feature; applying a pretrained, machine learning (ML) based, second classifier on (i) the risk scores, (ii) the one or more spatial feature values, and (iii) the at least one morphological feature, to predict an anomaly category of the first ROI, wherein said anomaly category is selected from a list consisting of: a type of a malignant tumor, a type of a benign tumor, an inflamed tissue and a healthy tissue.

8. The method according to any one of claims 1-7, wherein obtaining the subset of candidate pixel vectors comprises: analyzing a specific pixel vector of the plurality of pixel vectors, to calculate a maximal heating value of the corresponding location; and rejecting the specific pixel vector when the calculated maximal heating value does not surpass a predefined threshold.

9. The method of claim 8, wherein the tissue is a cutaneous, or sub-cutaneous tissue of a furry animal, and wherein obtaining the subset of candidate pixel vectors further comprises: analyzing a thermal feature value of the specific pixel vector, to determine consistency of the thermal feature value with heat dissipation by fur; and rejecting the specific pixel vector based on said consistency.

10. The method according to any one of claims 8-9, wherein the tissue is a cutaneous, or sub-cutaneous tissue of a furry animal, and wherein obtaining the subset of candidate pixel vectors further comprises: applying an ML-based, third classification model on a thermal feature value of the specific pixel vector, to predict a probability that the specific pixel vector depicts a fur of the animal; and rejecting the specific pixel vector based on said probability.

11. The method according to any one of claims 1-10, wherein determining a CD in the depicted area comprises: determining a window of pixel vectors surrounding a candidate pixel vector; calculating at least one statistical metric value, representing difference in temperature between (i) the candidate pixel vector and (ii) pixel vectors at the border of said window, along the measurement period; and identifying the candidate pixel vector as pertaining to a CD, when the calculated at least one statistical metric value surpasses a predetermined threshold, for at least a predetermined duration.

12. The method according to any one of claims 1-11, wherein calculating a thermal feature value of a specific pixel vector comprises: applying a polynomial fitting algorithm on the specific pixel vector, to fit values of member pixels into a polynomial function; and selecting the thermal feature value as a coefficient of a fourth-degree variable of said polynomial function.

13. The method according to any one of claims 1-12, wherein calculating a thermal feature value of a specific pixel vector comprises: calculating one or more Fourier series coefficients of the first ROI’s pixel vector; and selecting the thermal feature value from the one or more calculated Fourier series coefficients.

14. The method according to any one of claims 1-13, further comprisingEmploying a scanning light source, to obtain a scan data element representing a topography of the depicted area; and compensating at least one thermal feature value based on the topography, as represented by the scan data element.

15. The method according to any one of claims 3-14, wherein the spatial feature comprises one or more numerical values representing characteristics of a second, annular ROI, surrounding the first ROI.

16. The method according to any one of claims 2-15, further comprising: applying an ML-based ROI calculation module on at least one thermal image of the depicted area, to automatically define (i) the first ROI, suspected as anomalous, and (ii) a third ROI, beyond the first ROI, that is presumed healthy; extract one or more features representing the first ROI and third ROI, said features selected from spatial features, thermal features and morphological features; and categorizing the first ROI as anomalous based on said features of the first ROI and third ROI.

17. A system for categorizing anomaly of a tissue, the system comprising: a thermal imaging device, configured to obtain a sequence of thermal images depicting a preheated, or precooled area of the tissue, said sequence representing thermal changes in the depicted area over a predetermined measurement period; a non-transitory memory device, wherein modules of instruction code are stored, and at least one processor associated with the memory device, and configured to execute the modules of instruction code, whereupon execution of said modules of instruction code, the at least one processor is configured to: compute a plurality of pixel vectors in the sequence of images, wherein each pixel vector comprises member pixels that correspond to substantially the same location in the depicted area, along the measurement period; for each pixel vector, calculate one or more respective thermal feature values, characteristic of the thermal changes over time in the corresponding location; apply a filter on the plurality of pixel vectors, to obtain a subset of candidate pixel vectors; determine one or more cold dots (CDs) in the depicted area, based on the thermal feature values of candidate pixel vectors, wherein each CD comprises one or more member candidate pixel vectors; for each CD, apply a first classifier on the thermal feature values of the one or more member candidate pixel vectors, to predict a risk score representing a probability that said CD is anomalous; andcategorize anomaly of the depicted area based on the risk scores of the one or moreCDs.

18. The system of claim 17, further comprising a thermal element, associated with the at least one processor, wherein the at least one processor is further configured to control the thermal element so as to preheat, or precool the depicted area.

19. The system according to any one of claims 17-18, further comprising a scanning light source, associated with the at least one processor, wherein the at least one processor is further configured to: control the scanning light source to produce a scan the depicted area; obtain a scan data element, representing a topography of the depicted area; and compensate at least one thermal feature value, based on the topography, as represented by the scan data element.

20. The system according to any one of claims 17-19, wherein the at least one processor is further configured to categorize an anomaly of the depicted area by: receiving a definition of an ROI within the area, suspected to be anomalous; identifying one or more first CDs in the ROI; and categorizing the ROI as anomalous, based on the risk scores of the one or more first CDs.

21. The system according to any one of claims 17-20, wherein the at least one processor is further configured to: identify one or more second CDs in a peripheral region of the ROI; calculate one or more spatial feature values, characteristic of spatial distribution of the one or more first CDs and the one or more second CDs; and categorize the ROI as anomalous, further based on the spatial feature values.

22. The system according to any one of claims 17-21, further comprising a visible light imaging device, and wherein the at least one processor is further configured to:control the visible light imaging device to obtain a visible light image of the depicted area; and categorize the ROI by: analyzing the visible-light image to extract at least one morphological feature; and applying a pretrained, ML based, second classifier on (i) the risk scores, (ii) the one or more spatial feature values, and (iii) the at least one morphological feature, to predict an anomaly category of the ROI, wherein said anomaly category is selected from a type of a malignant tumor, a type of a benign tumor, an inflamed tissue and a healthy tissue.

Images