Machine learning systems for assessment, healing prediction, and treatment of wounds

Abstract

SOLUTION: Systems for assessing or predicting wound healing generate an image having pixels depicting a tissue region by receiving at least one signal from a light detection element indicating light of at least a first wavelength reflected from the tissue region including a wound, and a portion thereof with a mobile computer device, determine a reflectance intensity value of each pixel in the first wavelength, determine one or more quantitative features of a subset of the plurality of pixels based on the reflectance intensity value of each pixel in the subset, and generate at least one scalar value corresponding to a predicted healing amount of the wound or a portion thereof after the elapse of a prescribed period from generation of the image on the basis of the one or more quantitative features of the subset of the plurality of pixels by using one or more machine learning algorithms.EFFECT: A clinician can determine the prognosis of wound healing, select appropriate wound care therapy, or perform both.SELECTED DRAWING: Figure 30

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of U.S. Provisional Application No. 62 / 780,854, filed December 17, 2018, entitled "Predicting Healing of Diabetic Foot Ulcers at First Clinic Visit Using Artificial Intelligence," U.S. Provisional Application No. 62 / 780,121, filed December 14, 2018, entitled "Systems and Methods for High Precision Multi-Aperture Spectral Imaging," and U.S. Provisional Application No. 62 / 818,375, filed March 14, 2019, entitled "Systems and Methods for High Precision Multi-Aperture Spectral Imaging," each of which is expressly incorporated by reference in its entirety for all purposes.

[0002] STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT The inventions described in this disclosure were made in part with U.S. Government support awarded by the Biomedical Advanced Research and Development Authority (BARDA) in the Office of the Assistant Secretary for Preparedness and Response, U.S. Department of Health and Human Services under Contract No. HHSO100201300022C. The inventions described in this disclosure were also made in part with U.S. Government support awarded by the Defense Health Agency (DHA) under Contract No. W81XWH-17-C-0170 and / or Contract No. W81XWH-18-C-0114. The U.S. Government may own certain rights in this invention.

[0003] The systems and methods disclosed herein relate to medical imaging, and more specifically, to wound assessment, healing prediction and treatment using machine learning techniques. [Background technology]

[0004] Optical imaging is an emerging technology that has the potential to improve disease prevention, diagnosis and treatment in the clinic, hospital bed or operating room during critical care. Optical imaging techniques can non-invasively distinguish between tissues, or between tissues labeled with endogenous or exogenous contrast agents and living tissues, by measuring the photon absorption or scattering properties of each tissue at different wavelengths. These differences in photon absorption or scattering properties provide specific tissue contrasts, enabling the study of functional and molecular level activity that is indicative of health and disease.

[0005] The electromagnetic spectrum is a band of wavelengths or frequencies of electromagnetic radiation (e.g., light). From longest to shortest wavelengths, the electromagnetic spectrum includes radio waves, microwaves, infrared light (IR), visible light (i.e., light detectable by structures in the human eye), ultraviolet light (UV), x-rays, and gamma rays. Spectral imaging also refers to a branch of spectroscopy and photography that collects partial or complete spectra of spectral information at various locations in an image plane. Some types of spectral imaging systems can image one or more spectral bands. Multispectral imaging systems can image multiple spectral bands (usually a dozen or less spectral bands, each in a different spectral region) and collect measurements in the spectral bands for each pixel, and can refer to a bandwidth of tens of nm per spectral channel. Hyperspectral imaging systems, on the other hand, can measure many more spectral bands, e.g., as many as 200 or more spectral bands, and some types of hyperspectral imaging systems can continuously sample narrow bands (e.g., spectral bandwidth of a few nm or less) along a portion of the electromagnetic spectrum. Summary of the Invention [Means for solving the problem]

[0006] Aspects of the technology described herein relate to devices and methods that can be used to assess and / or classify tissue regions at or near a wound using non-contact, non-invasive, non-radioactive optical imaging. Such devices and methods may, for example, identify tissue regions corresponding to various tissue health classifications associated with a wound and / or measure healing predictive parameters of the wound or portions thereof, and output a visual display of the identified regions and / or parameters that can be used by a clinician to determine a prognosis of wound healing and / or select an appropriate wound care therapy. In some embodiments, devices and methods according to the technology of the present invention can provide such classification and / or prediction based on imaging at a single wavelength or multiple wavelengths. There has long been a need for such non-invasive imaging technology that can provide a physician with quantitatively predictive information about the healing of a wound or portions thereof.

[0007] In one embodiment, a system for assessing or predicting wound healing comprises: at least one light detecting element configured to collect light of at least a first wavelength reflected from a tissue region including a wound; and One or more processors in communication with the at least one light detecting element. The one or more processors: receiving a signal from the at least one light detecting element indicative of light at a first wavelength reflected from the tissue region; generating an image having a plurality of pixels representative of the tissue region based on the signals; measuring a reflected intensity value for each pixel at a first wavelength for at least a subset of the plurality of pixels based on the signal; determining one or more quantitative characteristics of the subset of the plurality of pixels based on the reflectance intensity values ​​of each pixel in the subset; The method is configured to use one or more machine learning algorithms to generate at least one scalar value corresponding to a healing prediction parameter or a healing evaluation parameter after a predetermined period of time based on the one or more quantitative features of the subset of the plurality of pixels.

[0008] In some embodiments, the wound is a diabetic foot ulcer. In some embodiments, the healing prediction parameter is a predicted amount of healing of the wound. In some embodiments, the healing prediction parameter is a predicted % area reduction of the wound. In some embodiments, the at least one scalar value comprises a plurality of scalar values, each scalar value corresponding to a likelihood of healing of an individual pixel of the subset or a likelihood of healing of a subgroup of pixels of the subset. In some embodiments, the one or more processors are further configured to output a visual representation of the plurality of scalar values ​​for display to a user. In some embodiments, the visual representation comprises the image in which each pixel of the subset is displayed with a particular visual representation selected based on the likelihood of healing of each pixel, with pixels with different likelihoods of healing being displayed with different visual representations. In some embodiments, the one or more machine learning algorithms comprise a SegNet pre-trained using an image database of wounds, burns or ulcers. In some embodiments, the image database of wounds comprises an image database of diabetic foot ulcers. In some embodiments, the image database of wounds comprises an image database of burns. In some embodiments, the predetermined period is 30 days. In some embodiments, the one or more processors are further configured to identify at least one patient health measure corresponding to a patient having the tissue region, and the at least one scalar value is generated based on the one or more quantitative features of the subset of the plurality of pixels and the at least one patient health measure. In some embodiments, the at least one patient health measure includes at least one variable selected from the group consisting of demographic variables, diabetic foot ulcer history variables, compliance variables, endocrine system variables, cardiovascular system variables, musculoskeletal system variables, nutritional status variables, infectious disease variables, renal variables, obstetric-gynecological variables, medication use variables, other disease variables, and clinical laboratory values.In some embodiments, the at least one patient health measure comprises one or more clinical features. In some embodiments, the one or more clinical features comprise at least one feature selected from the group consisting of a patient's age, a patient's degree of chronic kidney disease, a length of the wound on the day the image was created, and a width of the wound on the day the image was created. In some embodiments, the first wavelength is within the range of 420 nm ± 20 nm, 525 nm ± 35 nm, 581 nm ± 20 nm, 620 nm ± 20 nm, 660 nm ± 20 nm, 726 nm ± 41 nm, 820 nm ± 20 nm, or 855 nm ± 30 nm. In some embodiments, the first wavelength is within the range of 620 nm ± 20 nm, 660 nm ± 20 nm, or 420 nm ± 20 nm. In some embodiments, the one or more machine learning algorithms comprise a random forest ensemble. In some embodiments, the first wavelength is within the range of 726 nm±41 nm, 855 nm±30 nm, 525 nm±35 nm, 581 nm±20 nm, or 820 nm±20 nm. In some embodiments, the one or more machine learning algorithms include an ensemble of classifiers. In some embodiments, the system further includes an optical bandpass filter configured to pass at least the first wavelength of light. In some embodiments, the one or more processors are further configured to automatically segment a plurality of pixels in the image into wound pixels and non-wound pixels; and to select a subset of the plurality of pixels that includes the wound pixels. In some embodiments, the one or more processors are further configured to automatically segment the non-wound pixels into callus pixels and background pixels. In some embodiments, the one or more processors are further configured to automatically segment the non-wound pixels into callus pixels, normal skin pixels, and background pixels. In some embodiments, the one or more processors automatically segment the plurality of pixels using a segmentation algorithm that includes a convolutional neural network.In some embodiments, the segmentation algorithm is at least one of a U-Net including multiple convolution layers and a SegNet including multiple convolution layers. In some embodiments, the one or more quantitative features of the subset of the plurality of pixels include one or more aggregated quantitative features of the plurality of pixels. In some embodiments, the one or more aggregated quantitative features of the subset of the plurality of pixels are selected from the group consisting of a mean of the reflectance intensity values ​​of pixels of the subset, a standard deviation of the reflectance intensity values ​​of pixels of the subset, and a median of the reflectance intensity values ​​of pixels of the subset. In some embodiments, the one or more processors are further configured to create a plurality of transformed images by individually applying a plurality of filter kernels to the image by convolution; construct a three-dimensional matrix from the plurality of transformed images; and determine one or more quantitative features of the three-dimensional matrix, wherein the at least one scalar value is generated based on the one or more quantitative features of the subset of the plurality of pixels and based on the one or more quantitative features of the three-dimensional matrix. In some embodiments, the one or more quantitative features of the three-dimensional matrix are selected from the group consisting of a mean of the values ​​of the three-dimensional matrix, a standard deviation of the values ​​of the three-dimensional matrix, a median of the three-dimensional matrix, and a product of the mean and median of the three-dimensional matrix. In some embodiments, the at least one scalar value is generated based on the mean of the reflectance intensity values ​​of the pixels of the subset, the standard deviation of the reflectance intensity values ​​of the pixels of the subset, the median of the reflectance intensity values ​​of the pixels of the subset, the mean of the values ​​of the three-dimensional matrix, the standard deviation of the values ​​of the three-dimensional matrix, and the median of the three-dimensional matrix.In some embodiments, the at least one optical detecting element is further configured to collect light of at least a second wavelength reflected from the tissue region, and the one or more processors are further configured to receive a second signal from the at least one optical detecting element indicative of light of the second wavelength reflected from the tissue region; measure a reflected intensity value of each pixel at the second wavelength for at least a subset of the plurality of pixels based on the second signal; determine one or more additional quantitative features of the subset of the plurality of pixels based on the reflected intensity values ​​of each pixel at the second wavelength; and the at least one scalar value is generated based at least in part on the one or more additional quantitative features of the subset of the plurality of pixels.

[0009] In a second aspect, a system for wound assessment comprises: at least one light detecting element configured to collect light of at least a first wavelength reflected from a tissue region including a wound; and One or more processors in communication with the at least one light detecting element. The one or more processors: receiving a signal from the at least one light detecting element indicative of light at a first wavelength reflected from the tissue region; generating an image having a plurality of pixels representative of the tissue region based on the signals; measuring a reflected intensity value for each of the plurality of pixels at a first wavelength based on the signal; The method is configured to use a machine learning algorithm to automatically segment each of the plurality of pixels into at least a first subset of the plurality of pixels comprising wound pixels and a second subset of the plurality of pixels comprising non-wound pixels based on the reflectance intensity value of each of the plurality of pixels.

[0010] In some embodiments, the one or more processors are further configured to automatically segment a second subset of the plurality of pixels into at least two types of non-wound pixels, the at least two types of pixels being selected from the group consisting of callus pixels, normal skin pixels, and background pixels. In some embodiments, the machine learning algorithm comprises a convolutional neural network. In some embodiments, the machine learning algorithm is at least one of a U-Net comprising a plurality of convolutional layers and a SegNet comprising a plurality of convolutional layers. In some embodiments, the machine learning algorithm is trained on a dataset comprising a plurality of segmented images of wounds, ulcers, or burns. In some embodiments, the wound is a diabetic foot ulcer. In some embodiments, the one or more processors are further configured to output a visual representation of the segmented plurality of pixels for display to a user. In some embodiments, the visual representation comprises the image, with each pixel represented by a particular visual representation selected based on the segmented pixels, and wound pixels and non-wound pixels represented by different visual representations.

[0011] In another aspect, a method for predicting wound healing using a system for assessing or predicting wound healing comprises: illuminating a tissue region with light at least at a first wavelength and reflecting at least a portion of the light at the tissue region onto at least one light detecting element; generating at least one scalar value using the wound healing assessment or prediction system; and Determining a healing prediction parameter after a predetermined period of time Includes.

[0012] In some embodiments, illuminating the tissue area comprises activating one or more light emitting elements configured to emit light at least at the first wavelength. In some embodiments, illuminating the tissue area comprises exposing the tissue area to ambient light. In some embodiments, determining the healing prediction parameter comprises determining an expected % wound area reduction after a predetermined period of time. In some embodiments, the method further comprises: measuring one or more dimensions of the wound a predetermined period of time after determining the predicted amount of healing of the wound; measuring the actual amount of healing of the wound after the period of time; and updating at least one of one or more machine learning algorithms by providing at least the images and the actual amount of healing of the wound as training data. Further includes: In some embodiments, the method further comprises selecting a standard wound care regimen or an advanced wound care regimen based at least in part on the healing prediction parameter. In some embodiments, the step of selecting a standard or advanced wound care therapy comprises prescribing or applying one or more standard therapies selected from the group consisting of optimizing nutritional status, debridement to remove necrotic tissue by any means, maintaining a clean and moist granulation tissue bed with an appropriate moist dressing, therapy required to resolve any infection that may exist, addressing the lack of intravascular perfusion to the limb having the diabetic foot ulcer, reducing pressure load caused by the diabetic foot ulcer, and appropriate glycemic control if the healing prediction parameter indicates that the wound (preferably the diabetic foot ulcer) will not heal or close by more than 50% after 30 days, and prescribing or applying one or more advanced therapies selected from the group consisting of hyperbaric oxygen therapy, negative pressure wound therapy, bioengineered skin substitutes, synthetic growth factors, extracellular matrix proteins, matrix metalloproteinase regulators, and electrical stimulation therapy if the healing prediction parameter indicates that the wound (preferably the diabetic foot ulcer) will not heal or close by more than 50% after 30 days. [Brief description of the drawings]

[0013] [Figure 1A] 1 shows an example of light entering a filter at various chief ray angles of incidence.

[0014] [Figure 1B] 1B is a graph showing an example of the transmission efficiency obtained from the filter of FIG. 1A through which light is transmitted at various chief ray angles of incidence.

[0015] [Figure 2A] 1 shows an example of a multispectral image data cube.

[0016] [Figure 2B]An example is shown of how a particular multispectral imaging technique generates the data cube of FIG. 2A.

[0017] [Figure 2C] 2B shows an example of a snapshot imaging system capable of generating the data cube of FIG. 2A.

[0018] [Figure 3A] FIG. 1 shows a cross-sectional schematic diagram illustrating the optical design of an example of a multi-aperture imaging system with curved multi-bandpass filters according to the present disclosure.

[0019] [Figure 3B-3D] 3B shows an example of an optical design of optical components that constitute one optical path of the multi-aperture imaging system shown in FIG. 3A.

[0020] [Figure 4A-4E] 3A and 3B illustrate an embodiment of a multispectral, multiaperture imaging system having the optical design described in connection with FIG. 3A and FIG. 3B.

[0021] [Diagram 5] FIG. 4 illustrates another embodiment of a multispectral, multiaperture imaging system with the optical design described in connection with FIGS. 3A and 3B.

[0022] [Figure 6A-6C] FIG. 4 illustrates another embodiment of a multispectral, multiaperture imaging system with the optical design described in connection with FIGS. 3A and 3B.

[0023] [Figure 7A-7B] FIG. 4 illustrates another embodiment of a multispectral, multiaperture imaging system with the optical design described in connection with FIGS. 3A and 3B.

[0024] [Figure 8A-8B]Another embodiment of a multi-spectral, multi-aperture imaging system with the optical design described in connection with FIGS. 3A and 3B is shown.

[0025] [Figure 9A-9C] Another embodiment of a multi-spectral, multi-aperture imaging system with the optical design described in connection with FIGS. 3A and 3B is shown.

[0026] [Figure 10A-10B] Another embodiment of a multi-spectral, multi-aperture imaging system with the optical design described in connection with FIGS. 3A and 3B is shown.

[0027] [Figure 11A-11B] An example of a set of frequency bands that can pass through the filter of the multi-spectral, multi-aperture imaging system shown in FIGS. 3A - 10B is shown.

[0028] [Figure 12] A schematic block diagram of an imaging system that can be used in the multi-spectral, multi-aperture imaging system shown in FIGS. 3A - 10B is shown.

[0029] [Figure 13] A flowchart showing an example of a process for capturing image data using the multi-spectral, multi-aperture imaging system shown in FIGS. 3A - 10B is shown.

[0030] [Figure 14] A schematic block diagram of a workflow for processing image data, such as image data captured using the multi-spectral, multi-aperture imaging system shown in FIGS. 3A - 10B and / or using the process shown in FIG. 13, is shown.

[0031] [Figure 15]For example, parallax and parallax correction are illustrated for processing image data, such as image data captured using the multispectral, multi-aperture imaging system shown in Figures 3A-10B and / or using the process shown in Figure 13.

[0032] [Figure 16] FIG. 13 illustrates a workflow for pixel-by-pixel classification of multispectral image data, such as image data captured using the multispectral, multi-aperture imaging system shown in FIGS. 3A-10B and / or using the process shown in FIG. 13 and then processed according to FIGS. 14 and 15.

[0033] [Figure 17] A simplified block diagram of an example of a distributed computing system including the multispectral, multi-aperture imaging system shown in Figures 3A to 10B is shown.

[0034] [Figures 18A-18C] 1 illustrates an embodiment of a portable multispectral, multiaperture imaging system.

[0035] [Figure 19A-19B] 1 illustrates an embodiment of a portable multispectral, multiaperture imaging system.

[0036] [Figure 20A-20B] An example of a multispectral, multi-aperture imaging system for a compact USB 3.0 camera housed in a standard camera housing is shown.

[0037] [Figure 21] An example of a multispectral, multi-aperture imaging system that includes an additional light source to improve image alignment is shown.

[0038] [Figure 22]An example of the time course of the progression of healing of a diabetic foot ulcer (DFU) and corresponding measurements of DFU area, volume and debridement are shown.

[0039] [Figure 23] An example of the time course of healing progression of a non-healing diabetic foot ulcer (DFU) is shown with corresponding measurements of DFU area, volume and debridement.

[0040] [Figure 24] FIG. 1 illustrates a schematic diagram of an example machine learning system for generating a healing prediction based on one or more images taken of a DFU.

[0041] [Diagram 25] FIG. 1 illustrates a schematic diagram of an example machine learning system for generating a healing prediction based on one or more images taken of a DFU and one or more patient health measurements.

[0042] [Figure 26] 1 illustrates an example set of wavelength bands used in spectral and / or multispectral imaging for image segmentation and / or generation of predictive parameters in accordance with the techniques of the present invention.

[0043] [Figure 27] 1 is a histogram showing the effect of including clinical variables in an exemplary wound assessment method according to the present technology.

[0044] [Figure 28] 1 shows a schematic diagram of an example of an autoencoder in accordance with a machine learning system and method according to the present technique;

[0045] [Figure 29] 1 shows a schematic diagram of an example of a supervised machine learning algorithm in accordance with a machine learning system and method according to the present technique;

[0046] [Diagram 30] 1 shows a schematic diagram of an example of an end-to-end machine learning algorithm in accordance with the machine learning system and method of the present technology;

[0047] [Diagram 31] 1 is a bar graph illustrating the accuracy exhibited by several example machine learning algorithms in accordance with the present technique.

[0048] [Diagram 32] 1 is a bar graph illustrating the accuracy exhibited by several example machine learning algorithms in accordance with the present technique.

[0049] [Diagram 33] FIG. 1 shows a schematic diagram of an example process for predicting healing and generating a visual display of a conditional probability mapping according to a machine learning system and method according to the present technology.

[0050] [Diagram 34] 1 shows a schematic diagram of an example of a conditional probability mapping algorithm that includes one or more feature-wise linear transformation layers (FiLM).

[0051] [Diagram 35] 1 illustrates the accuracy exhibited by several image segmentation approaches for the creation of conditional probability of cure maps in accordance with the present techniques.

[0052] [Diagram 36] 1 illustrates an example set of convolution filter kernels used in an example single wavelength analysis method for predicting healing in accordance with the machine learning system and method of the present technology.

[0053] [Figure 37]1 shows an example of a truth mask created based on a DFU image for image segmentation according to the machine learning system and method of the present technique.

[0054] [Figure 38] 1 illustrates the accuracy demonstrated by an exemplary segmentation algorithm for wound images in accordance with the machine learning system and method of the present technology. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0055] Of the 26 million Americans living with diabetes, approximately 15-25% will develop a diabetic foot ulcer (DFU). DFUs limit the patient's mobility and reduce their quality of life. Up to 40% of patients who develop a DFU also develop a wound infection, which increases the risk of amputation and death. DFUs alone are associated with a 5% first-year mortality rate and a 42% 5-year mortality rate. This is reflected in the high annual risk of major amputation (4.7%) and minor amputation (39.8%). Furthermore, the annual cost of treating a single DFU ranges from approximately $22,000 to $44,000, and DFUs are associated with a total economic burden of $9-13 billion per year to the U.S. healthcare system.

[0056] DFUs with a 30-day percent area reduction (PAR) of >50% are generally considered to be healed by 12 weeks with standard care therapy. However, using such measurements requires an initial 4-week wound care period to determine whether more effective treatments (e.g., advanced care therapy) should be used. When a non-emergency wound such as a DFU first occurs, the usual clinical response to wound care is to place the patient on standard wound care therapy (e.g., correction of vascular defects, optimization of nutritional status, glycemic control, debridement, dressings and / or reduction of pressure load) for approximately 30 days from wound initiation and initial evaluation. At approximately day 30, the wound is assessed for healing (e.g., percent area reduction >50%). If the wound is not healing adequately, treatment is supported with one or more advanced wound management therapies. These advanced wound management therapies include growth factors, bioengineered tissues, hyperbaric oxygen, negative pressure, amputation, recombinant human platelet-derived growth factor (e.g., Regranex). TM gel), bioengineered human dermis substitutes (e.g. Dermagraft TM ), and / or bilayered living skin substitutes (e.g., Apligraf TM However, approximately 60% of DFUs fail to heal adequately after 30 days of standard wound care therapy. Furthermore, approximately 40% of fast-healing DFUs do not heal by 12 weeks, and the median healing time for DFUs is estimated to be 147 days for toe ulcers, 188 days for midfoot ulcers, and 237 days for heel ulcers.

[0057] If a DFU does not achieve the desired healing after 30 days of conventional or standard wound care therapy, it may be beneficial to start an advanced wound care therapy as soon as possible (e.g., within 30 days of starting wound therapy). However, using conventional assessment methods, physicians are usually unable to accurately identify DFUs that do not respond to 30 days of standard wound care therapy. Various strategies aimed at improving the treatment of DFUs are available, but are prescribed only after the standard wound care therapy has been implemented. Physiological measurement devices such as transcutaneous oxygen tension measurement, laser Doppler imaging, and indocyanine green video angiography have also been used in attempts to diagnose the healing potential of DFUs. However, such devices are not suitable for widespread use in the evaluation of DFUs and other wounds due to their inaccurate results, lack of useful data, poor sensitivity, and prohibitive cost. Clearly, earlier and more accurate means of predicting the healing of DFUs and other wounds are important to quickly determine the best treatment and reduce the time required for wound closure.

[0058] As outlined herein, the technology of the present invention provides a non-invasive, point-of-care, real-time imaging device capable of diagnosing the healing potential of DFUs, burns and other wounds. In various embodiments, the systems and methods of the technology of the present invention allow clinicians to determine the healing potential of a wound when it is presented or first assessed, or shortly thereafter. In some embodiments, the technology of the present invention can determine the healing potential of individual sites of a wound (such as a DFU or burn). Based on this predicted healing potential, a decision can be made on or about day 0 of treatment whether to perform a standard or advanced wound care therapy, without delaying for more than four weeks from the first presentation of the wound. Thus, the technology of the present invention may reduce the time to healing and reduce the chance of amputation.

[0059] An example of a spectral imaging system and a multispectral imaging system Various spectral and multispectral imaging systems that may be used in accordance with the methods of assessing, predicting, and treating DFUs and other wounds disclosed herein are described below. In some embodiments, images for wound assessment may be taken with a spectral imaging system configured to image light in a single wavelength band. In other embodiments, images may be taken with a spectral imaging system configured to capture two or more wavelength bands. In one particular example, images may be taken with a monochromatic, RGB, and / or infrared imaging device, such as those found in commercially available mobile devices. Yet another embodiment relates to spectral imaging using a multi-aperture system with curved, multi-bandpass filters positioned above each aperture. However, it will be understood that the methods of assessing, predicting, and treating wounds according to the present technology are not limited to the specific image acquisition devices disclosed herein, and may similarly be implemented using any imaging device capable of acquiring image data in one or more known wavelength bands.

[0060] The present disclosure also relates to techniques for implementing spectral unmixing and image alignment to generate a spectral data cube using image information received from such imaging systems. The techniques of the present disclosure address various challenges commonly found in spectral imaging in obtaining image data that show accurate information about the wavelength bands reflected from the object being imaged, as described below. In some embodiments, the systems and methods described herein can obtain images of a large area of ​​tissue (e.g., 5.9×7.9 inches) in a short time (e.g., within 6 seconds), and can achieve this without the need for injection of imaging contrast agents. In some aspects, for example, the multispectral imaging systems described herein are configured to obtain images of a large area of ​​tissue, e.g., 5.9×7.9 inches, within 6 seconds, and provide tissue analysis information, such as identification of multiple burn conditions, multiple wound conditions, multiple ulcer conditions or healing potential; or clinical characteristics including cancerous or non-cancerous conditions of the imaged tissue, wound depth, wound volume, debridement margins, and the presence of diabetic ulcers, non-diabetic ulcers, or chronic ulcers, in the absence of imaging contrast agents. Similarly, in some of the methods described herein, the multispectral imaging system of the present invention can acquire images from a large area of ​​tissue, e.g., 5.9 x 7.9 inches, in less than 6 seconds and output tissue analysis information such as identification of multiple burn conditions, multiple wound conditions, or healing potential; or clinical characteristics including cancerous or non-cancerous condition of the imaged tissue, wound depth, wound volume, debridement margins, presence of diabetic ulcers, non-diabetic ulcers, or chronic ulcers, in the absence of imaging contrast agents.

[0061] One challenge with existing solutions is that color distortion or parallax in the captured image can impair the quality of the image data. This can be especially problematic in applications that rely on the use of optical filters to accurately detect and analyze specific wavelengths of light. More specifically, color shading is a change in the wavelength of light as it passes through the image sensor area, as the angle of incidence of light on the color filter increases, the shorter the wavelength of light that passes through the color filter. Typically, this effect is observed in interference filters, which are fabricated by stacking multiple thin layers of different refractive indexes on a transparent substrate. Thus, longer wavelengths (such as red light) are more likely to be blocked at the edge of the image sensor due to the larger angle of incidence of the light rays, and incident light of the same wavelength is detected as a spatially non-uniform color as it passes through the image sensor. Without image correction, color shading appears as a change in color near the edge of the captured image.

[0062] The disclosed technology has many advantages over other multispectral imaging systems on the market because it is not limited in lens and / or image sensor configuration and their field of view or aperture size. As one skilled in the art would appreciate, changes to the lenses, image sensors, aperture size, or other components of the imaging systems disclosed herein may include other adjustments to the imaging system. The disclosed technology also provides an improvement over other multispectral imaging systems in that components that perform wavelength splitting or utilize the entire imaging system (e.g., optical filters, etc.) can be separated from components that convert light energy to digital output (e.g., image sensors, etc.). This can reduce cost, simplicity, and / or development time to reconfigure the imaging system for different multispectral wavelengths. The disclosed technology is also believed to be more robust than other multispectral imaging systems in that it can achieve imaging characteristics comparable to other multispectral imaging systems on the market in a smaller and lighter form factor. The disclosed techniques also have advantages over other multispectral imaging systems in that they can acquire multispectral images at snapshot, video or high-speed video rates. The disclosed techniques can also multiplex several spectral bands into each aperture, reducing the number of apertures required to acquire a certain number of spectral bands in an imaging data set, and can provide a more robust implementation of multispectral imaging systems based on multi-aperture technology due to cost savings from the reduced number of apertures and improved light collection (e.g., commercially available sensor arrays can use larger apertures that fit into fixed sizes and dimensions). Moreover, the disclosed techniques can provide all of these advantages without compromising on resolution or image quality.

[0063] FIG. 1A illustrates an example of a filter 108 positioned along a light path toward an image sensor 110, and light incident on the filter 108 at various ray incidence angles. Rays 102A, 104A, and 106A are shown as lines that pass through the filter 108 and are then refracted by a lens 112 onto the sensor 110. The lens 112 may be replaced with other image forming optics, such as, but not limited to, a mirror and / or an aperture. In FIG. 1A, the light in each ray is considered broadband, e.g., light having a spectral composition in a wide wavelength range, and only light of a particular wavelength is selectively transmitted by the filter 108. The three rays 102A, 104A, and 106A are incident on the filter 108 at different angles of incidence. For purposes of illustration, light ray 102A is shown as being substantially normal incidence on filter 108, light ray 104A has a larger angle of incidence than light ray 102A, and light ray 106A has a larger angle of incidence than light ray 104A. Each light ray 102B, 104B, 106B that passes through the filter exhibits a unique spectrum due to the angle-of-incidence-dependent transmission characteristics of filter 108, which are detected by sensor 110. This angle-of-incidence-dependent effect causes a shift in the bandpass characteristics of filter 108, with shorter wavelengths as the angle of incidence increases. Furthermore, the angle-of-incidence dependence may reduce the transmission efficiency of filter 108 and may change the spectral shape of the bandpass characteristics of filter 108. This combined effect is referred to as angle-of-incidence-dependent spectral transmission. FIG. 1B shows the spectrum of each of the rays of FIG. 1A detected by a spectrometer assumed to be at the sensor 110, illustrating the shift in the spectrum of the bandpass characteristic of the filter 108 as the angle of incidence increases. Curves 102C, 104C, and 106C show the center wavelength of the bandpass characteristic becoming shorter, and therefore the wavelength of the light emitted from the optical system in this example becoming shorter. As shown in this figure, the spectral shape of the bandpass characteristic and the transmittance peak also change with the angle of incidence. In certain consumer applications, image processing can be performed to remove the visible effects of such angle-dependent spectral transmission.However, such post-processing techniques cannot recover the exact wavelengths of light that actually entered filter 108. Thus, the resulting image data may not be usable for certain high precision applications.

[0064] Another challenge faced by certain existing spectral imaging systems is the time required to capture a complete spectral image data set, which is discussed in relation to Figures 2A and 2B. A spectral image sensor acquires the spectral irradiance I(x,y,λ) of a particular scene and collects a three-dimensional (3D) data set, commonly referred to as a data cube. Figure 2A shows an example of a spectral image data cube 120. As shown in this figure, the data cube 120 represents three-dimensional image data, where two spatial dimensions (x and y) correspond to the two-dimensional (2D) surface of the image sensor, and the spectral dimension (λ) corresponds to a particular wavelength range. The size of the data cube 120 is N x N y N λ and N x and N y is the number of sample points in each spatial dimension (x, y), and N λ is the number of sample points along the spectral axis λ. Because the data cube has a higher dimensionality than currently available 2D detector arrays (e.g., image sensors), a typical spectral imaging system simultaneously measures all sample points of the data cube 120 by taking time-sequential 2D slices (i.e., planes) of the data cube 120 (referred to herein as a "scanning" imaging system) or by dividing the data cube 120 during computation to obtain multiple two-dimensional data that can be reintegrated into the data cube 120 (referred to herein as a "snapshot" imaging system).

[0065] FIG. 2B shows an example of how a particular scanning spectral imaging technique generates data cube 120. Specifically, FIG. 2B shows portions 132, 134, and 136 of data cube 120 that can be collected in one detector integration time. A point-scanning spectrometer can, for example, image portion 132 that extends across the spectral plane λ at one spatial location (x,y). A point-scanning spectrometer can build data cube 120 by performing multiple integrations for each (x,y) location in the spatial dimensions. A filter wheel imaging system can, for example, image portion 134 that extends across spatial dimensions x and y but is located in only a single spectral plane λ. A wavelength-scanning imaging system, such as a filter wheel imaging system, can build data cube 120 by performing multiple integrations over the spectral plane λ. A line-scanning spectrometer can, for example, image portion 136 that extends across the entire spectral dimension λ and across one of the spatial dimensions (x or y) but is located at only one point in the other spatial dimension (y or x). A line-scan spectrometer can construct a data cube 120 by performing multiple integrations over the position of one of the spatial dimensions (y or x).

[0066] In applications where both the object and the imaging system are stationary (or relatively stationary during the exposure time), the scanning imaging systems described above have the advantage of providing a high resolution data cube 120. Line-scan and wavelength-scanning imaging systems may benefit from each spectral or spatial image being taken using the entire area of ​​the image sensor. However, motion of the imaging system and / or object between first and second exposures may result in artifacts in the resulting image data. For example, the same (x,y) location in the data cube 120 may actually represent a different physical location on the imaged object in the spectral dimension λ. Such artifacts may result in errors in subsequent analysis and / or the need for alignment (e.g., alignment of the spectral dimension λ so that a particular (x,y) location corresponds to the same physical location on the object).

[0067] In contrast, the snapshot imaging system 140 can capture the entire datacube 120 in one integration time or exposure, thus avoiding the effects of motion on image quality. FIG. 2C shows an example of an image sensor 142 and an optical filter array 144 (such as a color filter array (CFA)) that can be used to create a snapshot imaging system. In this example, the color filter array (CFA) 144 is a repeating pattern of color filter units 146 arranged on the surface of the image sensor 142. This method of obtaining spectral information can also be called a multispectral filter array (MSFA) or a spectrally resolved detector array (SRDA). In the example shown, the color filter unit 146 includes various color filters arranged in a 5×5 array, thereby generating 25 spectral channels in the resulting image data. The CFA can use various color filters to split the incoming light into bands for each color filter and direct the split light to the photoreceptors of each color on the image sensor. In this way, for a given color 148, only one of the 25 photoreceptors actually detects a signal indicative of light of that wavelength. Thus, with such a snapshot imaging system 140, 25 color channels may be generated in one exposure, but the amount of measurement data for each color channel is less than the total output of the image sensor 142. In some embodiments, the CFA may include a filter array (MSFA) and / or a spectrally resolved detector array (SRDA), and / or may include conventional Bayer filters, CMYK filters, or other absorptive or interferometric filters. One type of interferometric filter is a thin film filter array arranged in a grid, with each grid element corresponding to one or more sensor elements. Another type of interferometric filter is a Fabry-Perot filter.Nano-etched Fabry-Perot interference filters typically exhibit bandpass characteristics with a full width at half maximum (FWHM) of about 20-50 nm, but are advantageous in some embodiments because they exhibit a small roll-off at the transition from the center wavelength of the filter passband to the stopband. In addition, such filters exhibit a low optical density (OD) in the stopband, which can improve sensitivity to light outside the passband. This combined effect allows these particular filters to be sensitive to spectral regions that would be blocked by the high roll-off of a similar FWHM high optical density interference filter constructed from multiple thin film layers deposited by coating deposition techniques such as evaporation or ion beam sputtering. In embodiments using dye-based CMYK filters or RGB-based (Bayer) filters, it is preferable to have a small spectral roll-off and a large FWHM in the passband of each individual filter, which allows for unique spectral transmission for each wavelength across the entire spectrum being observed.

[0068] Thus, the data cube 120 obtained by the snapshot imaging system has one of two properties that can be problematic for high-precision imaging applications: First, the data cube 120 obtained by the snapshot imaging system has a size N larger than the (x,y) dimension of the detector array. x and N y The second problem is that the data cube 120 obtained by a snapshot imaging system has a resolution of N due to the small size ... x and N y The dimensions of the (x,y) may be the same as the (x,y) dimensions of the detector array. However, if interpolation was used in generating the data cube, this means that any particular value in the data cube is not an actual measurement of the wavelength of light incident on the sensor, but an estimate of that measurement based on surrounding values.

[0069] Another existing component used in multispectral imaging with a single exposure is the multispectral imaging beamsplitter. In such imaging systems, a beamsplitter cube splits the incoming light into different color bands, each of which is observed by separate image sensors. Although the design of the beamsplitter can be modified to tune the spectral bands being measured, it is not easy to split the incoming light into more than four beams without compromising the performance of the system. Thus, four spectral channels appears to be the practical limit of this approach. A closely related method splits the light using thin film filters instead of bulky beamsplitter cubes / prisms, but the cumulative transmission attenuation of multiple successive filters and spatial limitations limit this approach to about six spectral channels.

[0070] In some embodiments, the aforementioned challenges, among others, can be addressed by the multi-aperture spectral imaging system and associated image data processing techniques disclosed herein, which includes a multi-bandpass filter (preferably a curved multi-bandpass filter) that selectively transmits light incident through each aperture. This particular configuration can achieve all of the design goals of fast imaging speed, high resolution images, and high fidelity of detection wavelength. Thus, the optical design and associated image data processing techniques disclosed herein can be used in a portable spectral imaging system and / or can be used to image moving subjects and obtain data cubes suitable for high precision applications (e.g., clinical tissue analysis, biometric authentication, and episodic clinical signs). Such high precision applications can include diagnosing the stage (0-3) of malignant melanoma before metastasis, classifying the severity of wounds or burns in skin tissue, or tissue diagnosis of the severity of diabetic foot ulcers. Thus, the snapshot spectral acquisition and small form factor described in some embodiments allows the present invention to be used in clinical settings dealing with transient events, including diagnosis of various types of retinopathies (e.g., non-proliferative diabetic retinopathy, proliferative diabetic retinopathy, and age-related macular degeneration), as well as imaging of pediatric patients with high motion. Thus, one skilled in the art will appreciate that the use of the multi-aperture system with flat or curved multi-bandpass filters disclosed herein represents a significant technological advance over conventional spectral imaging implementations. More specifically, the multi-aperture system of the present invention may enable the collection of 3D spatial images of or relating to the curvature, depth, volume, and / or area of ​​an object based on the calculated perspective difference between each aperture. However, the multi-aperture method disclosed herein is not limited to any particular filter and may include flat and / or low-profile filters based on interference or absorption filtering.The invention disclosed herein can be modified to include flat filters in the image space of the imaging system when the angle of incidence to the appropriate lens or aperture is small or within the acceptable range. Filters may be placed at the entrance and / or exit pupil of the imaging lens or aperture stop, and such configurations would be readily employable by one skilled in the art of optical engineering.

[0071] Various aspects of the disclosure are described below with reference to specific examples and embodiments. These examples and embodiments are for illustrative purposes and are not intended to limit the disclosure. While the examples and embodiments described herein focus on specific calculations and algorithms for illustrative purposes, those skilled in the art will appreciate that these examples are for illustrative purposes only and are not intended to limit the invention. For example, while some examples are provided with respect to multispectral imaging, the multi-aperture imaging systems and associated filters disclosed herein can be configured to perform hyperspectral imaging in other implementations. Furthermore, while certain examples are presented as achieving advantages in portable applications and / or applications with moving target objects, it will be appreciated that the imaging system designs and associated processing techniques disclosed herein can provide highly accurate data cubes suitable for fixed imaging systems and / or analysis of relatively stationary targets.

[0072] Overview of Electromagnetic Spectral Range and Image Sensors Reference is made herein to particular colors or portions of the electromagnetic spectrum, and these wavelengths are hereinafter described as defined by ISO 21348, "Definitions of Irradiance Spectral Types." As will be described in more detail below, in certain imaging applications, wavelength ranges of particular colors may be passed collectively through particular filters.

[0073] Electromagnetic radiation in the wavelength range of 760 nm to 380 nm or about 760 nm to 380 nm is usually considered to be the "visible" spectrum, i.e., the portion of the spectrum that is perceptible by the color receptors of the human eye. In the visible spectrum, red light is usually considered to have a wavelength of 700 nm or about 700 nm, or to be in the range of 760 nm to 610 nm or about 760 nm to 610 nm. Orange light is usually considered to have a wavelength of 600 nm or about 600 nm, or to be in the range of 610 nm to 591 nm or about 610 nm to 591 nm. Yellow light is usually considered to have a wavelength of 580 nm or about 580 nm, or to be in the range of 591 nm to 570 nm or about 591 nm to 570 nm. Green light is usually considered to have a wavelength of 550 nm or about 550 nm, or in the range of 570 nm to 500 nm or about 570 nm to about 500 nm. Blue light is usually considered to have a wavelength of 475 nm or about 475 nm, or in the range of 500 nm to 450 nm or about 500 nm to about 450 nm. Purple (magenta) light is usually considered to have a wavelength of 400 nm or about 400 nm, or in the range of 450 nm to 360 nm or about 450 nm to about 360 nm.

[0074] Looking beyond the visible spectrum, infrared (IR) refers to electromagnetic radiation that has longer wavelengths than visible light and is generally not visible to the human eye. IR has wavelengths that extend from about 760 nm or about 1 mm to the nominal lower limit of the red visible spectrum, which is at 760 nm. Within this range, near infrared (NIR) refers to the portion of the spectrum adjacent to the red range, and is the wavelength range of about 760 nm to about 1400 nm or 760 nm to 1400 nm.

[0075] Ultraviolet (UV) light refers to a portion of electromagnetic radiation with wavelengths shorter than visible light and is generally invisible to the human eye. UV has wavelengths ranging from about 40 nm or the nominal lower limit of the visible spectrum at violet at 40 nm to about 400 nm. Within this range, near ultraviolet (NUV) refers to a portion of the spectrum adjacent to the violet range and has a wavelength range of about 400 nm to about 300 nm or 400 nm to 300 nm, mid ultraviolet (MUV) has a wavelength range of about 300 nm to about 200 nm or 300 nm to 200 nm, and far ultraviolet (FUV) has a wavelength range of about 200 nm to about 122 nm or 200 nm to 122 nm.

[0076] The image sensors described herein can be configured to detect electromagnetic radiation in any of the above ranges, depending on the particular wavelength range that is appropriate for a particular application. Typical silicon charge-coupled device (CCD) or complementary metal-oxide semiconductor (CMOS) sensors have spectral sensitivity that spans the entire visible spectrum, but also most of the near infrared (IR) spectrum and parts of the UV spectrum. In some implementations, back-illuminated or front-illuminated CCD or CMOS arrays can additionally or alternatively be used. For applications requiring high signal-to-noise ratio and scientific-grade measurements, some implementations can additionally or alternatively use scientific complementary metal-oxide semiconductor (sCMOS) cameras or electronically multiplied CCD cameras (EMCCD). In other implementations, sensors known to operate in a particular color range (e.g., short-wave infrared (SWIR), mid-wave infrared (MWIR), or long-wave infrared (LWIR)) and corresponding optical filter arrays can additionally or alternatively be used, depending on the intended application. These sensors and optical filter arrays may additionally or alternatively comprise cameras based on detector materials such as indium gallium arsenide (InGaAs) or indium antimony (InSb), or cameras based on microbolometer arrays.

[0077] The image sensors used in the multispectral imaging techniques disclosed herein may be used in conjunction with an optical filter array, such as a color filter array (CFA). Some types of CFAs can split incident light into red (R), green (G) and blue (B) visible regions and direct the split visible light to dedicated red, green or blue photodiode photoreceptors on the image sensor. A common example of a CFA is the Bayer pattern, which is a color filter array that arranges RGB color filters in a specific pattern on a rectangular grid of photodetectors. The Bayer pattern is composed of 50% green, 25% red and 25% blue, with rows of repeated red and green color filters alternating with rows of repeated blue and green color filters. Some types of CFAs (e.g., CFAs for RGB-NIR sensors) can also split NIR light and direct the split NIR light to dedicated photodiode photoreceptors on the image sensor.

[0078] Thus, the wavelength range of the filter components of the CFA may determine the wavelength range that each image channel exhibits in the captured image. Thus, in various embodiments, the red channel of the image may correspond to the red wavelength range of the color filter, may include some yellow and orange light, and may range from about 570 nm to about 760 nm or from 570 nm to 760 nm. In various embodiments, the green channel of the image may correspond to the green wavelength range of the color filter, may include some yellow light, and may range from about 570 nm to about 480 nm or from 570 nm to 480 nm. In various embodiments, the blue channel of the image may correspond to the blue wavelength range of the color filter, may include some violet light, and may range from about 490 nm to about 400 nm or from 490 nm to 400 nm. Those skilled in the art will appreciate that the exact start and end wavelengths (or portions of the electromagnetic spectrum) that define the colors (e.g., red, green, and blue) of the CFA may vary depending on the implementation of the CFA.

[0079] Additionally, typical visible CFAs transmit light outside the visible spectrum. Thus, many image sensors limit their sensitivity to infrared by placing a thin-film infrared-reflecting filter in front of the sensor that blocks infrared wavelengths but passes visible light. However, in some of the imaging systems disclosed herein, the thin-film infrared-reflecting filter may be omitted to allow infrared light to pass. Thus, the red, green and / or blue channels may be used to collect infrared wavelength bands. In some implementations, the blue channel may be used to collect a specific NUV wavelength band. Since the red, green and blue channels have different spectral responses associated with each channel having a unique transmission efficiency at each wavelength of the stacked spectral image, a known transmission profile may be used to provide a unique weighting response for each spectral band prior to unmixing. For example, the weighting response may include the known transmission response of the red, blue and green channels in the infrared and ultraviolet wavelength regions, allowing each channel to be used to collect bands from these wavelength regions.

[0080] As will be described in more detail below, additional color filters can be placed in front of the CFA along the light path to the image sensor to selectively extract specific bands of light incident on the image sensor. The color filters disclosed herein are combinations of (thin film) dichroic and / or absorptive filters, or dichroic and / or absorptive filters alone. Some of the color filters disclosed herein may be bandpass filters that pass frequencies within a certain region (in the passband) but block (attenuate) frequencies outside that region (in the blocking region). Some of the color filters disclosed herein may be multi-bandpass filters that pass multiple, discontinuous wavelength regions. These "frequency bands" may have narrower passbands, greater attenuation in the blocking region, and steeper spectral roll-off than the broad color range of the CFA filter. Roll-off is defined as the steepness of the spectral response at the transition from the passband to the blocking region of the filter. For example, the color filters disclosed herein can cover a passband of about 20 nm to about 40 nm or 20 nm to 40 nm. The particular configuration of the color filters can determine the wavelength bands that are actually incident on the sensor, which can improve the accuracy of the imaging techniques disclosed herein. The color filters described herein can be configured to selectively block or pass specific bands of electromagnetic radiation within the above-mentioned regions, depending on the particular wavelength bands that are appropriate for a particular application.

[0081] The term "pixel" is used herein to describe the output generated by an element of a two-dimensional detector array. In contrast, a photodiode, i.e., a single light-sensitive element, of a two-dimensional detector array behaves as a transducer that can convert photons into electrons via the photoelectric effect and then convert the electrons into a signal that can be used to determine a pixel value. An element of a data cube can be referred to as a "voxel" (e.g., an element of volume). A "spectral vector" refers to a vector that represents the spectral data (e.g., the spectrum of light received from a particular point in object space) at a particular (x,y) location of the data cube. A horizontal plane of a data cube (e.g., an image representing one spectral dimension) is referred to herein as an "image channel." In certain embodiments described herein, spectral video information may be captured, and the resulting data dimension can be N x N y N λ N t We can imagine the shape of a "hypercube" denoted by t is the number of frames captured during the video sequence).

[0082] Overview of an example of a multi-aperture imaging system with curved multi-bandpass filters FIG. 3A shows a schematic diagram of an example of a multi-aperture imaging system 200 with curved multi-bandpass filters according to the present disclosure. The schematic diagram shown includes a first image sensor area 225A (photodiodes PD1-PD3) and a second image sensor area 225B (photodiodes PD4-PD6). The photodiodes PD1-PD6 may be, for example, photodiodes formed on a semiconductor substrate (e.g., a CMOS image sensor). Typically, each photodiode PD1-PD6 may be a single unit of some material, semiconductor, sensor element, or other device capable of converting incident light into an electric current. This diagram shows only a small portion of the entire multi-aperture imaging system for purposes of explaining the structure and operation of the multi-aperture imaging system, and it will be appreciated that in an implementation, the image sensor area may include hundreds or thousands of photodiodes (and corresponding color filters). The first image sensor area 225A and the second image sensor area 225B may be implemented as separate sensors or as separate areas on the same image sensor, depending on the implementation. Although FIG. 3A shows two apertures and corresponding optical paths and sensor areas, it will be appreciated that the optical design principles illustrated in FIG. 3A can be extended to designs including three or more apertures and corresponding optical paths and sensor areas, depending on the implementation.

[0083] The multi-aperture imaging system 200 includes a first opening 210A that provides a first light path to a first sensor area 225A, and a second opening 210B that provides a first light path to a second sensor area 225B. The apertures may be adjustable to increase or decrease the brightness of the light reflected in the image, or the exposure time of a particular image may be changed by adjusting the apertures such that the brightness of the light incident on the image sensor area remains unchanged. The apertures may be located anywhere along the optical axis of the multi-aperture imaging system as would be reasonable for a person skilled in the art of optical design. The optical axis of the optical components located along the first optical path is shown by dashed line 230A, and the optical axis of the optical components located along the second optical path is shown by dashed line 230B, although it will be appreciated that the dashed lines do not represent the physical structure of the multi-aperture imaging system 200. Optical axis 230A and optical axis 230B are separated by a distance D, which can result in parallax between the images captured by first sensor area 225A and the images captured by second sensor area 225B. "Parallax" refers to the distance between two corresponding points on the left and right (or top and bottom) sides of a stereoscopic image, where the same physical point in object space appears at different locations in each image. Processing techniques that correct for and exploit this parallax are described in more detail below.

[0084] The optical axis 230A and the optical axis 230B pass through the center C of the corresponding aperture. Other optical components may also be arranged around these optical axes (e.g., the rotational symmetry points of the optical components may be arranged along this optical axis). For example, the first curved multi-bandpass filter 205A and the first imaging lens 215A may be arranged around the first optical axis 230A, and the second curved multi-bandpass filter 205B and the second imaging lens 215B may be arranged around the second optical axis 230B.

[0085] As used herein, the terms "on" and "above" when discussing the placement of optical elements refer to the location of a particular structure (e.g., a color filter or lens) through which light entering imaging system 200 from object space passes before reaching (or entering) another structure. To explain this, along a first optical path, curved multi-bandpass filter 205A is located above aperture 210A, which is located above imaging lens 215A, which is located above CFA 220A, which is located above first image sensor area 225A. Thus, light entering from object space (e.g., the physical space to be imaged) first passes through first curved multi-bandpass filter 205A, then through aperture 210A, then through imaging lens 215A, then through CFA 220A, and finally enters first image sensor area 225A. The second optical path (e.g., the optical path including the curved multi-bandpass filter 205B, the aperture 210B, the imaging lens 215B, the CFA 220B, and the second image sensor area 225B) may be similarly arranged. In another implementation, the aperture 210A, 210B and / or the imaging lens 215A, 215B may be located above the curved multi-bandpass filter 205A, 205B. Additionally, another implementation may not use a physical aperture and may rely on a clear aperture of the optics to control the brightness of the light imaged onto the sensor area 225A and the sensor area 225B. Thus, the lenses 215A, 215B may be located above the apertures 210A, 210B and the curved multi-bandpass filters 205A, 205B. In this implementation, the apertures 210A, 210B may be located above the lenses 215A, 215B, or the apertures 210A, 210B may be located below the lenses 215A, 215B, as may be determined by one of ordinary skill in the art of optical design.

[0086] The first CFA 220A, located above the first sensor area 225A, and the second CFA 220B, located above the second sensor area 225B, can act as filters that selectively pass certain wavelengths and can split the incoming visible light into red, green, and blue regions (denoted by the symbols R, G, and B, respectively). The light is "split" by passing only the selected wavelengths through the color filters of the first CFA 220A and the second CFA 220B. The split light is received by a dedicated red, green, or blue diode on the image sensor. Although red, blue, and green filters are commonly used, in other embodiments, different color filters can be used depending on the color channels required for the captured image data, such as filters that selectively pass ultraviolet, infrared, or near infrared light, as with an RGB-IR CFA.

[0087] As shown, each filter of the CFA is disposed above a single photodiode PD1-PD6. FIG. 3A also shows an example of a microlens (designated ML) that may be formed or otherwise disposed on each color filter to focus incident light onto the active detector area. Alternative implementations may have multiple photodiodes (e.g., a collection of two, four or more adjacent photodiodes) below a single filter. In the example shown, photodiodes PD1 and PD4 are located below a red filter and therefore output pixel information for a red channel, photodiodes PD2 and PD5 are located below a green filter and therefore output pixel information for a green channel, and photodiodes PD3 and PD6 are located below a blue filter and therefore output pixel information for a blue channel. Additionally, as described in more detail below, the particular color channel output by a given photodiode can be an even narrower frequency band based on the active light source and / or the particular frequency band passed through the multi-bandpass filters 205A, 205B, thereby allowing a given photodiode to output different image channel information with different exposures.

[0088] The imaging lenses 215A, 215B can be shaped to focus the image of the object scene on the sensor area 225A, 225B. Each imaging lens 215A, 215B is not limited to a single convex lens as shown in FIG. 3A, but can be composed of as many optical elements and surfaces as needed to form an image, and various types of imaging lenses or lens assemblies, either commercially available or custom designed, can be used. Each optical element or lens assembly can be formed or coupled to each other to be housed in series or stacked using an opto-mechanical barrel with a retaining ring or bezel. In some embodiments, the optical element or lens assembly can include one or more coupled lens groups, which can include two or more optical components bonded or otherwise coupled together. In various embodiments, the multi-bandpass filters described herein may be placed in front of a lens assembly of a multispectral imaging system, in front of a single lens of a multispectral imaging system, behind a lens assembly of a multispectral imaging system, behind a single lens of a multispectral imaging system, inside a lens assembly of a multispectral imaging system, inside a group of combined lenses of a multispectral imaging system, directly on the surface of a single lens of a multispectral imaging system, or directly on the surface of an element of a lens assembly of a multispectral imaging system. Additionally, apertures 210A and 210B may be eliminated and lenses 215A and 215B may be of the type typically used in digital single-lens reflex (DSLR) or mirrorless camera photography. Additionally, lenses 215A and 215B may be of the type used in machine vision using a C-mount or S-mount thread for attachment.Focus adjustment can be achieved, for example, by moving the imaging lenses 215A, 215B relative to the sensor areas 225A, 225B, or by moving the sensor areas 225A, 225B relative to the imaging lenses 215A, 215B, based on manual focus, contrast-based autofocus or other suitable autofocus techniques.

[0089] Each of the multi-bandpass filters 205A, 205B may be configured to selectively pass a narrow frequency band of light, for example, in some embodiments, a 10-50 nm frequency band (or in other embodiments, a broader or narrower frequency band of light). As shown in FIG. 3A, both of the multi-bandpass filters 205A, 205B may pass a frequency band λc (a "common frequency band"). In implementations with three or more optical paths, each of the multi-bandpass filters may pass this common frequency band. In this manner, each sensor area captures image information in the same frequency band (a "common channel"). As described in more detail below, this image information obtained in the common channel may be used to align a set of images captured by each sensor area. Some implementations may have one common frequency band and a corresponding common channel, or may have multiple common frequency bands and corresponding common channels.

[0090] Each of the multiple bandpass filters 205A, 205B can be configured to selectively pass one or more unique frequency bands in addition to the common frequency band λc. In this manner, the imaging system 200 can increase the number of different spectral channels collectively imaged by the sensor areas 205A, 205B beyond the number of spectral channels that can be imaged by a single sensor area. This aspect is advantageous in that the unique frequency bands λc can be selectively passed. u1 and a multi-bandpass filter 205A that passes a specific frequency band λ u2where λ u1 and λ u2 indicates different frequency bands. Although shown as passing two frequency bands in this figure, the multi-bandpass filters of the present disclosure can each pass more than one frequency band at a time. For example, as described below in connection with Figures 11A and 11B, in some implementations, each multi-bandpass filter can pass four frequency bands. In various embodiments, more frequency bands may be passed. For example, an implementation using four cameras may include a multi-bandpass filter configured to pass eight frequency bands. In some embodiments, the number of frequency bands may be, for example, four, five, six, seven, eight, nine, ten, twelve, fifteen, sixteen, or more.

[0091] The multiple bandpass filters 205A, 205B have curvatures selected to reduce the angle-of-incidence dependent spectral transmission at their respective sensor regions 225A, 225B. As a result, when receiving narrowband illumination from object space, each photodiode (e.g., a color filter covering the sensor region that passes that wavelength) located across the surface of the sensor region 225A, 225B that is sensitive to that wavelength will receive substantially the same wavelength of light without the wavelength shift seen in photodiodes located near the edge of the sensor as discussed above in connection with FIG. 1A. This configuration can produce more accurate spectral image data than would be possible using flat filters.

[0092] Figure 3B shows an example of an optical design for the optical components that make up one optical path of the multi-aperture imaging system shown in Figure 3A. More specifically, Figure 3B shows a custom achromatic doublet 240 that can be used as the multiple bandpass filters 205A, 205B. The custom achromatic doublet 240 passes light through a housing 250 to an image sensor 225. The housing 250 may include apertures 210A, 210B and imaging lenses 215A, 215B as previously described.

[0093] The achromatic doublet 240 is configured to correct optical aberrations due to the incorporation of the required surfaces for the multiple bandpass filter coatings 205A, 205B. The illustrated achromatic doublet 240 includes two lenses that can be made of glass or other optical materials with different dispersions and refractive indices. In other implementations, three or more lenses may be used. These achromatic doublet lens designs incorporate the multiple bandpass filter coatings 205A, 205B on a curved front surface 242, and incorporate the optical surfaces of a curved single lens onto which the filter coatings 205A, 205B are deposited to cancel optical aberrations, and the combined effect of the curved front surface 242 and curved back surface 244 of the achromatic doublet 240 limits the refractive or light gathering power, allowing these lenses alone, housed in the housing 250, to constitute the primary elements for light gathering. The achromatic doublet 240 can thus contribute to improving the accuracy of image data captured by the imaging system 200. These individual lenses can be mounted adjacent to one another, e.g., by bonding or cementing the individual lenses together, and shaped so that the aberrations of one lens are cancelled by the aberrations of the other lens. The curved front surface 242 or the curved back surface 244 of the achromatic doublet 240 can be coated with multiple bandpass filter coatings 205A, 205B. Alternative doublet designs may be implemented in the systems described herein.

[0094] Further variations of the optical designs described herein may be implemented. For example, in some embodiments, a single lens or other optical single lens, such as a convex lens (positive lens) or concave lens (negative lens) as shown in FIG. 3A, may be included in the optical path instead of the doublet 240 shown in FIG. 3B. FIG. 3C shows an example of an implementation in which a flat filter 252 is placed between the lens housing 250 and the sensor 225. The achromatic doublet 240 of FIG. 3C, which incorporates a flat filter 252 with multi-band transmission characteristics, corrects optical aberrations without significantly contributing to the optical power of each lens housed in the housing 250. FIG. 3D shows another example of an implementation in which a multi-bandpass coating is implemented by applying a multi-bandpass coating 254 to the front surface of the lens assembly housed in the housing 250. Thus, this multi-bandpass coating 254 may be applied to any curved surface of any optical element housed in the housing 250.

[0095] 4A-4E show an embodiment of a multispectral, multi-aperture imaging system 300 with the optical design described in connection with FIG. 3A and FIG. 3B. FIG. 4A shows a perspective view of imaging system 300 with housing 305 shown semi-transparently to reveal the internal components. The actual housing 305 may be larger or smaller than the housing 305 shown in the figure, depending, for example, on the desired size of embedded computing resources. FIG. 4B shows a front view of imaging system 300. FIG. 4C shows a cutaway side view of imaging system 300 taken along line CC of FIG. 4B. FIG. 4D shows a bottom view of imaging system 300, showing processing board 335. FIGS. 4A-4D are now described.

[0096] The housing 305 of the imaging system 300 may be contained within another housing. For example, in a portable implementation, the imaging system 300 may be contained within a housing that may have one or more handles shaped to allow the imaging system 300 to be held stably. An example of a portable implementation is shown in more detail in Figures 18A-18C and 19A-19B. The top surface of the housing 305 includes four openings 320A-320D. A different multi-bandpass filter 325A-325D is disposed above each opening 320A-320D and is held in place by a filter cap 330A-330B. The multi-bandpass filters 325A-325D may be curved or non-curved, and as described herein, each multi-bandpass filter passes a common frequency band and at least one unique frequency band, thereby enabling high-precision multispectral imaging in more spectral channels than would normally be captured by an image sensor covered with a color filter array. The image sensors, imaging lenses, and color filters described above are disposed within the camera housings 345A-345D. In some embodiments, the image sensors, imaging lenses, and color filters described above may be housed within a single camera housing, as shown, for example, in FIGS. 20A-20B. Thus, while the illustrated implementation uses separate sensors (e.g., one sensor housed within each of the camera housings 345A-345D), it will be appreciated that other implementations may use a single image sensor disposed over the entire area exposed through the openings 320A-320D. In this embodiment, the camera housings 345A-345D are secured to the imaging system housing 305 using a support 340, but other supports may be used in various implementations.

[0097] The top surface of the housing 305 supports an optional illumination board 310, which is covered with an optical diffusing element 315. The illumination board 310 is described in more detail in connection with FIG. 4E below. The diffusing element 315 may be made of glass, plastic, or other optical materials and diffuses the light emitted from the illumination board 310 so that the object space receives substantially uniform illumination in space. Uniform illumination of the target object may be beneficial in certain imaging applications, such as clinical analysis of tissue by imaging, since it provides a substantially uniform amount of illumination across the object's surface at each wavelength. In some embodiments, the imaging systems disclosed herein may utilize ambient light instead of or in addition to the light from the optional illumination board.

[0098] Because the illumination board 310 generates heat during use, the imaging system 300 includes a heat sink 350 with a number of heat dissipation fins 355. The heat dissipation fins 355 may extend into the space between each of the camera housings 345A-345D, and the top of the heat sink 350 may conduct heat away from the illumination board 310 to the fins 355. The heat sink 350 may be made from a suitable thermally conductive material. The heat sink 350 may assist in dissipating heat from other components, which may eliminate the need for a fan in some implementations of the imaging system.

[0099] A processing board 335 is secured within the housing 305 by a number of supports 365, which communicate with the cameras 345A-345D. The processing board 335 may control the operation of the imaging system 300. Although not shown, the imaging system 300 may also be configured with one or more memories, such as a memory for storing data generated using the imaging system, and / or a computer-executable instruction module for system control. The processing board 335 may be configured in various ways depending on the system design objectives. For example, the processing board may be configured (e.g., via a computer-executable instruction module) to control the activation of specific LEDs on the illumination board 310. In some implementations, a highly stable synchronous buck LED driver may be used, which allows software analog current control of the LEDs and detection of LED failures. In some implementations, the processing board 335 or a separate processing board may further provide image data analysis capabilities (e.g., via a computer-executable instruction module). Although not shown, the imaging system 300 may include a data interconnect between the sensor and the processing board 335 so that the processing board 335 can receive data from the sensor and process the data, and may include a data interconnect between the illumination board 310 and the processing board 335 so that the processing board can activate specific LEDs on the illumination board 310.

[0100] FIG. 4E illustrates an example of an illumination board 310 that may be included in imaging system 300, with the illumination board 310 separated from other components. The illumination board 310 includes four arms extending from a central portion, with three columns of LEDs arranged along each arm. The spacing between the LEDs in adjacent columns is offset laterally from one another to separate adjacent LEDs. Each LED column includes multiple rows of LEDs of different colors. Four green LEDs 371 are arranged in the center, with one green LED at each corner of the center. Starting from the innermost row (e.g., the row closest to the center), each column includes a row of two crimson LEDs 372 (for a total of eight crimson LEDs). Continuing radially outward, each arm has a row in the center column with one amber LED 374, a row on either side with two short blinking blue LEDs 376 (for a total of eight short blinking blue LEDs), another row in the center column with one amber LED 374 (for a total of eight amber LEDs), a row on either side with one non-PPG NIR LED 373 and one red LED 375 (for a total of four non-PPG NIR LEDs and four red LEDs), and a PPG NIR LED 377 in the center column (for a total of four PPG NIR LEDs). "PPG" LEDs refer to LEDs that are activated in multiple successive exposures to capture photoplethysmography (PPG) information representative of pulsatile blood flow in living tissue. It will be appreciated that various other colors and / or arrangements of LEDs may be used in the illumination board in alternative embodiments.

[0101] FIG 5 illustrates another embodiment of a multispectral, multi-aperture imaging system 300 with the optical design described in connection with FIG 3A and FIG 3B. Similar to the design of imaging system 300, imaging system 400 includes four optical paths, shown here as apertures 420A-420D with multiple bandpass filter lens groups 425A-425D secured to housing 405 by retaining rings 430A-430D. Imaging system 400 further includes an illumination board 410 secured to a front portion of housing 405 between retaining rings 430A-430D and a diffuser 415 disposed on illumination board 410 to aid in providing spatially uniform illumination of light onto the target object.

[0102] The illumination board 410 of the imaging system 400 has LEDs arranged in four arms of a cross, with each arm having two rows of closely spaced LEDs. In this manner, the illumination board 410 may be more compact than the illumination board 310 described above, and may be suitable for use with imaging systems having smaller form factor requirements. In this example configuration, each arm has one green LED and one blue LED in the outermost row, and moving inward, includes two rows of yellow LEDs, one row of orange LEDs, one row of red LEDs, one row of crimson LEDs, one row of amber LEDs, and one row of NIR LEDs. Thus, in this implementation, the LEDs are arranged such that the LEDs emitting longer wavelengths of light are located in the center of the illumination board 410, and the LEDs emitting shorter wavelengths of light are located at the ends of the illumination board 410.

[0103] 6A-6C illustrate another embodiment of a multispectral, multi-aperture imaging system 500 with the optical design described in connection with FIGS. 3A and 3B. More specifically, FIG. 6A illustrates a perspective view of the imaging system 500, FIG. 6B illustrates a front view of the imaging system 500, and FIG. 6C illustrates a cutaway side view of the imaging system 500 taken along line CC of FIG. 6B. The imaging system 500 includes similar components as previously described in connection with the imaging system 300 (e.g., a housing 505, an illumination board 510, a diffuser 515, multiple bandpass filters 525A-525D secured above the apertures by retaining rings 530A-530D, etc.), but in a shorter form factor (e.g., in one embodiment, fewer embedded computer components and / or smaller dimensions of embedded computer components). The imaging system 500 further includes a direct camera-to-frame mount 540 for rigidly and robustly fixing the position of the camera alignment.

[0104] 7A-7B show another embodiment of the multispectral, multi-aperture imaging system 600. FIG. 7A-7B show another possible arrangement of the ambient light sources 610A-610C of the multi-aperture imaging system 600. As shown, four lens assemblies with multiple bandpass filters 625A-625D with the optical design described in connection with FIG. 3A-3D can be arranged in a rectangular or square shape to be received by four cameras 630A-630D (including image sensors). Three rectangular light emitting elements 610A-610C can be arranged parallel to each other outside and between each lens assembly with multiple bandpass filters 625A-625D. These rectangular light emitting elements can be broad spectrum light emitting panels, each of which can be arranged with LEDs emitting light in different frequency bands.

[0105] 8A-8B show another embodiment of the multispectral, multi-aperture imaging system 700. FIG. 8A-8B shows another possible arrangement of the ambient light sources 710A-710D of the multi-aperture imaging system 700. As shown, the four lens assemblies with the multiple bandpass filters 725A-725D employing the optical design described in connection with FIG. 3A-3D can be arranged in a rectangular or square shape to be received by the four cameras 730A-730D (including image sensors). The four cameras 730A-730D are shown as an example of a closer arrangement, which may minimize the perspective difference between the lenses. The four rectangular light emitting elements 710A-710D can be arranged in a square shape surrounding the lens assemblies with the multiple bandpass filters 725A-725D. These rectangular light emitting elements may be broad spectrum light emitting panels, each of which may be arranged with LEDs emitting light in different frequency bands.

[0106] 9A-9C show another embodiment of a multispectral, multi-aperture imaging system 800. The imaging system 800 includes a frame 805 connected to a lens assembly front frame 830, which includes an aperture 820 and a support structure for a micro-video lens 825, which may include multiple bandpass filters using the optical design described with respect to FIGS. 3A-3D. The micro-video lens 825 receives light from four cameras 845 (including imaging lenses and image sensor areas) mounted on a lens assembly rear frame 840. Four linear rows of LEDs 811 are arranged along the four sides of the lens assembly front frame 830, with a diffusing element 815 for each LED on each of the four sides. FIGS. 9B and 9C show an example of possible dimensions in inches for the multi-aperture imaging system 800.

[0107] FIG. 10A illustrates another embodiment of a multispectral, multiaperture imaging system 900 with the optical design described in connection with FIGS. 3A-3D. The imaging system 900 can be implemented as a set of multiple bandpass filters 905 that can be attached to a multiaperture camera 915 of a mobile terminal 910. For example, a particular mobile terminal 910, such as a smartphone, can be equipped with a stereoscopic imaging system with two apertures leading to two image sensor areas. The multiaperture spectral imaging techniques disclosed herein can be implemented in such a device by providing a suitable set of multiple bandpass filters 905 that pass multiple narrow frequency bands of light to the sensor areas. The set of multiple bandpass filters 905 can be equipped with a light source (such as an LED array and a diffuser) that irradiates the object space with light in such frequency bands.

[0108] The imaging system 900 may further include a mobile application that configures the mobile terminal to generate and process the multispectral datacube (e.g., in a clinical tissue classification application, a biometric application, a material analysis application, or other application). Alternatively, the mobile application may configure the mobile terminal 910 to transmit the multispectral datacube over a network to a remote processing system and then receive and display the analysis results. An example of a user interface 910 for such an application is shown in FIG. 10B.

[0109] 11A-11B show an example of a set of frequency bands that can pass through four filters when implemented in the multispectral, multi-aperture imaging system shown in FIGS. 3A-10B, and the set of frequency bands is incident on an image sensor having, for example, a Bayer CFA (or another RGB CFA or RGB-IR CFA). JPEG2025013851000002.jpg36170These transmission curves also include the effect of quantum efficiency of the sensor used in this example. As shown, the set of four cameras captures eight unique channels or eight unique frequency bands. Each filter passes two common frequency bands for each camera (the two leftmost peaks) and two other frequency bands. In this implementation, the first and third cameras receive light in the first shared NIR frequency band (the rightmost peak), and the second and fourth cameras receive light in the second shared NIR frequency band (the second peak from the right). Each camera receives one unique frequency band ranging from about 550 nm to about 800 nm or from 550 nm to 800 nm. Thus, the cameras can capture eight unique spectral channels using a compact configuration. Graph 1010 in FIG. 11B shows the spectral irradiance of the LED board shown in FIG. 4E that may be used to illuminate the four cameras shown in FIG. 11A.

[0110] In this implementation, the eight frequency bands were selected based on generating spectral channels suitable for clinical tissue classification, but may be optimized for signal-to-noise ratio (SNR) and frame rate while limiting the number of LEDs (which transmit heat into the imaging system). The eight frequency bands include a common frequency band of blue light (the left-most peak in graph 1000) that passes through all four filters because tissue (e.g., animal tissue such as human tissue) exhibits higher contrast in blue wavelengths than in green and red wavelengths. More specifically, as shown in graph 1000, human tissue has the highest contrast when imaged with a frequency band centered at about 420 nm. The channel corresponding to this common frequency band is used for disparity correction, so that higher contrast allows for more accurate correction. For example, when performing disparity correction, the image processor can use a local or global method to find a set of disparities that maximizes a figure of merit that indicates similarity between local image patches or local images. Alternatively, the image processor can use a similar method to minimize a figure of merit that indicates dissimilarity. These figures of merit may be based on entropy, correlation, absolute difference or deep learning methods. The calculation of disparity by global methods may be performed iteratively and terminated when the figure of merit is stable. Local methods may calculate disparity pixel by pixel, using a fixed patch in one image as input for the figure of merit and evaluating different disparity values ​​for many patches cut from other images during testing. All such methods may constrain the range of disparities considered. The constraint may be based, for example, on knowledge of the depth and distance of the object. It may also be constrained based on the range of expected tilts of the object. The calculated disparity may be constrained by projective geometry, such as epipolar constraints. The output of the disparity calculated at a lower resolution may be used as an initial value or constraint for the calculation of the disparity at the next level of resolution to calculate the disparity.For example, a disparity calculated at a resolution of 4 pixels in a first calculation can be used to set a constraint of ±4 pixels in the next calculation of disparity at a higher resolution. Any algorithm that calculates from disparity benefits from higher contrast, especially when the contrast source is relevant for all viewpoints. Typically, a common frequency band can be selected to accommodate the highest contrast imaging of the materials expected to be imaged in a particular application.

[0111] Since the color separation between adjacent channels after capturing an image may not be perfect, in this implementation, an additional common frequency band is provided as a green frequency band adjacent to the blue frequency band, which passes through all the filters shown in graph 1000. The reason for using such a frequency band is that the pixels of the blue color filter have a wide spectral passband and therefore are sensitive to the green spectral region. This is usually shown as a spectral overlap and may be considered as an intentional crosstalk between adjacent RGB pixels. This spectral overlap allows the spectral sensitivity of the color camera to approach that of the human retina, making the final color space qualitatively similar to human vision. Thus, the common green channel for each filter allows the separation of the part of the signal generated by the blue photodiode that truly corresponds to the received blue light. This separation can be achieved using a spectral unmixing algorithm that factors in the transmittance of the multi-bandpass filters (shown as solid black lines in the T legend) and the corresponding transmittance of the CFA color filters (shown as dashed red, green, and blue lines in the Q legend, respectively). It will be appreciated that in some implementations, red light may be used as the common frequency band, in which case a second common channel may not be required.

[0112] 12 shows a high-level block diagram of an example of a compact imaging system 1100 with high resolution spectral imaging capabilities, the imaging system 1100 comprising a set of components including a processor 1120 coupled to a multi-aperture spectral camera 1160 and a light source 1165. A working memory 1105, a storage device 1110, an electronic display 1125 and a memory 1130 are also in communication with the processor 1120. As described herein, the imaging system 1100 may be able to capture more image channels by placing multiple different bandpass filters over the aperture of the multi-aperture spectral camera 1160 than would be possible with different color filters placed on the CFA of the image sensor.

[0113] The compact imaging system 1100 may be a device such as a mobile phone, a digital camera, a tablet computer, a personal digital assistant, etc. The compact imaging system 1100 may be a stationary device such as a desktop computer or a videoconferencing station using an internal or external camera for image capture. The compact imaging system 1100 may be a combination of an image capture device and a separate processing device that receives image data from the image capture device. A user may have multiple applications available on the compact imaging system 1100. These applications may include traditional photography applications, applications for still and video capture, dynamic color correction applications, brightness and shade correction applications, etc.

[0114] The image capture system 1100 includes a multi-aperture spectral camera 1160 for image capture. The multi-aperture spectral camera 1160 may be, for example, any of the devices shown in Figs. 3A-10B. The multi-aperture spectral camera 1160 may be connected to the processor 1120 so that images captured in different spectral channels can be sent from different sensor areas to the image processor 1120. Also, as described in more detail below, a single or multiple light sources 1165 may be controlled by the processor to emit light of specific wavelengths during a particular exposure. The image processor 1120 may be configured to perform various processes on the received captured images so as to output a high-quality parallax-corrected multi-spectral data cube.

[0115] The processor 1120 may be a general-purpose processing unit or a processor specifically designed for imaging applications. As shown, the processor 1120 is connected to a memory 1130 and a working memory 1105. In the embodiment shown in this figure, the memory 1130 stores an image capture control module 1135, a data cube creation module 1140, a data cube analysis module 1145, and an operating system 1150. These modules contain instructions that configure the processor to perform various image processing and device management tasks. The processor 1120 uses the working memory 1105 to store a working set of processor instructions contained in each module of the memory 1130. Alternatively, the processor 1120 uses the working memory 1105 to store dynamic data created during operation of the image capture system 1100.

[0116] As mentioned above, the processor 1120 is composed of several modules stored in the memory 1130. In some implementations, the imaging control module 1135 includes instructions for configuring the processor 1120 to adjust the focus position of the multi-aperture spectral camera 1160. The imaging control module 1135 further includes instructions for configuring the processor 1120 to capture images, such as multispectral images captured in different spectral channels and PPG images captured in the same spectral channel (e.g., NIR channel), with the multi-aperture spectral camera 1160. Non-contact PPG imaging typically uses near-infrared (NIR) wavelengths as illumination, taking advantage of the increased penetration of photons into tissue at these wavelengths. Thus, the imaging control module 1135, the multi-aperture spectral camera 1160 and the processor 1120 with the working memory 1105 are one of the means for capturing a series of spectral images and / or consecutive images.

[0117] The datacubing module 1140 includes instructions that configure the processor 1120 to generate a multispectral data cube based on the intensity signals received from each photodiode in the different sensor regions. For example, the datacubing module 1140 can estimate the disparity between the same regions of the imaged object based on the spectral channels corresponding to the common frequency bands passed by all the multiple bandpass filters, and use this disparity to align all the spectral images in all the captured channels with each other (e.g., the same point on a particular object is substantially at the same pixel location (x,y) in all the spectral channels). The aligned images can be combined to form a multispectral data cube, and the disparity information can be used to measure the depth of the separately imaged objects, e.g., the difference between the depth of normal tissue and the depth of the deepest point of a wound site. In some embodiments, the datacubing module 1140 can further perform spectral unmixing to determine which portions of the photodiode intensity signals correspond to which frequency bands passed by the filters, e.g., based on a spectral unmixing algorithm that factors in the transmittance of the filters and the quantum efficiency of the sensor.

[0118] The datacube analysis module 1145 can implement various techniques to analyze the multispectral datacube generated by the datacube creation module 1140 for a particular application. For example, some implementations of the datacube analysis module 1145 can provide the multispectral datacube (which may include depth information) to a machine learning model trained to classify each pixel according to a particular condition. The particular condition may be a clinical condition for tissue imaging, such as a burn condition (e.g., first degree burn, second degree burn, third degree burn, or normal tissue type), a wound condition (e.g., hemostasis, inflammation, proliferation, remodeling, or normal skin type), healing potential (e.g., a score reflecting the likelihood that the tissue will heal from the wound condition with or without a particular treatment), a perfusion state, a cancer state, or other tissue condition associated with the wound. The datacube analysis module 1145 can also analyze the multispectral datacube for biometric and / or material analysis purposes.

[0119] The operating system module 1150 configures the processor 1120 to manage memory and processing resources of the image capture system 1100. For example, the operating system module 1150 may include device drivers that manage hardware resources such as the electronic display 1125, the storage device 1110, the multi-aperture spectral camera 1160, the light source 1165, and the like. Thus, in some embodiments, the instructions included in the image processing module described above may not directly interact with these hardware resources, but instead may interact through standard subroutines or APIs included in the operating system component 1150. The instructions in the operating system 1150 may then directly interact with these hardware components.

[0120] Additionally, the processor 1120 may be configured to control the display 1125 to display the captured images and / or the analysis results of the multi-spectral datacube (e.g., classified images) to a user. The display 1125 may be external to the imaging device including the multi-aperture spectral camera 1160 or may be part of the imaging device. The display 1125 may also be configured to provide a viewfinder to the user prior to capturing an image. Additionally, the display 1125 may include an LCD screen or an LED screen and may implement touch-sensitive technology.

[0121] The processor 1120 may write data to the storage module 1110, such as, for example, data representing captured images, multispectral data cubes, and analysis results of the data cubes. In this figure, the storage module 1110 is shown as a conventional disk device, but one skilled in the art will appreciate that the storage module 1110 may be configured as any type of storage media device. For example, the storage module 1110 may include a disk drive, such as a floppy disk drive, a hard disk drive, an optical disk drive, a magneto-optical disk drive, or a solid-state memory, such as a FLASH memory, a RAM, a ROM, and / or an EEPROM. The storage module 1110 may further include multiple memory units, any one of which may be configured to be located inside the image capture device 1100 or may be located outside the image capture system 1100. For example, the storage module 1110 may include a ROM memory that includes system program instructions stored within the image capture system 1100. Additionally, the storage module 1110 may include a memory card or high-speed memory that is removable from the camera and configured to store captured images.

[0122] 12 shows a system including separate components such as a processor, an image sensor, and a memory, those skilled in the art will recognize that these separate components can be combined in various ways to achieve particular design goals. For example, in another embodiment, memory components may be combined with processor components to reduce cost and improve performance.

[0123] Further, while FIG. 12 shows two memory components, a memory component 1130 including several modules and a separate memory 1105 including a working memory, one skilled in the art will recognize that other memory structures may be utilized in some embodiments. For example, a ROM or static RAM memory may be utilized to store processor instructions implementing each module included in the memory 1130. Alternatively, the processor instructions may be read during system startup from a disk storage device integrated into the image capture system 1100 or connected via an external device port. The processor instructions may then be loaded into a RAM to facilitate execution of the instructions by the processor. For example, the working memory 1105 may be a RAM memory, and the instructions may be loaded into the working memory 1105 prior to execution of the instructions by the processor 1120.

[0124] Overview of an example of image processing technology FIG. 13 is a flow chart illustrating an example process 1200 for capturing image data using the multispectral, multi-aperture imaging system illustrated in FIGS. 3A-10B and 12. FIG. 13 illustrates an example of four exposures that can be used to generate a multispectral data cube as described herein: a visible light exposure 1205, an additional visible light exposure 1210, a non-visible light exposure 1215, and an ambient light exposure 1220. It will be appreciated that the exposures may be captured in any order and that some of the exposures may be omitted from or added to certain workflows described below. The process 1200 is described below with reference to the frequency bands illustrated in FIGS. 11A and 11B, although a similar workflow may be implemented using image data generated based on a different set of frequency bands. Additionally, in various embodiments, image acquisition and / or parallax correction may be enhanced by further implementing flat-field correction according to various known flat-field correction techniques.

[0125] To obtain the visible light exposure 1205, a control signal may be sent to the illumination board to light up the LEDs corresponding to the first five peaks (the five peaks on the left side corresponding to visible light in the graph 1000 of FIG. 11A). It may be necessary to stabilize the light output waveform of the particular LEDs simultaneously (e.g., for 10 ms). The capture control module 1135 may then start the exposure of the four cameras and continue this exposure for, for example, about 30 ms. The capture control module 1135 then stops the exposure and acquires data from the sensor areas (e.g., by transferring raw intensity signals from the photodiodes to the working memory 1105 and / or data store 1110). This data may include a common spectral channel used for parallax correction, as described herein.

[0126] To improve the signal-to-noise ratio, in some implementations, an additional visible light exposure 1210 may be taken using the same process as described for visible light exposure 1205. Taking two identical or nearly identical exposures may improve the signal-to-noise ratio, allowing for more accurate analysis of the image data. However, in implementations where the signal-to-noise ratio of only a single image is acceptable, this second exposure may be omitted. Also, in some implementations, taking two exposures in a common spectral channel may provide more accurate parallax correction.

[0127] In some implementations, a non-visible light exposure 1215 corresponding to NIR or IR light can be further captured. For example, the capture control module 1135 can activate two NIR LEDs corresponding to the two NIR channels shown in FIG. 11A. It may be necessary to stabilize the light output waveform of certain LEDs simultaneously (e.g., for 10 milliseconds). The capture control module 1135 can then start an exposure of the four cameras and continue this exposure for, for example, about 30 milliseconds. The capture control module 1135 then stops the exposure and acquires data from the sensor areas (e.g., by transferring raw intensity signals from the photodiodes to the working memory 1105 and / or data store 1110). Since this exposure can safely assume no change in the shape or position of the object relative to the exposures 1205 and 1210, the NIR channels can be aligned using previously calculated disparity values, so there is no need to have a common frequency band passing through all the sensor areas.

[0128] In some implementations, multiple exposures may be used to capture a sequence of images to generate PPG data indicative of deformation of the tissue site due to pulsatile blood flow. In some implementations, such PPG exposures may be captured at non-visible wavelengths. Combining the PPG data with the multispectral data may improve the accuracy of certain medical imaging analyses, but capturing the PPG data may increase the time required for the image capture process. In some implementations, the increased time required for the image capture process may result in movement of the portable imaging device and / or the object, which may introduce errors. Thus, in certain implementations, capturing the PPG data may be omitted.

[0129] In some implementations, an ambient light exposure 1220 may be further taken. In this exposure, all LEDs may be turned off and an image may be taken using ambient lighting (e.g., sunlight or light from other sources). The capture control module 1135 may then initiate an exposure of the four cameras and continue this exposure for a desired time (e.g., about 30 milliseconds). The capture control module 1135 may then stop the exposure and acquire data from the sensor area (e.g., by transferring raw intensity signals from the photodiodes to the working memory 1105 and / or data store 1110). The effect of ambient light may be removed from the multispectral data cube by subtracting the intensity values ​​of the ambient light exposure 1220 from the numerical values ​​of the visible light exposure 1205 (or the visible light exposure 1205 corrected for signal-to-noise ratio by the second exposure 1210) and also from the numerical values ​​of the non-visible light exposure 1215. Such processing can improve the accuracy of downstream analysis by isolating portions of the resulting signal indicative of light emitted by the light source and light reflected from the object / tissue site. In some implementations, this step may be omitted if sufficient analytical accuracy is achieved using only visible light exposures 1205, 1210 and non-visible light exposures 1215.

[0130] It will be appreciated that the particular exposure times listed above are examples in one implementation and that in other implementations the exposure times may vary depending on the image sensor, the intensity of the light source, and the object being imaged.

[0131] Figure 14 shows a simplified block diagram of a workflow 1300 for processing image data, such as image data captured using the multispectral, multi-aperture imaging system shown in Figures 3A-10B and 12 and / or using process 1200 shown in Figure 13. Although workflow 1300 shows the output of two RGB sensor areas 1301A, 1301B, the number of sensor areas in workflow 1300 can be increased and there can be sensor areas corresponding to different CFA color channels.

[0132] The RGB sensor outputs from the two sensor regions 1301A, 1301B are stored in 2D sensor output modules 1305A, 1305B, respectively. These sensor region values ​​are sent to non-linear mapping modules 1310A, 1310B, which can perform parallax correction by determining the disparity between images captured using a common channel and then applying this determined disparity across all channels to align all spectral images with each other.

[0133] The output of these non-linear mapping modules 1310A, 1310B are then provided to a depth calculation module 1335, which can calculate the depth of a particular region of interest in the image data. For example, the depth may represent the distance between the object and the image sensor. In some implementations, the depth of the object relative to some other object than the image sensor can be measured by calculating and comparing multiple depth values. For example, the deepest depth of the wound bed can be measured, and the depth of the normal tissue surrounding the wound bed (deepest, shallowest or average depth) can also be measured. The deepest depth of the wound can be measured by subtracting the normal tissue depth from the wound bed depth. This depth comparison can also be performed at other points in the wound bed (e.g., all or some of the given sampling points), thereby constructing a 3D map of the wound depth at various points (shown in FIG. 14 as z(x,y) where z is the depth value). Larger disparities impose higher computational burden on algorithms for calculating depth as described above, although in some embodiments larger disparities may improve depth calculations.

[0134] The output of these nonlinear mapping modules 1310A, 1310B are then also provided to a linear equation module 1320, which can process the sensed values ​​as a set of linear equations for spectral unmixing. In one implementation, the Moore-Penrose pseudo-inverse equation can be used as a function of at least the quantum efficiency of the sensor and the transmittance values ​​of the filters to calculate the actual spectral values ​​(e.g., the intensity of light of a particular wavelength incident on each image point (x,y)). Spectral unmixing can be used in implementations requiring high accuracy, such as clinical diagnostics and other biological applications. Spectral unmixing can also be applied to provide estimates of photon flux and signal-to-noise ratio.

[0135] Based on the parallax-corrected spectral channel images and spectral unmixing, the workflow 1300 can generate a spectral datacube 1325, such as in the form shown in the figure of F(x,y,λ), where F denotes the light intensity at a particular image location (x,y) at a particular wavelength or frequency band λ.

[0136] FIG. 15 illustrates parallax and parallax correction for processing image data, such as image data captured using the multispectral, multi-aperture imaging system shown in FIGS. 3A-10B and 12 and / or using the process shown in FIG. 13. A first set of images 1410 shows image data of the same physical location on the same object captured at four different sensor regions. As shown, the object locations are not at the same location in these raw images based on the (x,y) coordinate frame of the photodiode grid of the image sensor region. A second set of images 1420 shows the same object location after parallax correction, where the images are at the same (x,y) location in the coordinate frame of the aligned images. It will be appreciated that such alignment may include cropping certain data from edge regions of the images that do not completely overlap one another.

[0137] FIG. 16 illustrates a workflow 1500 for pixel-by-pixel classification of multispectral image data, such as image data captured using the multispectral, multi-aperture imaging system shown in FIGS. 3A-10B and 12 and / or using the process shown in FIG. 13 and then processed according to FIGS. 14 and 15 .

[0138] In block 1510, a multispectral, multi-aperture imaging system 1513 can capture image data indicative of physical points 1512 on an object 1511. In this example, the object 1511 includes tissue of a patient having a wound. The wound can include a burn, a diabetic ulcer (e.g., a diabetic foot ulcer), a non-diabetic ulcer (e.g., a pressure ulcer or a slow-healing wound), a chronic ulcer, a post-operative incision, an amputation site (before or after an amputation), a cancerous lesion, or damaged tissue. When PPG information is included, the imaging system disclosed herein provides a method for evaluating conditions involving changes in tissue blood flow and changes in pulse rate, including tissue perfusion; cardiovascular health; injuries such as ulcers; peripheral arterial disease; and respiratory health.

[0139] In block 1520, the data captured by the multispectral, multi-aperture imaging system 1513 may be processed to generate a multispectral datacube 1525 having multiple different wavelengths 1523, which may include multiple different images (PPG data 1522) of the same wavelength corresponding to different times. For example, the image processor 1120 may be configured via the datacube creation module 1140 to generate the multispectral datacube 1525 according to the workflow 1300. Some implementations may be associated with depth values ​​at various points in the spatial dimensions, as previously described.

[0140] In block 1530, the multispectral data cube 1525 can be analyzed as input data 1525 to a machine learning model 1532 to generate a classification mapping 1535 of the imaged tissue. The classification mapping allows each pixel in the image data (each pixel representing a particular point on the imaged object 1511 after registration) to be assigned a particular tissue classification or a particular healability score. Different classifications or scores can be indicated in the classified image output by using visually distinct colors or patterns. Thus, even if many images of the object 1511 are taken, the output can be a single image of the object (e.g., a regular RGB image) overlaid with multiple visual indications of the pixel-by-pixel classification.

[0141] In some implementations, the machine learning model 1532 may be an artificial neural network. An artificial neural network is a computational entity inspired by biological neural networks but adapted for implementation by a computer device. Although dependencies between inputs and outputs are not easily found, artificial neural networks can be used to model complex relationships between inputs and outputs and to find certain patterns in data. A neural network typically includes an input layer, one or more intermediate layers ("hidden layers"), and an output layer, each of which includes many nodes. The number of nodes may vary from layer to layer. A neural network that includes two or more hidden layers is considered to be "deep". The nodes of each layer are connected to all or some of the nodes of the next layer, and the weights of these connections are typically learned from data during a training process, for example using backpropagation to adjust the parameters of the neural network to generate expected outputs from corresponding inputs in the labeled training data. An artificial neural network can thus be thought of as an adaptive system configured to change its structure (e.g., connections and / or weights) during training based on the information flowing through it, with the hidden layer weights encoding meaningful patterns in the data.

[0142] A fully connected neural network is a neural network in which each node in the input layer is connected to each node in the next layer (the first hidden layer), each node in the first hidden layer is connected to each node in the next hidden layer, and so on until each node in the final hidden layer is connected to each node in the output layer.

[0143] A CNN is a type of artificial neural network that, like the previously described artificial neural networks, is composed of multiple nodes with learnable weights. Unlike the previously described artificial neural networks, each layer of a CNN may have nodes arranged in three dimensions, with width, height, and depth corresponding to the pixel values ​​of a 2×2 array of each video frame (e.g., width and height) and the number of video frames in the sequence (e.g., depth). The nodes of each layer may be locally connected to a small region of the previous layer, called a receptive field, of width and height. The weights of the hidden layer may take the form of a convolution filter applied to the receptive field. In some embodiments, the convolution filter may be two-dimensional, so that the convolution (or convolution transformation of an image) can be repeated with the same filter for each frame of the input volume or a specified subset of frames. In another embodiment, the convolution filter may be three-dimensional, so that it spans the full depth of multiple nodes of the input volume. Each node in each convolutional layer of a CNN shares weights, duplicating the convolutional filters of a given layer across the width and height of the input volume (e.g., across all frames), thereby reducing the total number of trainable weights and improving the applicability of the CNN to datasets outside the training data. Values ​​from certain layers may be pooled to reduce the number of calculations in the next layer (e.g., values ​​representing certain pixels may be propagated to the next layer, while others are discarded), and furthermore, as the CNN progresses in depth, a pooling mask may be reintroduced to the discarded values, returning the number of data points relative to the original dimensions. Multiple layers, some of which may be fully connected, may be stacked to form a CNN structure.

[0144] During training, the artificial neural network can be applied to the paired data in the training data to vary parameters and predict the output of a particular pair given the input. For example, the training data can include a multispectral data cube (input) and a labeled classification mapping (expected output), such as by a clinician designating wound areas that correspond to particular clinical conditions, which may include a healed (1) or unhealed (0) label that is assigned some time after the initial imaging of the wound and actual healing is confirmed. Alternative implementations of the machine learning model 1532 can be trained to make other types of predictions, such as predicting the likelihood that a wound will heal and reduce the wound area by a certain percentage within a certain time period (e.g., at least a 50% reduction in wound area within 30 days), or predict the wound condition classified as hemostasis, inflammation, pathogen colonization, proliferation, remodeling, normal skin, etc. In some implementations, patient measurements may be further incorporated into the input data to further improve classification accuracy, or the training data may be segmented based on the patient measurements and used on different patients with the same measurements to train the machine learning model 1532 on different examples. The patient measurements may include textual information or medical history or aspects thereof describing characteristics of the patient or the patient's health, such as, for example, the area of ​​the wound, lesion or ulcer; the patient's BMI; the patient's diabetic status; the presence of peripheral vascular disease or chronic inflammation in the patient; the number of other wounds the patient currently has or has had; whether the patient is receiving or has recently received immunosuppressive drugs (e.g., chemotherapy) or other medications that positively or negatively affect wound healing rate; HbA1c; stage IV chronic renal failure; whether the patient has type 1 or type 2 diabetes; chronic anemia; asthma; medication use; smoking status; diabetic neuropathy; deep vein thrombosis; previous myocardial infarction; transient ischemic attack; and sleep apnea; and any combination thereof.These measurements can be converted into vector representations with appropriate processing, such as vector embeddings of words, vectors of binary values ​​indicating whether a patient meets a particular measurement (e.g., whether or not he or she has type 1 diabetes), or numerical values ​​indicating the extent of each measurement in the patient.

[0145] At block 1540, a classification mapping 1535 can be output to a user. In this example, the classification mapping 1535 uses a first color 1541 to indicate pixels classified according to a first state and a second color 1542 to indicate pixels classified according to a second state. The classification and resulting classification mapping 1535 can exclude background pixels based on, for example, object recognition, background color identification, and / or depth values. As shown in the figure, some implementations of the multispectral, multiaperture imaging system 1513 can project the classification mapping 1535 onto the tissue site. Such an embodiment can be particularly beneficial when the classification mapping includes a visual indication of the recommended resection margin and / or resection depth.

[0146] Such methods and systems may assist clinicians and surgeons in the process of managing skin wounds, such as making decisions on burn excision, amputation level, lesion removal, and wound triage. The embodiments described herein can be used to identify and / or classify the severity of decubitus ulcers, hyperemia, limb deterioration, Raynaud's phenomenon, scleroderma, chronic wounds, abrasions, lacerations, bleeding, ruptures, puncture wounds, penetration wounds, skin cancer (basal cell carcinoma, squamous cell carcinoma, malignant melanoma, or actinic keratosis), or any type of tissue change in which the characteristics and properties of the tissue differ from normal. The devices described herein can also be used to monitor normal tissue; aid and improve wound treatment methods (e.g., allowing for faster and more sophisticated approaches to determining debridement margins); and evaluate the progress of wound or disease recovery (especially after treatment has been applied). Some embodiments described herein provide devices capable of identifying normal tissue adjacent to damaged tissue, determining margins and / or depth during resection, monitoring the healing process after implantation of a prosthetic device such as a left ventricular assist device, assessing the viability of tissue or regenerative cell grafts, or monitoring post-operative recovery (particularly following reconstructive surgery). Additionally, embodiments described herein can be used to assess wound changes or post-wound normal tissue development, particularly following administration of therapeutic agents such as steroids, hepatocyte growth factor, fibroblast growth factor, antibiotics, regenerative cells (e.g., isolated or enriched cell populations including stem cells, endothelial cells, and / or endothelial progenitor cells).

[0147] Overview of an example distributed computing environment Figure 17 shows a simplified block diagram of an example of a distributed computing system 1600 including a multispectral, multi-aperture imaging system 1605, which may be any of the multispectral, multi-aperture imaging systems shown in Figures 3A-10B and 12. As shown, a datacube analysis server 1615 may include one or more computers, perhaps arranged as a server cluster or server farm, with the memory and processors that make up these computers either located within one computer or distributed across many computers (including computers located remotely from one another).

[0148] The multispectral and multi-aperture imaging system 1605 may include networking hardware (e.g., wireless internet, satellite, Bluetooth or other communication) for communicating with the user devices 1620 and the data cube analysis server 1615 via the network 1610. For example, in some implementations, the processor of the multispectral and multi-aperture imaging system 1605 may be configured to control image capture and then transmit the raw data to the data cube analysis server 1615. In other implementations, the processor of the multispectral and multi-aperture imaging system 1605 may be configured to control image capture, perform spectral unmixing and parallax correction to generate a multispectral data cube, and then transmit the multispectral data cube to the data cube analysis server 1615. In some implementations, all processing and analysis may be performed locally on the multispectral and multi-aperture imaging system 1605, and the multispectral data cube and resulting analysis results may be transmitted to the data cube analysis server 1615 for performing aggregate analysis and / or for use in training or retraining machine learning models. Thus, the datacube analysis server 1615 may provide the latest machine learning models to the multispectral and multiaperture imaging system 1605. The processing load for the final result of analyzing the multispectral datacube may be shared in various ways between the multiaperture imaging system 1605 and the datacube analysis server 1615 depending on the processing power of the multiaperture imaging system 1605.

[0149] The network 1610 may include any suitable network, such as an intranet, the Internet, a cellular network, a local area network, other similar networks, or combinations thereof. The user device 1620 may include any network-equipped computing device, such as a desktop computer, a laptop, a smart phone, a tablet, an e-reader, a video game console, etc. For example, when tissue classification is required, the results (e.g., classified images) measured by the multi-aperture imaging system 1605 and the data cube analysis server 1615 may be transmitted to a patient or physician designated user device, a hospital information system that stores the patient's electronic medical records, and / or a centralized public health database (e.g., maintained by the Centers for Disease Control and Prevention, etc.).

[0150] An example of the implementation result background: Burn morbidity and mortality pose a challenge to wounded combatants and their medical personnel. Burn injury rates in historical combat casualties range from 5-20%, with approximately 20% of these casualties requiring complex burn surgery, such as at the U.S. Army Institute of Surgical Research (ISR) Burn Center. Burn surgery requires specialized training, and is therefore performed by ISR staff rather than U.S. Army hospital staff. Limited numbers of burn specialists greatly complicate the logistics of providing medical care to burned soldiers. Thus, the use of novel objective methods for detecting burn depth preoperatively and intraoperatively can ensure that a greater number of medical personnel (including personnel from non-ISR organizations) are available to provide medical care to burned soldiers during ongoing combat. Increased medical personnel availability can expand the scope of more complex burn care to advance the medical role for burned combatants.

[0151] To begin to address this need, we developed a novel cart-based imaging device that uses multispectral imaging (MSI) and artificial intelligence (AI) algorithms to aid in the preoperative assessment of burn healability. The device can image a large area of ​​tissue (e.g., 5.9 x 7.9 in2) in a short time (e.g., within 6 seconds, 5 seconds, 4 seconds, 3 seconds, 2 seconds, or 1 second) and does not require the injection of imaging contrast agents. In this study of the general public, the accuracy of the device in assessing burn healability was shown to be superior (e.g., reaching 70-80%) to clinical judgment by burn specialists.

[0152] method: Civilian patients with burns of various severity were imaged within 72 hours of burn injury and at several time points up to 7 days after burn injury. The exact burn severity in each image was determined using healing assessments or punch biopsies over a 3-week period. The accuracy of the device in identifying and distinguishing healing from non-healing burn tissue in first-, second-, and third-degree burns was analyzed on a pixel-by-pixel basis.

[0153] result: Data was collected from 38 civilian patients, totaling 58 burns and 393 images. The AI ​​algorithm achieved a sensitivity of 87.5% and a specificity of 90.7% in predicting non-healing burn tissue.

[0154] Conclusion: The accuracy of the novel device and its AI algorithm in determining the healing potential of burns was shown to be superior to the clinical judgment of burn specialists. Future studies will focus on redesigning the device to make it portable and evaluating its use in an intraoperative setting. Design modifications to make the device portable include reducing the size of the device to that of a portable system, increasing the field of view, shortening the acquisition time for a single snapshot, and evaluating the device for use in an intraoperative setting using a porcine model. These developments were performed using a benchtop multispectral imaging (MSI) subsystem that showed comparable capabilities in basic imaging studies.

[0155] Additional light source for image alignment In various embodiments, one or more additional light sources may be used in conjunction with any of the embodiments disclosed herein to improve the accuracy of image alignment. FIG. 21 illustrates an example embodiment of a multi-aperture spectral imager 2100 including a projector 2105. In some embodiments, the projector 2105 or other suitable light source may be, for example, one of the light sources 1165 described above with respect to FIG. 12. In embodiments including an additional light source such as a projector 2105 for alignment, the method may further include additional exposures. The additional light source such as the projector 2105 may project one or more dots, stripes, grids, random spots, or other suitable spatial patterns in a single spectral band, multiple spectral bands, or broadbands, individually or cumulatively visible through all cameras of the imager 2100, into the field of view of the imager 2100. For example, the projector 2105 may project shared or common channel light, broadband illumination, or cumulatively visible illumination that may be used to verify the accuracy of image alignment calculated based on the common band approach described above. As used herein, "cumulatively visible illumination" refers to a selected plurality of wavelengths whose pattern is converted by each image sensor of a multispectral imaging system. For example, cumulatively visible illumination can include a plurality of wavelengths whose pattern is converted by each channel, even if the plurality of wavelengths does not include a wavelength that is common to all channels. In some embodiments, the type of pattern projected by projector 2105 may be selected based on the number of apertures into which the pattern is imaged.For example, if the pattern is only visible through one aperture, it may be preferable for the pattern to be relatively dense (e.g., it may have a relatively narrow autocorrelation of about 1-10 pixels, about 20 pixels, less than about 50 pixels, less than about 100 pixels, etc.), whereas if the pattern is imaged through multiple apertures, a less dense pattern or a pattern with a less narrow autocorrelation may be useful. In some embodiments, an additional exposure taken with the projected spatial pattern is included in the parallax calculation with the intention of improving the accuracy of the alignment over an embodiment in which the exposure is taken without the projected spatial pattern. In some embodiments, the additional light source can project a stripe pattern of a single spectral band, multiple spectral bands, or a broadband (e.g., a shared or common channel, or a broadband illumination, etc.) that can be individually or cumulatively visible through all cameras into the field of view of the imager, and use this to improve the alignment of the images based on the phase of the stripe pattern. In some embodiments, the additional light source can improve image alignment by projecting multiple unique spatial arrangements of dots, grids and / or specks of a single spectral band, multiple spectral bands or broadband (e.g., shared or common channel, or broadband illumination, etc.) into the field of view of the imaging device, which can be individually or cumulatively visible through all cameras. In some embodiments, the method further includes an additional sensor with an aperture or multiple apertures that can detect the shape of a single or multiple objects in the field of view. For example, the additional sensor may use LIDAR, light field or ultrasound technology to further improve the accuracy of image alignment using the common band approach described above. The additional sensor may be a single aperture or multi-aperture sensor capable of sensing light field information, or other signals such as ultrasound or pulsed laser.

[0156] Implementing machine learning for wound assessment, healing prediction and treatment An example of an embodiment of a machine learning system and method for wound assessment, healing prediction and treatment is described below. Any of the various imaging devices, systems, methods, techniques and algorithms described herein may be applied to the field of wound imaging and wound analysis. The implementation described below may include obtaining one or more images of the wound in one or more known wavelength bands, and based on the one or more images, performing one or more of the following: segmenting the image into wound and non-wound portions, predicting the percentage reduction in the wound area after a predetermined period of time, predicting the healing potential of individual parts of the wound after a predetermined period of time, displaying a visual indication related to the segmentation or prediction as described above, indicating whether a standard wound care therapy or an advanced wound care therapy should be selected, and other functions.

[0157] In various embodiments, the wound assessment system or clinician can determine the appropriate level of wound care therapy based on the results of the machine learning algorithms disclosed herein. For example, if the output of the wound healing prediction system indicates that the imaged wound will close more than 50% within 30 days, the wound healing prediction system can request or notify the medical professional or patient to apply a standard care therapy. If the output of the wound healing prediction system indicates that the imaged wound will not close more than 50% within 30 days, the wound healing prediction system can request or notify the medical professional or patient to use one or more advanced wound care therapies.

[0158] In treating existing wounds, wounds such as diabetic foot ulcers (DFUs) may be administered one or more standard wound care regimens, such as standard of care (SOC) as defined by the Centers for Medicare and Medicaid, during the first 30 days of treatment. The standard of care regimen (SOC) may include one or more of the following as an example of a standard wound care plan: optimization of nutritional status; debridement by any means to remove necrotic tissue; maintenance of a clean and moist granulation tissue bed with an appropriate moist dressing; treatment required for resolution of any infection that may be present; addressing the lack of intravascular perfusion to the limb associated with the DFU; reducing pressure load on the DFU; and appropriate glycemic control. During the first 30 days of standard of care regimen (SOC), measurable signs of healing of the DFU are defined as a decrease in the size of the DFU (wound surface area or wound volume), a decrease in the amount of exudate from the DFU, and a decrease in the amount of necrotic tissue within the DFU. An example of the progression of healing of a DFU is shown in FIG. 22.

[0159] Generally, advanced wound care (AWC) is administered when no healing is observed during the first 30 days of standard of care (SOC). The Centers for Medicare and Medicaid have not published a description or definition of advanced wound care (AWC), but it can be considered any treatment that does not fall under standard of care (SOC) as defined above. Advanced wound care (AWC) is an area of ​​intense research and innovation, with new options being introduced almost continuously for clinical use. Therefore, coverage for advanced wound care (AWC) is determined on a case-by-case basis, and some patients may not be eligible for reimbursement for treatments that are considered advanced wound care (AWC). Based on this understanding, advanced wound care (AWC) may include, but is not limited to, one or more of the following: hyperbaric oxygen therapy; negative pressure wound therapy; bioengineered skin substitutes; synthetic growth factors; extracellular matrix proteins; matrix metalloproteinase regulators; and electrical stimulation therapy. An example of the healing progression of a non-healing DFU is shown in FIG.

[0160] In various embodiments, the wound assessment and / or wound healing prediction described herein may be based solely on one or more images of the wound or based on one or more images of the wound in combination with patient health data (e.g., one or more health measurements, clinical characteristics, etc.). The techniques described herein may take a single image or a set of multispectral images (MSIs) of a patient tissue site, such as an ulcer or other wound, process the images using the machine learning systems described herein, and output one or more healing prediction parameters. A variety of healing parameters may be predicted by the techniques of the present invention. Some examples of healing prediction parameters are: (1) a binary yes / no value regarding whether the ulcer heals to a reduction in ulcer area of ​​more than 50% (or another desired threshold (%) according to clinical standards) within 30 days (or another desired time period according to clinical standards); (2) the percentage that indicates the likelihood of the ulcer healing to a greater than 50% reduction in ulcer area (or another desired threshold (%) according to clinical standards) within 30 days (or another desired time period according to clinical standards); and (3) A prediction of the actual ulcer area that is expected to decrease within 30 days (or another desired time period according to clinical standards) due to healing of the ulcer. These include, but are not limited to: In another example, a system in accordance with the present technology may provide binary yes / no values ​​or percentages indicating the likelihood of healing for smaller portions of a wound, such as individual pixels or subsets of pixels in an image of the wound, and these yes / no values ​​or percentages indicating the likelihood of healing may indicate whether the individual portions of the wound are tissue that is likely to heal after a specified period of time or tissue that is likely not to heal after a specified period of time.

[0161] FIG. 24 illustrates one example of an approach for providing such a healing prediction. As shown, a multispectral imaging sensor is used to provide images of a wound, or a set of multispectral images, taken at different wavelengths, at different times, or simultaneously, to provide input and output values ​​to a neural network, such as an autoencoder neural network (a type of artificial neural network, described in more detail below). This type of neural network can reduce the number of features represented by the input, and herein, the number of values ​​(e.g., numerical values) representing pixel values ​​in a single or multiple input images. Such feature counts can be provided to, for example, a fully connected feedforward artificial neural network or a machine learning classifier, such as the system illustrated in FIG. 25, to output a healing prediction for the imaged ulcer or other wound.

[0162] FIG. 25 illustrates an example of another approach for providing a healing prediction as described above. As shown, an image (or a set of multispectral images taken at different wavelengths at different times or simultaneously using a multispectral image sensor) is provided as input to a neural network, such as a convolutional neural network ("CNN"). A CNN works with a two-dimensional ("2D") array of pixel values ​​(e.g., numbers along the height and width of the image sensor used to capture the image data) and provides an output that is a one-dimensional ("1D") representation of the image. The numbers can indicate a classification of each pixel in the input image, for example, a classification of each pixel according to one or more physiological conditions related to an ulcer or other wound.

[0163] As shown in FIG. 25, the patient measurements data repository can store other types of information about the patient, referred to herein as patient measurements, clinical variables, or health measurements. The patient measurements can include textual information describing characteristics of the patient, such as the area of ​​the ulcer; the patient's body mass index (BMI); the number of other wounds the patient currently has or has had; diabetic status; whether the patient is receiving or has recently received immunosuppressants (e.g., chemotherapy) or other medications that positively or negatively affect wound healing rates; HbA1c; chronic renal failure stage IV; whether the patient has type 1 or type 2 diabetes; chronic anemia; asthma; medication use; whether the patient smokes; diabetic neuropathy; deep vein thrombosis; previous myocardial infarction; transient ischemic attack; and sleep apnea; and any combination thereof. However, various other measurements can be used. Various exemplary measurements are shown in Table 1 below. [Table 1-1] [Table 1-2] [Table 1-3] [Table 1-4]

[0164] These measurements can be converted into vector representations by suitable processing, such as vector embeddings of words, vectors of binary values ​​indicating whether a patient has a particular measurement (e.g., whether or not the patient has type 1 diabetes), or numerical values ​​indicating the extent of each measurement in the patient. In various embodiments, any one of these patient measurements or a combination of all or a subset of these patient measurements can be used to improve the accuracy of the healing prediction parameters generated by the systems and methods of the present technology. In an exemplary test, image data was obtained at the first visit for treatment of a DFU, and it was found that the percent area reduction of the DFU could be accurately predicted with an accuracy of about 67% when the image data was analyzed alone without considering clinical variables. Predictions based on patient history alone were approximately 76% accurate, using the most important features of wound area, BMI, number of previous wounds, HbA1c, stage IV chronic renal failure, type 1 or type 2 diabetes, chronic anemia, asthma, medication use, smoking status, diabetic neuropathy, deep vein thrombosis, previous myocardial infarction, transient ischemic attack, and sleep apnea. When these medical variables were combined with imaging data, the accuracy of predictions was observed to increase to approximately 78%.

[0165] In one exemplary embodiment shown in Figure 25, image data represented in one dimension (1D) can be concatenated with patient measurements represented as vectors. This concatenated value can then be provided as an input to a fully connected neural network, which outputs healing prediction parameters.

[0166] The system shown in Figure 25 can be considered as a single machine learning system with multiple machine learning models and a vector-represented patient measurement generator. In some embodiments, the entire machine learning system can be trained in an end-to-end manner so that the CNN and fully connected network can adjust parameters via backpropagation, and the vector-represented patient measurements can be added to the values ​​passed through the CNN and fully connected network to create healing prediction parameters from the input image.

[0167] An example of a machine learning model Artificial neural networks are man-made in the sense that they are computational entities inspired by biological neural networks but modified for implementation by computer devices. Although dependencies between inputs and outputs are not easily found, artificial neural networks can be used to model complex relationships between inputs and outputs and to find certain patterns in data. A neural network typically includes an input layer, one or more intermediate layers ("hidden layers"), and an output layer, each of which contains a large number of nodes. The number of nodes in each layer may vary. A neural network that includes two or more hidden layers is considered to be "deep". The nodes in each layer are connected to all or some of the nodes in the next layer, and the weights of these connections are usually learned based on the training data during the training process, for example using backpropagation to adjust the parameters of the neural network to generate the expected output from the corresponding inputs in the labeled training data. Thus, an artificial neural network may be an adaptive system configured to change its structure (e.g., connections and / or weights) during training based on the information flowing through the artificial neural network, and the hidden layer weights can be viewed as encoding meaningful patterns in the data.

[0168] A fully connected neural network is a neural network in which each node in the input layer is connected to each node in the next layer (the first hidden layer), each node in the first hidden layer is connected to each node in the next hidden layer, and so on until each node in the final hidden layer is connected to each node in the output layer.

[0169] An autoencoder is a neural network that includes an encoder and a decoder. A particular autoencoder aims to compress input data in an encoder and then restore the encoded data in a decoder, resulting in an output that is a good and complete reconstruction of the original input data. An example of an autoencoder neural network described herein, such as the autoencoder neural network shown in FIG. 24, can use pixel values ​​of an image of a wound (e.g., structured in a vector or matrix format) as input to its input layer. The next layer or layers, i.e., the "encoder layers", encode the input information by reducing the dimensionality of the information (e.g., by representing the input values ​​using fewer dimensions than the n dimensions of the original input values), and another layer or layers ("decoder layers") that follow these encoder layers decode the encoded information to generate feature vectors as output at the output layer. An example of a training process for an autoencoder neural network may be unsupervised learning, which learns the parameters of the hidden layer that generate output data that matches the input data provided. Thus, the number of nodes in the input layer and the number of nodes in the output layer usually match. Dimensionality reduction allows an autoencoder neural network to learn the most important features of an input image, where the autoencoder's middle layer (or another middle layer) represents the "feature reduced" input values. In some examples, an autoencoder neural network can reduce an image of, for example, about 1 million pixels (where each pixel value can be considered a separate feature of the image) to a feature set of about 50 numbers. The dimensionality reduction method of representing images can also be used in other machine learning models, such as the classifier shown in FIG. 25, a suitable CNN or other neural network, to output healing prediction parameters.

[0170] A CNN is a type of artificial neural network, and like the previously described artificial neural networks, it is composed of multiple nodes with learnable weights between each node. Unlike the previously described artificial neural networks, each layer of a CNN may have nodes arranged in three dimensions, with width, height, and depth corresponding to the pixel values ​​of a 2×2 array of each image frame (e.g., width and height) and the number of image frames in an image sequence (e.g., depth). In some embodiments, the nodes of each layer may be locally connected to a small region of the previous layer, called a receptive field, of width and height. The weights of the hidden layer may take the form of a convolution filter applied to the receptive field. In some embodiments, the convolution filter may be two-dimensional, so that the convolution (or convolution transformation of an image) can be repeated with the same filter for each frame of the input volume or a specified subset of frames. In another embodiment, the convolution filter may be three-dimensional, so that it spans the full depth of multiple nodes of the input volume. Each node in each convolutional layer of a CNN shares weights, duplicating the convolutional filters of a given layer across the width and height of the input volume (e.g., across all frames), thus reducing the total number of trainable weights and improving the applicability of the CNN to data sets outside the training data. Values ​​from a particular layer may be pooled to reduce the number of calculations in the next layer (e.g., values ​​representing a particular pixel may be passed on to the next layer, while the other values ​​are discarded), and furthermore, as the CNN progresses in depth, a pooling mask may be reintroduced to the discarded values, returning the number of data points relative to the original magnitude. Multiple layers, some of which may be fully connected, may be stacked to form a CNN structure. During training, an artificial neural network may be applied to the paired data in its training data to change parameters and predict the output of a particular pair given the input.

[0171] Artificial intelligence refers to a system running on a computer that is capable of performing tasks that would normally require human intelligence. The artificial intelligence system disclosed herein is capable of performing image analysis (and other data analysis) that would otherwise require the skill and intelligence of a human physician. The artificial intelligence system disclosed herein has the advantage of being able to make such predictions at the patient's first visit, rather than making the patient wait 30 days for a wound healing assessment.

[0172] The ability to learn is an important aspect of intelligence, since systems without this ability cannot usually gain knowledge from experience. Machine learning is a field of computer science that gives computers the ability to learn without being explicitly programmed, for example, allowing artificial intelligence systems to learn complex tasks and adapt to changes in the environment. The machine learning system disclosed herein can be trained with a large amount of labeled training data to determine the healing potential of a wound. Through such machine learning, the artificial intelligence system disclosed herein can learn new relationships between the appearance of a wound (captured as image data, such as by MSI) and the healing potential of the wound.

[0173] The artificial intelligence and machine learning systems disclosed herein include computer hardware, one or more memories, and one or more processors, for example, as described with respect to the various imaging systems and imaging devices disclosed herein. Any of the machine learning systems and / or methods according to the present techniques may be implemented on or in communication with the processors and / or memories of the various imaging systems and imaging devices disclosed herein.

[0174] An example of the implementation of multispectral imaging of diabetic foot ulcers (DFUs) In an exemplary application of the machine learning systems and methods disclosed herein, following imaging on day 0, a machine learning algorithm as described above was used to predict the % area reduction (PAR) of imaged wounds at day 30. To perform this prediction, a machine learning algorithm was trained to take multispectral imaging (MSI) data and clinical variables as inputs, and output a scalar value representing the predicted % area reduction. After 30 days, individual wounds were assessed to measure the true % area reduction. The predicted % area reduction was compared to the measured true % area reduction in an assessment of wound healing over a 30-day period. The performance of the algorithm was measured using the coefficient of determination (R 2 ) was used to score the score.

[0175] The machine learning algorithm in this application was a bagged ensemble of decision tree classifiers adapted to use data from DFU's image database. Other suitable ensembles of classifiers, such as the XGBoost algorithm, may be implemented as well. DFU's image database contained 29 individual images of diabetic foot ulcers from 15 subjects. The true percent area reduction at day 30 for each image was known. The algorithm was trained using leave-one-out cross-validation (LOOCV). R 2 The score was calculated after combining the prediction results for the test images obtained from each LOOCV split.

[0176] The MSI data consisted of 2D images of eight channels, each of which represented diffusely reflected light from tissues at a particular wavelength filter. The field of view of each channel was 15 cm × 20 cm, with a resolution of 1044 pixels × 1408 pixels. The eight wavelength bands included 420 nm ± 20 nm; 525 nm ± 35 nm; 581 nm ± 20 nm; 620 nm ± 20 nm; 660 nm ± 20 nm; 726 nm ± 41 nm; 820 nm ± 20 nm; and 855 nm ± 30 nm, where "±" indicates the half-width of each spectral channel. The eight wavelength bands are shown in Figure 26. As quantitative features, the mean, median, and standard deviation of all pixel values ​​were calculated from each channel.

[0177] Additionally, clinical variables were obtained from each subject: age, degree of chronic kidney disease, length of DFU on day 0, and width of DFU on day 0.

[0178] For MSI datacubes using 1–8 channels, separate algorithms were created using features extracted from every possible combination of the 8 channels (wavelength bands) (in total, C1(8)+C2(8)+...+C8(8)=255 different feature sets). 2 The values ​​were calculated and ordered from smallest to largest. 2 The 95% confidence intervals of R values ​​were calculated from the predicted results of the algorithms trained on each feature set. To determine whether there was an improvement above chance for a particular feature set, feature sets were identified that did not include 0.0 within the 95% confidence interval of the results of the algorithm trained on that feature set. In addition, the same analysis was performed 255 more times including all clinical variables for each feature set. To determine whether the clinical variables influenced the performance of the algorithms, a t-test was used to compare the R values ​​obtained from the 255 algorithms trained with the clinical variables. 2The average values ​​of were compared to 255 algorithms trained without including clinical variables. The results of the analysis are shown in Tables 2 and 3 below. Table 2 shows the performance of the feature set that does not include clinical variables and includes only image data. [Table 2]

[0179] As shown in Table 2, the best performing feature set that did not include clinical features was one that included only three of the eight channels available in the MSI data. It was observed that the 726 nm wavelength band was included in each of the top five feature sets. The bottom five feature sets each included only one wavelength band. Although the 726 nm wavelength band was included in each of the top five feature sets, it was observed that the worst performance was observed when only this 726 nm wavelength band was used. Table 3 below shows the performance of the feature sets that include image data and the clinical variables age, degree of chronic kidney disease, DFU length at day 0, and DFU width at day 0. [Table 3]

[0180] The best performing feature set including clinical variables was one that included all eight channels available in the MSI data. All five of the top feature sets included the 855 nm wavelength band. Histograms of the models including and excluding clinical variables are shown in Figure 27 with vertical lines indicating the mean of each distribution.

[0181] R between all feature sets not including clinical variables by comparing the importance of clinical features 2 The average R 2 We determined whether the R obtained from models trained on feature sets that did not include clinical variables was equivalent to the mean 2The mean R obtained from models trained on feature sets including clinical variables was 0.31, whereas the mean R obtained from models trained on feature sets including clinical variables was 0. 2 The mean value of was found to be 0.32. The difference between the means was calculated by t-test, with a p-value of 0.0443. Therefore, the model trained with clinical variables was found to be significantly more accurate than the model trained without clinical features.

[0182] Extracting features from image data Although the above application examples have described the extraction of pixel mean, standard deviation, and median, various other features may be extracted from the image data and used to create the healing prediction parameters. Feature types include local features, semi-local features, and global features. Local features may indicate the texture of the image patch, while global features may include contour representations, shape descriptors, and texture features. Global texture features and local texture features provide different information about a particular image due to different support in computing the texture. In some cases, global features allow the whole object to be generalized with a single vector. On the other hand, local features are computed at multiple points in the image, allowing for robust analysis of partially occluded or low quality images. However, specialized classification algorithms may be required to handle the variation in the number of feature vectors per image.

[0183] Local features may include, for example, scale-invariant feature transform (SIFT), speeded-up robust features (SURF), features from accelerated segment test (FAST), binary robust invariant scalable keypoints (BRISK), Harris corner detection operator, binary robust independent elementary features (BRIEF), oriented FAST and rotated BRIEF (ORB), and KAZE features. Semi-local features may include, for example, edges, splines, lines, and moments within a small window. Global features may include, for example, color, Gabor features, wavelet features, Fourier features, texture features (e.g., first, second, and higher moments), neural network features from 1D, 2D, and 3D convolutional or hidden layers, and principal component analysis (PCA).

[0184] Application example in RGB-based DFU imaging As a further example for obtaining a healing prediction parameter, a similar MSI method may be used based on RGB data obtained from a photographic digital camera or the like. In this case, the algorithm may obtain data from the RGB image, and may also obtain data from the subject's medical history or other clinical variables, and may output a healing prediction parameter, such as a conditional probability that a DFU will or will not respond to a standard wound care regimen over a 30-day period. In some embodiments, this conditional probability is the probability that the DFU being treated is a non-healing ulcer, given that x is the input data for the model characterized by a parameter θ, as follows: P モデル (y=“non-healing”│x;θ)

[0185] The scoring method for the RGB data may be similar to the scoring method in the MSI application described above. In one example, the predicted non-healing areas may be compared to the true non-healing areas measured by a 30-day healing assessment of wounds such as DFUs. The results of this comparison indicate the performance of the algorithm. The method for performing such a comparison may be based on the clinical outcome of the output image.

[0186] In this application example, four outcomes can be obtained for each healing prediction parameter generated by the healing prediction algorithm. A true positive (TP) outcome indicates that the wound area is reduced by less than 50% (e.g., the DFU is a non-healing ulcer) and the algorithm predicts that the wound area is reduced by less than 50% (e.g., the machine learning device outputs a prediction that it is non-healing). A true negative (TN) outcome indicates that the wound area is reduced by at least 50% (e.g., the DFU is a healing ulcer) and the algorithm predicts that the wound area is reduced by at least 50% (e.g., the machine learning device outputs a prediction that it is healing). A false positive (FP) outcome indicates that the wound area is reduced by at least 50%, but the algorithm predicts that the wound area is reduced by less than 50%. A false negative (FN) outcome indicates that the wound area is reduced by less than 50%, but the algorithm predicts that the wound area is reduced by at least 50%. After prediction and evaluation of actual cure, these results can be summarized using the performance measures of accuracy, sensitivity and specificity shown in Table 4 below. [Table 4]

[0187] In a retrospective study, an image database of DFUs was obtained, containing 149 individual images of diabetic foot ulcers from 82 subjects. 69% of the DFUs in this dataset were considered "healed" because they had reached the target of 50% wound area reduction at 30 days. The mean wound surface area was 3.7 cm. 2 and the median wound area was 0.6 cm 2 It was.

[0188] Color photographic images (RGB images) were used as input data for the developed model. The RGB images consisted of three-channel 2D images, each of which represented the light diffusely reflected from tissue at wavelengths utilized by conventional color camera sensors. The images were taken by clinicians using a handheld digital camera. The imager, working distance, and field of view (FOV) selections varied between images. Before training the algorithm, each image was manually cropped to position the ulcer in the center of the field of view (FOV). After cropping, the images were interpolated to a size of 3 channels × 256 pixels × 256 pixels. During this interpolation process, we did not control the preservation of the aspect ratio of the original images. However, the aspect ratio can be maintained during these preprocessing steps if desired. A clinical dataset (e.g., clinical variable values ​​and health measurements) was also obtained from each subject, including medical history, past wounds, and blood test results.

[0189] To perform this analysis, we developed two algorithms whose initial goal is to identify new representations of image data that can be combined with patient health measurements in traditional machine learning classification approaches. There are many methods available for creating these image representations, including principal component analysis (PCA) and scale-invariant feature transform (SIFT). In one example, we compressed the images using a separately trained unsupervised approach and performed machine learning to predict DFU healing. In a second example, we used an end-to-end supervised approach to predict DFU healing.

[0190] The unsupervised feature extraction approach described above used, for example, an autoencoder algorithm consistent with the method illustrated in FIG. 24. An example of an autoencoder is shown diagrammatically in FIG. 28. The autoencoder included an encoder module and a decoder module. The encoder module was a 16-layer VGG convolutional network. At the 16th layer, a compressed image representation was presented. The decoder module was a 16-layer VGG network with an added upsampling function and removed the pooling function. JPEG2025013851000011.jpg18170

[0191] The autoencoder was pre-trained using PASCAL visual object classes (VOC) data and fine-tuned using the DFU images in our example dataset. Each individual image, containing 3 channels × 256 pixels × 256 pixels (total number of pixels: 65,536), was aggregated into a vector per element and compressed into multiple vectors representing 50 data points. After training, the same encoder-decoder algorithm was used for all images in the dataset.

[0192] After extracting the image compression vector, the image compression vector was used as input to a second supervised machine learning algorithm. Combinations of image features and patient features were tested using various machine learning algorithms, including logistic regression, k-nearest neighbors, support vector machines, and various decision tree models. An example of a supervised machine learning algorithm using the image compression vector and patient clinical variables as inputs to predict DFU healing is shown diagrammatically in FIG. 29. The machine learning algorithm may be any of a variety of known machine learning algorithms, such as multi-layer perceptron, quadratic discriminant analysis, naive Bayes, or an ensemble of such algorithms.

[0193] An end-to-end machine learning approach that was investigated as an alternative to the unsupervised feature extraction approach described above is shown diagrammatically in Figure 30. In this end-to-end approach, a 16-layer VGG CNN was modified by concatenating patient health measurements data to the image vectors in the first fully connected layer. In this way, the encoder module and the subsequent machine learning algorithm can be trained simultaneously. Other methods involving global variables (e.g., patient health measurements or clinical variables) have also been proposed to improve the performance of the CNN or to repurpose it. The most commonly used method is the feature-wise linear modulation (FiLM) generative model. The supervised machine learning algorithm was trained using k-fold cross-validation. The results for each image were calculated as one of the following: true positive, true negative, false positive, and false negative. These results were summarized using the performance measures listed in Table 4.

[0194] The prediction accuracy of the unsupervised feature extraction (autoencoder) and machine learning approaches shown in Figures 28 and 29 was determined using seven machine learning algorithms and three input feature combinations shown in Figure 31. Each algorithm was trained using three-fold cross-validation and reported as the average accuracy (± 95% confidence interval). Of the algorithms trained using this approach, only two algorithms were able to exceed the accuracy baseline. The accuracy baseline was set when a simple classifier predicted all DFUs as healing ulcers. The two algorithms that were able to exceed the baseline were logistic regression and support vector machine, which included a combination of image data and patient data. Patient health measurements that were important for predicting DFU healing and used in these models included wound area; body mass index (BMI); number of previous wounds; hemoglobin A1c (HbA1c); renal failure; type 1 or type 2 diabetes; anemia; asthma; medication use; smoking status; diabetic neuropathy; deep vein thrombosis (DVT); previous myocardial infarction (MI); and combinations of these.

[0195] Using the end-to-end machine learning approach shown in Figure 30, the results showed significantly better performance than the baseline, as shown in Figure 32. However, compared to the unsupervised approach, the performance of this approach was not significantly better, but the average accuracy was higher than the other methods tested.

[0196] Predicting healing of a subset of wound areas In a further exemplary embodiment, the system and method according to the present technology can not only generate a single healing probability for the entire wound, but can also predict the area of ​​tissue in each individual wound that is not expected to heal after 30 days with standard wound care. To obtain this output, a machine learning algorithm was trained to take MSI or RGB data as input and generate healing prediction parameters for a portion of the wound (e.g., individual pixels in an image of the wound or a subset thereof). The present technology can be further trained to output a visual display, such as an image highlighting areas of ulcerated tissue that are predicted not to heal within 30 days.

[0197] FIG. 33 illustrates an example process for predicting healing and generating a visual display. As described elsewhere herein, the method illustrated in FIG. 33 obtains a spectral data cube. The data cube is sent to machine learning software for processing. The machine learning software can implement all or part of pre-processing, machine learning wound assessment model, and post-processing. The machine learning module outputs a conditional probability map that is processed by a post-processing module (e.g., probability thresholding) to generate a visually outputtable result in the form of a classified image for the user. As shown as an image output to the user in FIG. 33, the system of the present invention can display an image of the wound to the user, where pixels of areas expected to heal and pixels of areas expected not to heal are displayed with different visual displays.

[0198] The process shown in FIG. 33 was applied to a set of DFU images, and the predicted non-healed areas were compared to the true non-healed areas measured by a 30-day healing assessment of the DFU. The results of this comparison indicate the performance of the algorithm. The method for performing this comparison was based on the clinical outcomes of the output images. The DFU image database contained 28 individual images of diabetic foot ulcers from 19 subjects. In each image, the true area of ​​the wound that did not heal after 30 days of standard wound care was known. The algorithm was trained using a leave-one-out cross-validation (LOOCV) method. The results for each image were calculated as one of the following: true positive, true negative, false positive, and false negative. The results were summarized using the performance measures listed in Table 4.

[0199] A convolutional neural network was used to create a conditional probability map for each input image. The algorithm includes an input layer, a convolutional layer, a deconvolutional layer, and an output layer. The MSI data or RGB data is usually input to the convolutional layer. The convolutional layer is usually composed of a convolutional stage (e.g., an affine transformation) whose output is used as an input to a detection stage (e.g., a nonlinear transformation such as rectified linear [ReLU]), the result of which may be used for the next convolutional stage and the detection stage. The result obtained may be downsampled by a pooling function or may be used directly as the result of the convolutional layer. The result obtained from the convolutional layer is provided as the input to the next layer. The deconvolutional layer usually starts with an inverse pooling layer, which is followed by a convolutional stage and a detection stage. These layers are usually composed in the following order: input layer, convolutional layer, deconvolutional layer. This configuration is often considered as an encoder layer arranged first, followed by a decoder layer. The output layer typically consists of multiple fully connected neural networks applied to each vector in any of the dimensions of the tensor output from the previous layer. The results from such fully connected neural networks are aggregated into a matrix called a conditional probability map.

[0200] Each entry in the conditional probability map represents a region of the original DFU image. Each region may be a one-to-one mapping with each pixel of the input MSI image, or it may be an n-to-1 mapping, where n represents an aggregate of several pixels of the original image. The conditional probability value in this map represents the probability that the tissue contained in that region of the image will not respond to standard wound care. This result is obtained by segmenting the pixels in the original image, and the predicted non-healing regions are segmented from the predicted healing regions.

[0201] The results from a particular layer in the convolutional neural network can be modified with information from another source. In this example, this modification can be done using clinical data from the subject's medical history or treatment plan (e.g., patient health measurements or clinical variables as described herein) as the source. Thus, the results from the convolutional neural network can be adjusted depending on the magnitude of variables other than the imaging data. To implement this, a feature-wise linear transformation layer (FiLM layer) can be incorporated into the network architecture as shown in FIG. 34. The FiLM layer is a machine learning algorithm trained to learn the parameters of an affine transformation, which is applied to one of multiple layers in the convolutional neural network. The input to this machine learning algorithm is a numerical vector, in this case a numerical vector representing the clinically meaningful medical history of the patient in the form of patient health measurements or clinical variables. The training of this machine learning algorithm may be done simultaneously with the training of the convolutional neural network. One or more FiLM layers using different inputs and machine learning algorithms can be applied to different layers of the convolutional neural network.

[0202] The input data for conditional probability mapping included multispectral imaging (MSI) data and color photographic images (RGB images). The MSI data consisted of eight channel 2D images, each of which represented the diffusely reflected light from tissue at a specific wavelength filter. The field of view for each channel was 15 cm × 20 cm, with a resolution of 1044 pixels × 1408 pixels. The eight wavelengths included 420 nm ± 20 nm, 525 nm ± 35 nm, 581 nm ± 20 nm, 620 nm ± 20 nm, 660 nm ± 20 nm, 726 nm ± 41 nm, 820 nm ± 20 nm, and 855 nm ± 30 nm, as shown in Figure 26. The RGB images included three channel 2D images, each of which represented the diffusely reflected light from tissue at wavelengths used in conventional color camera sensors. The field of view of each channel was 15 cm x 20 cm in size and had a resolution of 1044 pixels x 1408 pixels.

[0203] To segment images based on cure probability, we used a CNN architecture called SegNet. According to the SegNet developers, the model uses RGB images as input and outputs a conditional probability map. We further modified the SegNet architecture to allow the use of 8-channel MSI images in the input layer. We also modified the SegNet architecture to include a FiLM layer.

[0204] To demonstrate the ability to segment DFU images into healed and non-healed regions, we developed various deep learning models, each utilizing different inputs. These models used two types of input features: only MSI data or only RGB images. In addition to varying input features, we also varied the manner in which the algorithm was trained. Some of these changes included pre-training the deep learning model with the PASCAL visual object classes (VOC) dataset, pre-training the deep learning model with a database of images of different types of tissue wounds, pre-specifying the kernels of the input layer using a filter bank, early stopping, random image augmentation during the training of the algorithm, and averaging the results of the random image augmentation during inference to generate a single aggregated conditional probability map.

[0205] The top two deep learning models using one of the two input features were identified to perform better than chance. Replacing the RGB data with MSI data improved the results; the number of image-based errors decreased from 9 to 7. However, both the MSI and RGB data methods proved viable for creating conditional probability maps of the cure potential of DFUs.

[0206] In addition to evaluating whether the SegNet architecture can segment images of wounds with the desired accuracy, we also evaluated whether images of other types of wounds, though unexpected, would be suitable for use in a training system to segment images of DFUs or other wounds based on the conditional healing probability mapping. As previously mentioned, the SegNet CNN architecture may be suitable for segmenting DFU images when pre-trained using DFU image data as training data. However, in some cases, a suitable training image set of a large number of images may not be available for a particular type of wound. Figure 35 shows an example of a color image of a DFU (A) and four examples of how the DFU was segmented into areas predicted to heal and areas predicted not to heal by various segmentation algorithms. In image (A), taken on the first day of evaluation, the area of ​​the wound identified as a non-healing wound in the evaluation performed four weeks later is shown in dashed lines. Images (B)-(E), which correspond to different segmentation algorithms, show the predicted non-healing tissue in color. As shown in image (E), the SegNet algorithm was highly accurate in predicting non-healing tissue regions that closely matched the dashed outlines in image (A) corresponding to regions determined to be non-healing by actual inspection, even though it was pre-trained on a database of burn images rather than a database of DFU images. In contrast, a naive Bayesian linear model trained on DFU image data (image (B)), a logistic regression model trained on DFU image data (image (C)), and a SegNet pre-trained on PACAL VOC data (image (D)) all performed less well, showing significantly larger and less accurately outlined non-healing tissue regions in images (B)-(D).

[0207] An example of single wavelength analysis of a DFU image In another implementation, it was found that the assessment of wound percent area reduction (PAR) at day 30 and / or segmentation in the form of a conditional probability map can also be performed based on single wavelength band image data, without using MSI or RGB image data. To implement this method, a machine learning algorithm was trained to take as input features extracted from single wavelength band images and output a scalar value indicating the predicted percent area reduction.

[0208] All images were obtained from subjects following an Institutional Review Board (IRB) approved clinical trial protocol. The dataset included 28 individual images of diabetic foot ulcers from 17 subjects. All subjects were imaged at their first visit for wound treatment, and all wounds were at least 1.0 cm wide in their largest dimension. Only subjects who were prescribed standard wound care therapy were included in the study. Healing of DFUs was assessed by clinicians during routine follow-up visits to determine true percent area reduction (PAR) after 30 days of treatment. During this healing assessment, images of the wounds were collected and compared to images taken on day 0 to determine true percent area reduction (PAR).

[0209] Various machine learning algorithms may be used, such as classifier ensembles. Two machine learning algorithms for regression were used in this analysis. One algorithm was a bagged ensemble of decision tree classifiers (bagged decision trees), and the other was a random forest ensemble. All features used to train the machine learning regression models were obtained from pre-treatment DFU images taken at the first visit for treatment of the DFUs included in this study.

[0210] Eight grayscale images of each DFU were acquired at unique wavelengths in the visible or near infrared spectrum. The field of view for each image was approximately 15 cm x 20 cm with a resolution of 1044 pixels x 1408 pixels. The eight unique wavelengths were selected using an optical bandpass filter set with wavelength bands of 420 nm ± 20 nm, 525 nm ± 35 nm, 581 nm ± 20 nm, 620 nm ± 20 nm, 660 nm ± 20 nm, 726 nm ± 41 nm, 820 nm ± 20 nm, and 855 nm ± 30 nm, as shown in FIG. 26.

[0211] The raw image, 1044 pixels by 1408 pixels, contained a reflectance intensity value for each pixel. Quantitative features were calculated based on the reflectance intensity values ​​and included the first and second moments of the reflectance intensity values ​​(e.g., mean and standard deviation). In addition, the median was also calculated.

[0212] After these calculations are performed, the filter sets may be applied individually to the raw image to create multiple transformed images. In a particular example, a total of 512 filters with dimensions of 7 pixels by 7 pixels or another suitable kernel size may be used. FIG. 36 shows an example of a set of 512 filter kernels with a size of 7 pixels by 7 pixels that may be used in an example implementation. This example, but not limited to, filter set may be obtained via training a convolutional neural network (CNN) for segmentation of the DFU. The 512 filters shown in FIG. 36 were obtained from the first kernel set of the input layer of the CNN. The "learning" of these filters was adjusted by updating the weights to avoid large deviations from the Gabor filters included in the filter bank.

[0213] Filters may also be applied to the raw images by convolution. From the 512 images resulting from the application of these convolution filters, a single three-dimensional matrix may be constructed with dimensions of 512 channels by 1044 pixels by 1408 pixels. Additional features may be further calculated from this three-dimensional matrix. For example, in some embodiments, the mean, median and standard deviation of the intensity values ​​of the three-dimensional matrix may be calculated as additional features for input to a machine learning algorithm.

[0214] In addition to the six features mentioned above (e.g., the mean, median and standard deviation of pixel values ​​of the raw image, and the mean, median and standard deviation of pixel values ​​of a three-dimensional matrix constructed by applying a convolution filter to the raw image), additional features and / or linear or non-linear combinations of such features may be further included as desired. For example, the product of two features or the ratio of two features may be used as a new input feature to the machine learning algorithm. In one example, the product of the mean and median may be used as an additional input feature.

[0215] The algorithm was trained using leave-one-out cross-validation (LOOCV). One DFU image was set aside as a test set, and the remaining DFU images were used as a training set. After training, the model was used to predict the percent area reduction of the set of DFU images. After the prediction was completed, the set of DFU images was added back to the full set of DFU images, and the process was repeated with another set of DFU images. The LOOCV method was repeated until each DFU image had been used once as a holdout set (test set). After collecting the test set results for each cross-validation fold, the overall performance of the model was calculated.

[0216] The predicted percent area reduction for each DFU image was compared to the measured true percent area reduction for each DFU over a 30-day healing period. The performance of the algorithm was evaluated using the coefficient of determination (R 2) was used to score the score. 2 The usefulness of each wavelength was determined by using the R 2 The R value is an index showing the proportion of variance in the percent reduction in the area of ​​the DFU that is explained by features extracted from the DFU images. 2 The value is defined by the following formula:

number

number

number

[0217] The goal of this study was to determine whether using each of the eight different wavelengths independently in a regression model would produce results that were significantly better than chance. To determine whether there was an improvement above chance for a particular feature set, the R of the algorithm trained on that feature set was used. 2A feature set was identified that did not include zero within the 95% confidence interval of the values. To perform this evaluation, eight independent experiments were performed in which machine learning models were trained using the original six features, i.e., the mean, median, and standard deviation of the raw images; and the mean, median, and standard deviation of a three-dimensional matrix created by transforming the raw images by applying a convolution filter. Random forest and bagged decision tree models were trained. As a result, we reported algorithms that performed well in cross-validation. The results of these eight models were reexamined to see whether the lower limit of the 95% confidence interval (CI) was above zero. If the lower limit of the 95% confidence interval (CI) was not above zero, additional features created by a nonlinear combination of the original six features were employed.

[0218] Using the initial six features, seven of the eight wavelengths tested could be used to create a regression model that explained a large portion of the variance in the percent area reduction from the DFU dataset. These seven wavelengths were ranked from most effective to least effective as follows: 660 nm; 620 nm; 726 nm; 855 nm; 525 nm; 581 nm; 420 nm. When the product of the mean and median of the three-dimensional matrix was included as an additional feature, the last wavelength listed in the table below (820 nm) was found to be effective. The results of this testing are summarized in Table 5. [Table 5]

[0219] Thus, it has been demonstrated that the imaging and analysis systems and methods described herein appear to be capable of accurately generating one or more healing predictive parameters even when based on images from a single wavelength band. In some embodiments, the use of a single wavelength band may facilitate the calculation of one or more aggregate quantitative features derived from the images, such as the mean, median or standard deviation of the raw image data and / or the mean, median or standard deviation of a set of multiple images or a three-dimensional matrix generated by applying one or more filters to the raw image data.

[0220] An exemplary wound image segmentation system and method As previously described, the machine learning techniques described herein can be used to reliably predict parameters associated with wound healing, such as overall wound healing (e.g., % area reduction) and / or healing associated with a portion of a wound (e.g., probability of healing associated with an individual pixel or subset of pixels of a wound image), by analyzing a spectral image containing reflectance data at individual or multiple wavelengths. Additionally, some of the methods disclosed herein can predict wound healing parameters based at least in part on aggregated quantitative features, e.g., statistical quantities such as mean, standard deviation, median, etc., calculated based on a subset of pixels in the wound image that are determined to be pixels corresponding to wound tissue regions, i.e., "wound pixels," as opposed to calluses, normal skin, background, or other non-wound tissue regions. Thus, to improve or optimize the accuracy of such predictions based on a set of wound pixels, it is preferable to accurately select the subset of wound pixels in the wound image.

[0221] Segmentation of an image (e.g., an image of a DFU) into wound and non-wound pixels has traditionally been performed manually, for example by a physician or other clinician inspecting the image and selecting a set of wound pixels based on the image. However, such manual segmentation is time-consuming, inefficient, and subject to human error. For example, formulas used to calculate area and volume lack the accuracy and precision required to measure raised areas of a wound. Furthermore, determining the true border of a wound and classifying tissue within the wound (e.g., epithelial proliferation) requires advanced capabilities. Since changes in measurements in a wound are often critical information for determining the effectiveness of treatment, errors in initial wound measurements can lead to inaccurate treatment decisions.

[0222] To address such concerns, systems and methods according to the present technology are suitable for automatic wound margin detection and tissue type identification in wound regions. In some embodiments, systems and methods according to the present technology can be configured to automatically segment wound images into at least wound and non-wound pixels, with aggregate quantitative features calculated based on a subset of the wound pixels to a desired degree of accuracy. Furthermore, it may be desirable to implement a system or method capable of segmenting wound images into wound and non-wound pixels and / or further segmenting the wound or non-wound pixels into one or more subclassifications without requiring the development of additional healing prediction parameters.

[0223] Color photographs of wounds may be used to develop a dataset of images of diabetic foot ulcers. A variety of color camera systems may be used in acquiring this data. In one implementation, a total of 349 images were used. A trained physician or other clinician may use a software program to identify and label the wound, callus, normal skin, background, and / or other types of pixels in each wound image. These labeled images, known as truth masks, may contain multiple colors corresponding to the number of types of labeling in the image. FIG. 37 shows an example image of a DFU (left) and its corresponding truth mask (right). The example truth mask shown in FIG. 37 includes purple areas corresponding to background pixels, yellow areas corresponding to callus pixels, and cyan areas corresponding to wound pixels.

[0224] Based on a set of ground truth images, a convolutional neural network (CNN) can be used to automatically segment the aforementioned types of tissue. In some embodiments, the structure of the algorithm may be a shallow layer U-net with multiple convolutional layers. In one implementation, 31 convolutional layers have been used to obtain the desired segmentation results. However, various other algorithms for image segmentation may be applied to obtain the desired output.

[0225] In one example of segmentation implementation, the image database of the DFU was randomly divided into three sets, and a training set of 269 images was used to train the algorithm, a test set of 40 images was used to select hyperparameters, and a validation set of 40 images was used for validation. The algorithm was trained using the steepest descent method, and the accuracy of the test set was monitored. When the accuracy of the test set reached a maximum, the training of the algorithm was stopped. The validation set was then used to judge the results of the algorithm.

[0226] The results from each image in the validation set obtained by the U-net algorithm were compared to the corresponding truth mask. This comparison was performed by comparing pixels to pixels. For each of the three tissue types, the comparison results were summarized using the following classification: True positives (TP) included the total number of pixels in the truth mask where the tissue of interest was present, but the machine learning model predicted the presence of the same tissue. True negatives (TN) included the total number of pixels in the truth mask where the tissue of interest was not present, but the machine learning model predicted the presence of the same tissue. False positives (FP) included the total number of pixels in the truth mask where the tissue of interest was not present, but the machine learning model predicted the presence of the same tissue. False negatives (FN) included the total number of pixels in the truth mask where the tissue of interest was present, but the machine learning model predicted the presence of the same tissue. These results were summarized using the following measures:

[0227]

number

[0228]

number

[0229]

number

[0230] In some embodiments, the algorithm may be trained for multiple epochs, and a number of epochs intermediate between overfitting and underfitting may be determined to optimize accuracy. In one implementation described herein, the image segmentation algorithm was trained for 80 epochs. Training was monitored and it was found that accuracy on the test dataset was best with 73 epochs.

[0231] The performance of the U-net segmentation algorithm was measured by accuracy and found to be better than chance. Furthermore, when predicting only one tissue type using a simple classifier, U-net outperformed all three available simple classifiers. Regardless of the overfitting issue, the performance of this machine learning model on the validation set was shown to be viable based on the metrics used to summarize the results.

[0232] Figure 38 shows three examples of wound image segmentation results using the U-net segmentation algorithm combined with the methods described herein. In each of the three examples, an image of a DFU is shown in the right column, and the automated image segmentation output produced by the U-net segmentation algorithm trained as described herein is shown in the middle column. The manually created truth masks corresponding to each DFU image are shown in the left column of Figure 38. These images visually demonstrate the high accuracy of segmentation using the methods described herein.

[0233] term All of the methods and tasks described herein may be performed by a computer system and may be fully automated. In some cases, the computer system includes multiple separate computers or computing devices (e.g., physical servers, workstations, storage arrays, cloud computing resources, etc.) that communicate and interoperate over a network to perform the functions described herein. Any such computing device typically includes a processor (or multiple processors) that executes program instructions or program modules stored in memory or other non-transitory computer-readable storage media or non-transitory computer-readable storage devices (e.g., solid-state storage devices, disk drives, etc.). Various functions disclosed herein may be embodied in such program instructions or may be implemented in the form of application-specific circuitry (e.g., ASICs or FPGAs) in the computer system. When a computer system includes multiple computing devices, these devices may be located in the same location or in different locations. Results of the methods and tasks disclosed herein may be persistently stored by converting physical storage devices, such as solid-state memory chips or magnetic disks, into various forms. In some embodiments, the computer system may be a cloud-based computing system in which processing resources are shared by multiple separate business entities or other users.

[0234] The processes disclosed herein may be initiated in response to an event, such as upon request initiated by a user or system administrator, on a predefined schedule, a dynamically determined schedule, or in response to some other event. Once the processes are initiated, executable program instructions stored on one or more non-transitory computer-readable media (e.g., hard drives, flash memory, removable media, etc.) may be loaded into memory (e.g., RAM) of a server or other computing device. The executable instructions may then be executed by a hardware-based computer processor included in the computing device. In some embodiments, the processes disclosed herein, or portions thereof, may be implemented on multiple computing devices and / or multiple processors, either serially or in parallel.

[0235] Depending on the embodiment, certain acts, events or functions of any of the processes or algorithms described herein may be performed in a different order, may be added to one another, may be combined with one another, or may be omitted altogether (e.g., not all of the acts and events described herein are required to perform the algorithms described herein). Furthermore, in certain embodiments, multiple acts or events may be performed simultaneously rather than sequentially, for example, using multithreaded processing, interrupt processing, or multiple processors or multiple processor cores, or on other parallel structures.

[0236] The various exemplary logic blocks, modules, routines, and algorithmic steps described in connection with the embodiments disclosed herein may be implemented as electronic devices (e.g., ASIC or FPGA devices), computer software running on computer hardware, or a combination thereof. Additionally, the various exemplary logic blocks and modules described in connection with the embodiments disclosed herein may be implemented by devices such as processor devices, digital signal processors ("DSPs"), application specific integrated circuits ("ASICs"), field programmable gate arrays ("FPGAs") or other programmable logic devices, discrete gate logic or transfer logic, discrete hardware components, or combinations of these components designed to perform the functions described herein. The processor device may be a microprocessor, but in other aspects, the processor device may be a controller, a microcontroller, or a state machine, combinations thereof, and the like. The processor device may include electrical circuitry configured to process computer-executable instructions. In another embodiment, the processor device includes an FPGA or other programmable device that performs logical operations without processing computer-executable instructions. The processor unit may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a combination of multiple microprocessors, a combination of a DSP core and one or more microprocessors, or other such configurations. Although described herein primarily with respect to digital technology, the processor unit may include primarily analog components. For example, all or some of the rendering techniques described herein may be implemented using analog circuitry, or a combination of analog and digital circuitry.The computing environment can include any type of computer system, including, but not limited to, microprocessor-based computer systems, mainframe computers, digital signal processors, portable computing devices, device controllers, computer engines within appliances, to name a few.

[0237] The components of the methods, processes, routines or algorithms described in connection with the embodiments disclosed herein may be embodied directly in hardware, directly in software modules executed by a processor device, or directly in a combination of the hardware and software modules. The software modules may be included in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, a hard disk, a removable disk, a CD-ROM, or other forms of non-transitory computer-readable storage media. A typical storage medium may be coupled to the processor device such that the processor device can read information from and write information to such storage medium. In this aspect, the storage medium may be integral to the processor device. The processor device and the storage medium may be included in an ASIC. The ASIC may be included in a user terminal. In this aspect, the processor device and the storage medium may be present as discrete components of the user terminal.

[0238] As used herein, conditional terms such as "can," "may," "potential," "may," "for example," and the like, generally mean that a particular embodiment includes a particular feature, component, or step, and another embodiment does not include the feature, component, or step, unless otherwise indicated or interpreted otherwise in the context in which the terms are used. Thus, such conditional terms generally do not imply that one or more embodiments in any way require the feature, component, or step, or that one or more embodiments necessarily include decision logic, regardless of whether a particular feature, component, or step is included or performed in a particular embodiment, in the presence or absence of other inputs or promptings. Additionally, terms such as "including," "comprising," "having," and the like, are synonymous and are used in an inclusive sense in an open-ended manner, and do not exclude additional components, features, acts, operations, and the like. Additionally, the term "or" is also used in an inclusive sense (rather than an exclusive sense), e.g., when the term "or" is used to list components, it means one, some, or all of the listed components.

[0239] Unless otherwise indicated, it is understood that disjunctive language such as "at least one of X, Y, or Z" is generally used herein to indicate that a particular item, term, etc. may be X, Y, or Z, or any combination thereof (e.g., X, Y, or Z). Thus, such disjunctive language is generally not intended, and should not be intended, to imply that in a particular embodiment, the presence of at least one of X, at least one of Y, and at least one of Z is required.

[0240] While novel features applied in various embodiments have been shown, described and pointed out in the foregoing detailed description, it will be understood that the form and details of the foregoing apparatus or algorithms may be variously omitted, substituted and modified without departing from the scope of the present disclosure. Since some of the features described herein may be used or practiced separately from other features, it will be readily understood that certain embodiments described herein may be embodied in forms that do not provide all of the features and advantages described herein. All changes that come within the meaning and range of equivalency of the claims are encompassed within the scope of these claims.

Claims

1. A system for assessing or predicting wound healing, comprising: and one or more processors including an application executable by a mobile computing device, the mobile computing device in communication with at least one light detecting element. wherein the one or more processors are configured to perform the following operations according to processor-executable instructions included in the application, the operations comprising: receiving at least one signal from the light detecting element indicative of light of at least a first wavelength reflected from a tissue region including a wound or a portion thereof; generating an image having a plurality of pixels representative of the tissue region based on the at least one signal; measuring a reflected intensity value of each pixel at a first wavelength for at least a subset of the plurality of pixels based on the signal; determining one or more quantitative features of the subset of pixels based on the reflectance intensity values ​​of each pixel in the subset; and using one or more machine learning algorithms to generate at least one scalar value based on the one or more quantitative features of the subset of the plurality of pixels, the at least one scalar value corresponding to a predicted amount of healing of the wound or portion thereof after a predetermined period of time following generation of the image. system.

2. The system of claim 1 , wherein the operations further comprise determining the predicted amount of healing of the wound or portion thereof after a predetermined period of time.

3. The system of claim 1 , wherein the predicted amount of healing is a predicted percentage area reduction of the wound or portion thereof.

4. The system of claim 1 , wherein the predetermined period is 30 days.

5. 2. The system of claim 1, wherein the operations further include identifying at least one patient health measure corresponding to a patient having the tissue region, and the at least one scalar value is generated based on the one or more quantitative features of the subset of the plurality of pixels and the at least one patient health measure.

6. 6. The system of claim 5, wherein the at least one patient health measurement value includes at least one variable selected from patient demographic variables, compliance variables, endocrine system variables, cardiovascular system variables, musculoskeletal system variables, nutritional status variables, infectious disease variables, renal variables, obstetric-gynecological system variables, medication use variables, other disease variables, or clinical test values.

7. 6. The system of claim 5, wherein the at least one patient health measure comprises at least one characteristic selected from the group consisting of a patient's age, a patient's degree of chronic kidney disease, a length of the wound or portion thereof on the date the image was created, and a width of the wound or portion thereof on the date the image was created.

8. 2. The system of claim 1, wherein the first wavelength is within a range of 620 nm±20 nm, 660 nm±20 nm, or 420 nm±20 nm, and the one or more machine learning algorithms include a random forest ensemble.

9. 2. The system of claim 1, wherein the first wavelength is within a range of 726 nm±41 nm, 855 nm±30 nm, 525 nm±35 nm, 581 nm±20 nm, or 820 nm±20 nm, and the one or more machine learning algorithms include an ensemble of classifiers.

10. The operation further comprises automatically segmenting pixels in the image into wound pixels and non-wound pixels; and The system of claim 1 , further comprising selecting a subset of the plurality of pixels that includes pixels of the wound.

11. 11. The system of claim 10, wherein the plurality of pixels are automatically segmented using a segmentation algorithm comprising at least one of a U-Net comprising multiple convolutional layers and a SegNet comprising multiple convolutional layers.

12. 2. The system of claim 1, wherein the one or more quantitative features of the subset of the plurality of pixels are selected from the group consisting of a mean of the reflectance intensity values ​​of the pixels of the subset, a standard deviation of the reflectance intensity values ​​of the pixels of the subset, and a median of the reflectance intensity values ​​of the pixels of the subset.

13. The operation further comprises: applying a plurality of filter kernels individually to the image by convolution to produce a plurality of transformed images; constructing a three-dimensional matrix from the plurality of transformed images; and determining one or more quantitative characteristics of the three-dimensional matrix; The system of claim 1 , wherein the at least one scalar value is generated based on the one or more quantitative features of the subset of the plurality of pixels and based on the one or more quantitative features of the three-dimensional matrix.

14. 14. The system of claim 13, wherein the one or more quantitative features of the three-dimensional matrix are selected from the group consisting of a mean of the numerical values ​​of the three-dimensional matrix, a standard deviation of the numerical values ​​of the three-dimensional matrix, a median of the three-dimensional matrix, and a product of the mean and median of the three-dimensional matrix.

15. 15. The system of claim 14, wherein the at least one scalar value is generated based on a mean value of the reflectance intensity values ​​of the pixels of the subset, a standard deviation of the reflectance intensity values ​​of the pixels of the subset, a median value of the reflectance intensity values ​​of the pixels of the subset, a mean value of the numerical values ​​of the three-dimensional matrix, a standard deviation of the numerical values ​​of the three-dimensional matrix, and a median value of the three-dimensional matrix.

16. The operation further comprises: receiving a second signal from the at least one light detecting element indicative of light at a second wavelength reflected from the tissue region; measuring a reflected intensity value of each pixel at a second wavelength for at least a subset of the plurality of pixels based on the second signal; and determining one or more additional quantitative features of the subset of the plurality of pixels based on the reflected intensity value of each pixel at a second wavelength; The system of claim 1 , wherein the at least one scalar value is generated based at least in part on the one or more additional quantitative features of the subset of the plurality of pixels.

17. The system of claim 1 , wherein the at least one scalar value is generated on the same day as a first day after a predetermined period of time.

18. The operation further comprises: receiving an indication of an actual amount of healing of the wound, or portion thereof, after the predetermined period of time, the indication being determined based on measuring one or more dimensions of the wound, or portion thereof, after the predetermined period of time has elapsed since determining an expected amount of healing of the wound, or portion thereof; and 10. The system of claim 1, further comprising updating at least one of the one or more machine learning algorithms by providing images of at least the wound or portion thereof and the actual amount of healing as training data.

19. 10. The system of claim 1, wherein the operations further comprise selecting between a standard wound care therapy and an advanced wound care therapy based at least in part on a predicted amount of healing of the wound, or portion thereof, prior to the passage of the predetermined period of time.

20. The choice between standard wound care therapy and advanced wound care therapy is: When the predicted healing rate indicates that the wound or portion thereof will be more than 50% healed or closed after 30 days, prescribing one or more standard therapies selected from improving nutritional status, debridement to remove necrotic tissue, maintaining granulation tissue with a dressing, treatments required to resolve any infection that may have developed, addressing deficiencies in intravascular perfusion to the limb bearing the wound or portion thereof, reducing pressure load on the wound or portion thereof, or glycemic control; and and prescribing one or more advanced therapies selected from hyperbaric oxygen therapy, negative pressure wound therapy, bioengineered skin substitutes, synthetic growth factors, extracellular matrix proteins, matrix metalloproteinase regulators, and electrical stimulation therapy when the predicted healing amount indicates that the wound or portion thereof will not be more than 50% healed or closed after 30 days.

20. The system of claim 19.

21. The system of claim 1 , wherein the mobile computing device comprises a smartphone or a tablet.

22. 1. A method for assessing or predicting wound healing, comprising: Under the control of one or more processors of a mobile computing device; receiving at least one signal from a photodetector element of the mobile computing device indicative of light of at least a first wavelength reflected from a tissue region including the wound or a portion thereof; generating an image having a plurality of pixels representative of the tissue region based on the at least one signal; measuring a reflected intensity value of each pixel at the first wavelength for at least a subset of the plurality of pixels based on the signal; determining one or more quantitative features of the subset of pixels based on the reflectance intensity values ​​of each pixel in the subset; and using one or more machine learning algorithms to generate at least one scalar value based on the one or more quantitative features of the subset of the plurality of pixels, the at least one scalar value corresponding to a predicted amount of healing of the wound or portion thereof after a predetermined period of time following generation of the image. method.

23. 23. The method of claim 22, further comprising determining the predicted amount of healing of the wound or portion thereof after a predetermined period of time.

24. 23. The method of claim 22, wherein the healing amount is the predicted percent area reduction of the wound or portion thereof.

25. The method of claim 22 , wherein the mobile computing device comprises a smartphone or a tablet.

Images