Identification of Neurological States Using Image Analysis

Abstract

It is described to identify a patient's neurological state using image analysis. An image of patient tissue displays a plurality of nerves and a plurality of layers of the patient tissue. Using a trained model, each nerve among the plurality of nerves and each layer among the plurality of layers of the patient tissue in the image are labeled. Based on the labels assigned to the plurality of nerves and the plurality of layers of the patient tissue, the amount of nerve area relative to the tissue area is determined. Based on the amount of nerve area relative to the tissue area, the neurological state is determined.

Description

Technical Field

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 369,739, filed Jul. 28, 2022, the disclosure of which is incorporated herein by reference.

Background Art

[0002] Cancer treatments such as chemotherapy often damage a patient's nerves. In particular, chemotherapy can cause peripheral neuropathy. Peripheral neuropathy can be a debilitating side effect of cancer treatment that can occur when chemotherapeutic agents damage the peripheral nerves that are outside the brain and spinal cord. Symptoms of peripheral neuropathy can include pain, burning, tingling, numbness, electric shock, prickling, temperature sensitivity, etc. Peripheral neuropathy can start in the extremities (e.g., hands and feet) and move upward.

[0003] Diagnosing peripheral neuropathy can require determining the state of the nerves. Conventionally, medical professionals have used nerve conduction to determine the nerve state. Nerve conduction involves measuring the speed of conduction of electrical impulses through the nerves. In particular, one electrode stimulates the nerve with a gentle electrical impulse and another electrode located downstream records the result. This can be used to determine the nerve conduction velocity (NCV). NCV can be used to determine nerve damage and nerve destruction. However, this can be a costly, invasive, time-consuming, and error-prone process.

Summary of the Invention

[0004] Embodiments of systems, apparatuses, devices, methods, and / or computer program products for identifying neuropathy using image analysis, and / or combinations and sub-combinations thereof are provided herein.

[0005] A given embodiment includes a diagnostic method for determining a nerve condition. The method includes receiving an image of a patient's skin. The image shows a plurality of nerves and a plurality of layers of the patient's skin. The method further includes using a trained model to label each nerve among the plurality of nerves and each layer among the plurality of layers in the image of the patient's skin. The trained model is trained to identify differences between the nerves and the plurality of layers of the skin tissue. The method further includes determining an amount of nerve area relative to a layer of a skin tissue area based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin, and determining the nerve condition based on the amount of nerve area relative to the layer of the tissue area.

[0006] In some embodiments, the amount of nerve area relative to a tissue area (e.g., the area of the epidermis) corresponds to nerve density, and determining the nerve condition includes determining the presence of peripheral neuropathy when the nerve density is lower than a threshold amount.

[0007] In some embodiments, labeling each nerve among the plurality of nerves and each layer among the plurality of layers in the image of the patient's skin includes using the model to determine whether a pixel among the plurality of pixels in the input image corresponds to a nerve among the plurality of nerves or a respective layer among the plurality of layers of the patient's skin, and assigning to each pixel among the plurality of pixels a respective label indicating whether each respective pixel corresponds to a nerve or a respective layer among the plurality of layers of the patient's skin.

[0008] In some embodiments, the method further includes training a model. Training the model includes assigning training labels to each nerve in each training image of a plurality of training images and to each layer of a plurality of tissue layers. Training the model further includes inputting the plurality of training images into a convolutional neural network to label the nerves of the tissue and each of the plurality of layers in each training image of the plurality of training images, and comparing each respective label of the nerves of the tissue and each of the plurality of layers in each training image of the plurality of training images assigned by the convolutional neural network with the respective training label of the plurality of labels corresponding to the nerves and each layer. Assigning training labels to each nerve of the tissue and each of the plurality of layers in each image of the plurality of training images may include applying one or more filters to each image of the plurality of training images to identify the nerves in each image of the plurality of training images. Training the model may further include inputting a mirrored or rotated version of each training image of the plurality of training images into the convolutional neural network to label the nerves of the tissue and each of the plurality of layers in the mirrored or rotated version of each training image of the plurality of training images.

[0009] In some embodiments, a set of images of the plurality of images includes replicas of a single image, each replica including a contrast, brightness, crop, distortion, or rotation that varies from the single image.

[0010] In some embodiments, each image includes a z-stack image.

[0011] Another given embodiment includes a diagnostic system for determining a neurological state. The system includes a memory containing stored instructions and a processor coupled to the memory. The instructions, when executed by the processor, cause the processor to receive an image of a patient's skin. The image displays a plurality of nerves and a plurality of layers of the patient's skin. The instructions, when executed by the processor, further cause the processor to label each nerve of the plurality of nerves and each layer of the plurality of layers in the patient's skin in the image using a trained model. The trained model is trained to identify differences between the nerves and the plurality of layers of the skin tissue. The instructions, when executed by the processor, further cause the processor to determine the amount of nerve area relative to the layer area of the skin tissue region based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin, and to determine the neurological state based on the amount of nerve area relative to the layer area of the tissue region.

[0012] In some embodiments, the amount of nerve area relative to a tissue region (e.g., the epidermis) corresponds to nerve density, and determining the neurological state includes determining the presence of peripheral neuropathy when the nerve density is lower than a threshold amount.

[0013] In some embodiments, labeling each nerve of the plurality of nerves and each layer of the plurality of layers in the image of the patient's skin includes using the model to determine whether a pixel of the plurality of pixels in the input image corresponds to a nerve of the plurality of nerves or a respective layer of the plurality of layers of the patient's skin, and assigning a respective label indicating whether each respective pixel corresponds to a nerve or a respective layer of the plurality of layers of the patient's skin to each pixel of the plurality of pixels.

[0014] In some embodiments, when executed, the instructions cause the processor to further train the model. Training the model includes assigning training labels to each nerve in each training image of a plurality of training images and to each layer of a plurality of tissue layers. Training the model further includes inputting the plurality of training images into a convolutional neural network to label the nerves of the tissue and each layer of the plurality of layers in each training image of the plurality of training images, and comparing each respective label of the nerves of the tissue and each layer of the plurality of layers in each training image of the plurality of training images assigned by the convolutional neural network with the respective training labels of the plurality of labels corresponding to the nerves and each layer. Assigning training labels to each nerve of the tissue and each layer of the plurality of layers in each image of the plurality of training images may include applying one or more filters to each image of the plurality of training images to identify the nerves in each image of the plurality of training images. Training the model may further include inputting a mirrored or rotated version of each training image of the plurality of training images into the convolutional neural network to label the nerves of the tissue and each layer of the plurality of layers in the mirrored or rotated version of each training image of the plurality of training images.

[0015] In some embodiments, a set of images of the plurality of images includes replicas of a single image, each replica including a contrast, brightness, crop, distortion, or rotation that varies from the single image.

[0016] In some embodiments, each image includes a z-stack image.

[0017] Another given embodiment is a non-transitory machine-readable medium having instructions stored therein that, when executed by at least one computing device, cause the at least one computing device to perform operations including receiving an image of a patient's skin. The image displays a plurality of nerves and a plurality of layers of the patient's skin. The operations further include using a trained model to label each nerve among the plurality of nerves and each layer among the plurality of layers in the image of the patient's skin. The trained model is trained to identify differences between the nerves and the plurality of layers of the skin tissue. The operations further include determining an amount of nerve area relative to a layer of the skin tissue area based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin, and determining a nerve state based on the amount of nerve area relative to the layer of the tissue area.

[0018] In some embodiments, the amount of nerve area relative to a tissue area (e.g., the epidermis) corresponds to nerve density, and determining the nerve state includes determining the presence of peripheral neuropathy when the nerve density is lower than a threshold amount.

[0019] In some embodiments, labeling each nerve among the plurality of nerves and each layer among the plurality of layers in the image of the patient's skin includes using the model to determine whether a pixel among the plurality of pixels in the input image corresponds to a nerve among the plurality of nerves or to each respective layer among the plurality of layers of the patient's skin, and assigning to each pixel among the plurality of pixels a respective label indicating whether each respective pixel corresponds to a nerve or to each respective layer among the plurality of layers of the patient's skin.

[0020] In some embodiments, the operation further includes training a model. Training the model includes assigning training labels to each nerve in each training image among a plurality of training images and to each layer of a plurality of tissue layers. Training the model includes inputting the plurality of training images into a convolutional neural network to label the nerves of the tissue and each layer among the plurality of layers in each training image among the plurality of training images, and comparing each respective label of the nerves of the tissue and each layer among the plurality of layers in each training image among the plurality of training images assigned by the convolutional neural network with the respective training label corresponding to the nerve and each layer among the plurality of labels. Assigning training labels to each nerve of the tissue and each layer among the plurality of layers in each image among the plurality of training images may include applying one or more filters to each image among the plurality of training images to identify the nerves in each image among the plurality of training images. Training the model may further include inputting a mirrored or rotated version of each training image among the plurality of training images into the convolutional neural network to label the nerves of the tissue and each layer among the plurality of layers in the mirrored or rotated version of each training image among the plurality of training images.

[0021] In some embodiments, the set of images among the plurality of images includes replicas of a single image, each replica including a contrast, brightness, crop, distortion, or rotation that varies from the single image.

[0022] In some embodiments, each image includes a z-stack image.

Brief Description of the Drawings

[0023] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate the present disclosure and, together with the description, further serve to explain the principles of the present disclosure and to enable those skilled in the art to make and use the present disclosure.

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

[0024] The drawing in which an element first appears is typically indicated by one or more digits at the leftmost end of the corresponding reference number. In the drawings, like reference numbers may indicate identical or functionally similar elements.

DETAILED DESCRIPTION OF THE INVENTION

[0025] Systems, apparatuses, devices, methods, and / or computer program product embodiments for identifying a nerve state using image analysis, and / or combinations and sub - combinations thereof are provided herein.

[0026] As described above, conventional methods for discriminating nerve states can be costly, invasive, time-consuming, and error-prone. Specifically, manual evaluation of images showing tissue by a trained pathologist can be a time-consuming process. Human vision / interpretation is limited, may lack consistency, and can include significant bias. For example, if a patient is being treated for neuropathy using a given treatment, a pathologist may assume that the treatment is effective when interpreting an image of the patient's tissue. Additionally, different pathologists may use different methodologies to interpret the images and thus the interpretations may vary. Further, many pathologists consider only a subset of the nerves or nerve types in a given image. All of these can lead to inconsistent diagnoses and treatments.

[0027] The embodiments described herein solve these problems by using image analysis to discriminate a patient's nerve state. In a given embodiment, a processor receives an image of patient skin tissue. The image shows a plurality of nerves and a plurality of layers of the patient skin tissue. Using a trained model, the processor labels each nerve and each layer among the plurality of nerves and the plurality of layers of the patient skin tissue in the image. The trained model is trained to discriminate differences between the nerves and the plurality of layers of the skin tissue. Further, the processor determines an amount of nerve area relative to a layer of the skin tissue area based on the labels assigned to the plurality of nerves and the plurality of layers of the patient skin tissue. The processor determines the nerve state based on the amount of nerve area relative to a layer of the skin tissue area.

[0028] The embodiments described in this specification use a trained model configured to implement a neural network to determine a patient's neurological state. Specifically, the neural network implemented by the trained model provides fast, accurate, and consistent identification of the nerves and layers of the patient's skin tissue. This may involve determining the patient's nerve density or the number of nerve fiber crossings between the layers of the patient's tissue. The neurological state may be an indicator of peripheral neuropathy. Further, the neurological state may indicate whether drug treatment is inducing neuropathy or, alternatively, whether ongoing treatment for neuropathy is effective.

[0029] Figure 1 is a block diagram of a system for identifying a neurological state using image analysis. The system may include a server 100, a client device 110, and a database 120. The devices of the system may be connected via a network. For example, the devices of the system may be connected via a wired connection, a wireless connection, or a combination of a wired and a wireless connection. In an exemplary embodiment, one or more portions of the network may be an ad hoc network, an intranet, an extranet, a virtual private network (VPN), a local area network (LAN), a wireless LAN (WLAN), a wide area network (WAN), a wireless wide area network (WWAN), a metropolitan area network (MAN), a portion of the Internet, a portion of the public switched telephone network (PSTN), a cellular telephone network, a wireless network, a WiFi network, a WiMax network, any other type of network, or a combination of two or more such networks. Alternatively, the server 100, the client device 110, and the database 120 may be located on a single physical machine or a virtual machine.

[0030] In some embodiments, the server 100 and the database 120 may be present in a cloud computing environment. In other embodiments, the server 100 may be present within the cloud computing environment, while the database 120 may be present outside the cloud computing environment. Further, in other embodiments, the server 100 may be present outside the cloud computing environment, while the database 120 may be present within the cloud computing environment.

[0031] The server 100 may include an analysis engine 102 and a model 104. The analysis engine 102 may train the model 104 to label nerves and tissue layers in an image of skin tissue. The tissue layers of the skin may include the dermis, epidermis, and cornea. The model 104 may implement deep learning algorithms such as neural networks, convolutional neural networks (CNNs), artificial neural networks (ANNs), recurrent neural networks (RNNs), and deep reinforcement learning. Further, the analysis engine 102 may use the output of the model 104 to determine the nerve state of a given patient.

[0032] The database 120 can be one or more data storage devices configured to store structured data and unstructured data. The database 120 can store files for the training model 104. Further, the database 120 can store data associated with the layers of nerves and tissues labeled by the model 104. In one example, the label can be a mask overlaid on an image of the tissue. The data can be an image of the tissue including different masks for the nerve and tissue layers. The database 120 can also store data associated with the nerve state of a patient determined by the analysis engine 102. The data can be, for example, the nerve density in the epidermis. In particular, the nerve density can be determined based on the amount of nerve area / coverage rate over a tissue region identified in an image of the tissue. Alternatively or additionally, the data can be the number of nerve fiber crossings over multiple layers (e.g., from the dermis to the epidermis).

[0033] The client device 110 can be configured to interface with the server 100 and the database 120. For this purpose, the client device 110 can send a request to the server 100 to train the model 104. The request can include the location of a file for training the model 104 within the database 120. Alternatively, the client device 110 can search for the file and include this file in the request. The file can be an image of the tissue and can be received in one or more batches. As a non-limiting example, the image can be an image of the skin tissue of a mouse. In this example, the file includes two batches. The first batch includes 81 images of 12 mice and two treatment groups. The second batch includes 99 images of 19 mice in four treatment groups. The treatment groups in this embodiment indicate the current nerve state of each mouse.

[0034] Server 100 can receive requests from client device 110. Analysis engine 102 can search for files indicated in requests from database 120. As shown above, the files can be images of tissues (e.g., mouse skin tissue). The files can be in any of various formats such as PDF, JPEG, TIF, GIF.

[0035] To increase the amount of available training data, analysis engine 102 can generate multiple replicated versions of each image retrieved from database 120. For example, analysis engine 102 can generate rotated versions or mirrored versions of each image. Further, analysis engine 102 can generate one or more versions of each image with various contrast, brightness, crop, distortion, edge, shape, and / or texture filters.

[0036] Analysis engine 102 can identify and label the nerves and layers of the tissue within each image (including different versions of each image) and instruct model 104 to do so. Model 104 can implement a deep learning algorithm to classify and label each image pixel as part of a layer of tissue, nerve, or background. As a non-limiting example, the deep learning algorithm can be a CNN.

[0037] The CNN can be trained in two phases: a forward phase and a backward phase. The forward phase includes convolutional layers, pooling layers, and fully connected layers. Analysis engine 102 can instruct the CNN to classify and label each input image pixel as part of a layer, nerve, or background in order to train the CNN. The input images can be obtained from training files that include different versions of each image.

[0038] The convolutional layer can apply filters to the input image to generate a feature map. In particular, in the convolutional layer, the CNN can perform feature extraction on the input image. The features can include some parts of the input image. For example, the features can be different edges or shapes of the input image. The CNN can extract different types of features and generate different types of feature maps. For example, the CNN can apply an array of numbers (e.g., a kernel) across different parts of the input image. A kernel is sometimes called a filter. As shown above, different types of filters can be applied to the input image to generate different feature maps. For example, a filter for identifying shapes in the input image can be different from a filter for edge detection. Therefore, different kernels can be applied to identify shapes in the input image compared to edge detection. Each kernel can contain a different array of numbers. The values of the filter or kernel can be randomly assigned and optimized over time (e.g., using a gradient descent algorithm). The kernel can be applied as a sliding window across different parts of the input image. The kernel can be summed with a given part of the input image to generate an output value. The output value can be included in the feature map. The feature map can include output values from different kernels applied to each part of the input image. The generated feature map can be a two-dimensional array.

[0039] The pooling layer can generate a reduced feature map. In particular, in the pooling layer, the CNN can reduce the dimensionality of each feature map generated in the convolutional layer. The CNN can extract some parts of a given feature map and discard the rest. When an image is pooled, important features are retained. For example, a feature map can include activated regions and deactivated regions. The activated regions can contain detected features, and the deactivated regions can indicate that that part of the image did not contain features. Pooling can remove the deactivated regions. In this way, the size of the image is reduced. The CNN can use max pooling or average pooling in the pooling layer to perform these operations. Max pooling retains the higher values of some parts of the feature map while discarding the rest of the values. Average pooling retains the average value of different parts of the feature map. Thus, the CNN can generate a reduced feature map for each of the feature maps generated in the convolutional layer.

[0040] The CNN can include additional convolutional layers. In the additional convolutional layers, the CNN can generate additional feature maps based on the reduced feature maps generated in the pooling layer. Further, the CNN can include additional pooling layers. In the additional pooling layers, the CNN can generate further reduced feature maps based on the feature maps generated in the additional convolutional layers.

[0041] The convolutional layer can also apply the Rectified Linear Unit (ReLU) function to the input image. The ReLU function is applied to the input image to remove linearity from the input image. For example, the ReLU function can remove all black elements from the input image and retain only gray and white. This causes the colors in the input image to change more abruptly and removes linearity from the input image.

[0042] Convolutional and pooling layers can be used for feature learning. Feature learning enables the CNN to identify the desired features in the input image and thus allows the input image to be accurately classified. Therefore, by optimizing the convolutional and pooling layers, the CNN can apply the correct filters to the input image to extract the necessary features required for classifying the input image.

[0043] Next, the fully connected layer can classify the features of the image using weights and biases to generate an output. The CNN can classify and label each pixel of the input image as part of a layer of tissue, nerve, or background. The CNN can output the input image with labeled layers, nerves, and background. The label can be a mask applied to the image indicating whether each pixel is part of a particular layer, nerve, or tissue. The mask is described in more detail with respect to FIG. 5.

[0044] Specifically, in the fully connected layer, the CNN can flatten the reduced feature map generated in the pooling layer into a one-dimensional array (or vector). The fully connected layer is a neural network. The CNN can perform a linear transformation on the one-dimensional array. The CNN can perform the linear transformation by applying weights and biases to the one-dimensional array. First, the weights and biases are randomly initialized and can be optimized over time. The CNN can perform a non-linear transformation, such as an activation layer function (e.g., softmax or sigmoid), to classify and label each input image pixel as part of a layer of tissue, nerve, or background.

[0045] In the reverse phase, the CNN can use backpropagation to determine whether the CNN was able to correctly label each pixel of the input image. Backpropagation involves optimizing the input parameters so that the CNN can classify the document more accurately. The input parameters can include values such as kernels, weights, biases, etc. Gradient descent can be used to optimize the parameters. In particular, gradient descent can be used to optimize the classification and labeling of the CNN for each pixel in the input image.

[0046] Gradient descent is an iterative process for optimizing a CNN. Gradient descent updates the parameters of the CNN, causing the CNN to classify and label each pixel in the input image based on the updated parameters, and to verify the classification and labeling of each pixel in the input image.

[0047] For this purpose, the CNN can use backpropagation to verify the classification of the pixels in the input image. In particular, an expert in a particular field (e.g., a pathologist) can determine whether the model 104 has correctly labeled the layers, nerves, and background. The expert in the particular field can provide feedback regarding the accuracy of the labeling from the client device 110 to the analysis engine 102. Alternatively or additionally, the analysis engine 102 can use the metadata associated with each image to verify the labeling of the layers, nerves, and background. For example, the metadata can include pre-assigned labels assigned to the layers, nerves, and background of each image. A pathologist or a different third party can generate the pre-assigned labels for each image. The pre-assigned labels can be masks assigned to different layers, nerves, and backgrounds.

[0048] The CNN can compare the classification and labeling assigned to each pixel in the input image by the CNN with the information contained in the metadata or feedback provided by an expert in a particular field. Based on the comparison, the CNN can use the gradient descent algorithm to update the values of the filters, weights, and biases and re-run the forward pass on the input image.

[0049] The analysis engine 102 can instruct the CNN to label and classify each pixel in each image within a training file that includes different versions of each image. The CNN iteratively optimizes its parameters, re-classifies and re-labels each pixel in each image within the training file via the forward pass, and verifies the classification and labeling of each pixel via the backward pass until a desired accuracy threshold is reached.

[0050] The CNN can be considered to be sufficiently trained when it reaches the desired accuracy threshold. For this purpose, the model 104 can be considered to be sufficiently trained.

[0051] The client device 110 may send a request to the server 100 asking to classify and label new images of the patient's tissue. The request may be to determine the patient's neurological state based on the images of the patient's tissue. The request may include new images. Alternatively, the request may include the location of the new images in the database 120. The new images may not be part of the images used to train the model 104.

[0052] The server 100 may receive the request. If the request includes the location of the new images in the database 120, the analysis engine 102 may retrieve the new images from the database 120. The analysis engine 102 may instruct the model 104 to classify and label each pixel of the new image as part of a particular layer, nerve, or background. The model 104 may implement a deep learning algorithm (e.g., CNN) to classify and label each pixel of the new image. The model 104 may output an image including labels for different layers, nerves, and backgrounds of the tissue in the image.

[0053] The analysis engine 102 can determine the nerve density within the epidermis using the markings. The nerve density can indicate nerve damage. In this regard, the analysis engine 102 can determine the patient's nerve density based on the amount / ratio of the area of the patient's nerves to the total area of the patient's epidermis based on the image output by the model 104. Specifically, the analysis engine 102 can identify the nerves labeled within the image output by the model 104 and determine what portion of the tissue area is covered by the nerves. The analysis engine 102 can generate a determination value regarding the patient's nerve condition based on the patient's nerve density. For example, the analysis engine 102 can diagnose a patient with chemotherapy-induced peripheral neuropathy based on the nerve density being lower than a threshold amount. Alternatively, the analysis engine 102 can diagnose a patient with chemotherapy-induced peripheral neuropathy based on the nerve density being lower (e.g., by a threshold or ratio) than a second (potentially earlier or pre-treatment) determination value of the nerve density by the analysis engine 102. Additionally, the analysis engine 102 can determine the effectiveness of treatment for peripheral neuropathy by comparing a first (e.g., pre-treatment) determination value and a second (e.g., post-treatment) determination value of the nerve density by the analysis engine 102. Further, a pathologist can determine peripheral neuropathy through manual analysis or comparison of the nerve density determination values by the analysis engine 102.

[0054] Alternatively or additionally, the analysis engine 102 may also use the labels in the images output by the model 104 to determine the number of nerve fiber crossings from the dermis layer to the epidermis layer. A low number of nerve fiber crossings from the dermis layer and the epidermis layer may indicate nerve damage. Thus, the analysis engine 102 may identify the epidermis layer in the patient's tissue labeled in the image output by the model 104, and the dermis layer in the patient's tissue labeled in the image output by the model 104. The analysis engine 102 may generate a boundary between the epidermis layer and the dermis layer in the image output by the model 104. The analysis engine 102 may determine the number of nerve fiber crossings based on identifying a grouping of pixels labeled as nerves extending from the dermis layer to the epidermis layer. This determination value of the nerve fiber crossings by the analysis engine 102 may be closely correlated with the counting of nerve fiber crossings according to the IENF counting rule. The IENF is described in relation to FIG. 13.

[0055] Based on the number of nerve fiber crossings, the analysis engine 102 may generate a determination regarding the patient's nerve condition. For example, the analysis engine 102 may diagnose a patient with chemotherapy-induced peripheral neuropathy based on the number of nerve fiber crossings being lower than a threshold amount. In one example, the number of nerve crossings may include a normalized number of nerve crossings determined by dividing the determined number of nerve crossings by a specific pixel length of the epidermis / dermis boundary. Alternatively, the analysis engine 102 may determine whether ongoing treatment for the patient's diagnosed neuropathy is effective, or whether a drug has induced the progression of neuropathy.

[0056] FIG. 2 is a block diagram illustrating the preprocessing of an input image before being transmitted to the model 104 according to some embodiments. FIG. 2 will be described with reference to FIG. 1.

[0057] In some embodiments, the input image can be an image of skin tissue. The input image can be part of a training file or a new image of a patient's tissue. The image of the skin tissue can be obtained using a skin biopsy. Images of the tissue can be captured at different focal distances. The focal plane is the distance between the lens and the focus of the camera that captures the image.

[0058] The analysis engine 102 can perform preprocessing on the input image before providing the input image to the model 104 for classification and labeling. The analysis engine 102 can receive or retrieve images 200 and 202 of the same tissue at different focal planes. The analysis engine 102 can compile a z-stack image 204 by combining images 200 and 202. For this purpose, the z-stack image 204 can be a composite image having a greater depth than images 200 and 202. The z-stack image 204 can be the input image.

[0059] As illustrated by element 206, two to n images may be captured at various focal planes. The analysis engine 102 can combine the two to n images to generate a z-stack image 208. The z-stack image 208 can be a composite of the two to n images.

[0060] FIG. 3 is a block diagram illustrating the generation of various versions of an image according to some embodiments. FIG. 3 will be described with reference to FIG. 1.

[0061] In some embodiments, the analysis engine 102 may generate various versions of each image in the training file by varying the contrast, brightness, crop, distortion, rotation, etc. of each image. For example, image 300 may be included in the training file. The analysis engine 102 may vary one or more of the contrast, brightness, crop, distortion, rotation, etc. of image 300 to generate image 302. Similarly, image 304 may be included in the training file. The analysis engine 102 may vary one or more of the contrast, brightness, crop, distortion, rotation, etc. of image 304 to generate image 306.

[0062] Images 302 and 306 may be included in the training file along with images 300 and 304. This allows for an increase in the number of training files and thus improves the model 104. Accordingly, each of images 300 - 306 may be used to train the model 104 as described above with respect to FIG. 1. This enables the model 104 to be trained to label and classify each image pixel for various contrast, brightness, crop, distortion, rotation, etc.

[0063] FIG. 4 is a block diagram illustrating the generation of various versions of an image according to some embodiments. FIG. 4 will be described with reference to FIG. 1.

[0064] In some embodiments, the analysis engine 102 may generate variant versions of each image in the training file by rotating and mirroring each initial image. For example, image 400 may be included as an initial image in the training file. The analysis engine 102 may flip image 400 to generate image 402. The analysis engine 102 may mirror image 400 to generate image 404. The analysis engine 102 may mirror image 402 to generate image 406. The analysis engine 102 may rotate image 404 to generate image 408. The analysis engine 102 may flip image 408 to generate image 410. The analysis engine 102 may mirror image 408 to generate image 412. The analysis engine 102 may mirror image 410 to generate image 414.

[0065] Images 400 - 414 may be included in the training file. This makes it possible to increase the number of training files. Thus, images 400 - 414 may be used to train the model 104 as described above with respect to FIG. 1. This enables the model 104 to be trained to label and classify each pixel of each image regardless of the position of the tissue layers, nerves, and background in the image.

[0066] FIG. 5 shows an exemplary image illustrating pre - assigned labels according to some embodiments. FIG. 5 will be described with reference to FIG. 1.

[0067] In some embodiments, a domain expert may use a software application to pre - assign labels to the layers, nerves, and background in an image of a tissue. The domain expert may pre - assign labels for each of the images in the training file. The pre - assigned labels may be part of the metadata of the training file used to verify the labeling and classification of the model 104.

[0068] For example, the image 500 can be an image of skin tissue within a training file. A domain expert can use a software application to assign a first label to the dermis layer 504 of the tissue in the image, a second label to the epidermis layer 506 of the tissue in the image, a third label to the corneal layer 508 of the tissue in the image, a fourth label to the nerve 510 of the tissue in the image, and a fifth label to the background 514 of the image.

[0069] The image 502 includes labels. The labels can be masks overlaid on the layers, nerves, and background. The first label can be a blue mask, the second label can be a purple mask, the third label can be a yellow mask, the fourth label can be a green mask, and the fifth label can be a white mask. Thus, the blue mask can be overlaid on the pixels corresponding to the dermis layer, the purple mask can be overlaid on the pixels corresponding to the epidermis layer, the yellow mask can be overlaid on the pixels corresponding to the corneal layer, the green mask can be overlaid on the pixels corresponding to the nerve, and the white mask can be overlaid on the pixels corresponding to the background. However, this is one embodiment, and different colors can be used for the various masks as needed.

[0070] FIG. 6 illustrates the identification of nerves in an image of tissue, according to some embodiments.

[0071] In some embodiments, pre-assigning labels to the images within the training file can include identifying nerves in the images of the tissue. Specifically, the application can be executed by the client device 110. Alternatively, the application can reside on the server 100 and the client device 110 can access the application. The application can be used to pre-assign the labels within the training file.

[0072] The application can first automatically detect nerves within the image 600 based on the brightness and contrast of the image 600. For example, the nerves within the image 600 can be darker than the background. Thus, the application can detect darker regions of the image 600 based on the brightness and contrast of the pixels within the image 600. The darker pixels can be labeled as nerves. Such detection can be based on a pre-trained deep learning model trained on previously labeled images, or on a given pixel having a brightness and / or contrast value that is a threshold percentage different from at least one adjacent pixel or subset of adjacent pixels, or having a brightness value that is above or below a given threshold.

[0073] For example, each pixel in the image can be assigned a value corresponding to one or a combination of intensity, brightness, or contrast. Each pixel having a value above a specific threshold (or conversely, below the threshold) can be identified as a pixel corresponding to a nerve. Each pixel having a value below a specific threshold (or conversely, above it) can be identified as a pixel corresponding to tissue. The threshold can be a predetermined absolute threshold. Alternatively, the threshold can be based on the average or mean value across all pixels in the image. In some aspects, background pixels or other pixels outside the boundary of the tissue sample can be assigned a "null" value.

[0074] Once values are assigned to all pixels in the image, the nerve density can be determined by dividing the number of pixels corresponding to nerves by the number of pixels corresponding to tissue. Alternatively, the ratio of the number of nerve pixels or tissue pixels to the total number of pixels in the image can be used.

[0075] In some embodiments, nerves can be identified by applying one or more filters to an image of tissue. For example, a nerve can be identified in image 600 by applying one or more filters corresponding to one or more parameters. The parameters can include luminance, contrast, size, color, edges, shape, and texture filters. A pre - assigned label (e.g., a fourth label or a green mask) can be assigned to the identified nerve and this label can then be used for further analysis. In some embodiments, the filters are operable to be optimized for a given image.

[0076] FIG. 7 is a block diagram showing a neural network according to some embodiments. FIG. 7 will be described with reference to FIG. 1.

[0077] As described above, model 104 can implement a deep - learning algorithm. The deep - learning algorithm can be a neural network (e.g., a convolutional neural network) 700. The neural network 700 can include an input layer, a plurality of hidden layers, and an output layer.

[0078] If the neural network 700 is a CNN, the input layer can include a matrix of pixels of the input image. The input image can be part of a training file or a new image of a patient's tissue. The matrix of pixels of the input image can be sent to hidden layer 1.

[0079] One hidden layer can be a convolutional layer. As described above, the convolutional layer can apply a filter to a matrix of pixels of the input image to generate a feature map. The feature map can be sent to a further hidden layer that can be a pooling layer. As described above, the pooling layer can generate a downsampled feature map. The neural network 700 can include additional hidden layers (e.g., additional convolutional layers or pooling layers) to generate further feature maps and downsampled feature maps. Although two hidden layers are explicitly shown in FIG. 7, it is understood that the neural network 700 can include any number of hidden layers (such as 2, 3, 4,... n hidden layers), and each of the hidden layers can have various shapes and / or sizes.

[0080] The downsampled feature map can be sent to the output layer. The output layer can be a fully connected layer. As described above, in the fully connected layer, the CNN can flatten the downsampled feature map generated by the pooling layer into a one-dimensional array (or vector). The fully connected layer can use weights and biases to classify the features of the image and generate an output. The output can be the classified and labeled input image. That is, the output can be an image that includes the label or mask predicted / classified by the neural network 700.

[0081] When the model 104 is trained, when an output is generated, the neural network 700 can verify the classified and labeled input image based on the metadata of the input image. The metadata can include pre-assigned labels. The neural network 700 can identify any errors in the classification and label assigned to the input image, attempt to improve the parameters (e.g., biases and weights), and re-classify and label the input image. The neural network 700 can repeatedly classify and label each image in the training file.

[0082] FIG. 8 is a diagram of a feature map generated by a neural network according to some embodiments. FIG. 8 will be described with reference to FIGS. 1 and 7.

[0083] In some embodiments, the hidden layer of neural network 700 may generate a feature map 804 of a matrix of pixels 800 representing the input image. For example, an array of digits 802 may be applied across different portions of the matrix of pixels 800 to generate the feature map 804.

[0084] The neural network 700 may reduce the feature map. Additionally, the neural network 700 may generate additional feature maps based on the reduced feature map, and may further reduce the additional feature maps until a fully connected layer is formed. As described above, the fully connected layer may generate an output of the classified and labeled input image.

[0085] As a non-limiting example, the input image 806 may be transformed into a matrix of pixels, a feature map, and a reduced feature map, as shown by transformation 808. The fully connected layer may output a classified and labeled image 810.

[0086] Figs. 9-12 illustrate images of tissue labeled by a pathologist and model 104 according to some embodiments. Specifically, Figs. 9-12 show the ability of model 104 to automatically label images that would otherwise require significant effort and time by a pathologist (e.g., in the case of nerve fiber crossings) or that are virtually impossible for a pathologist (e.g., in the case of nerve fiber density). Additionally, model 104 can improve the accuracy and consistency of the labels provided by model 104 as compared to manually labeling the images by one or more pathologists.

[0087] For example, image 900 may be an image of a patient's tissue. The patient's tissue may include a dermis layer, an epidermis layer, a corneal layer, and nerves. Image 900 may also include a background. A pathologist may use software to label image 900 and generate image 902. Image 902 may have labels assigned to the dermis layer, epidermis layer, corneal layer, nerves, and background.

[0088] Furthermore, the image 900 can be labeled by the model 104. The model 104 can label each pixel of the image 900 as part of the dermis layer, epidermis layer, corneal layer, nerve, or background to generate the image 904. As shown by element 906, the model 104 can automatically perform the labeling function that has been conventionally performed by a pathologist, and can better identify the size and shape of the corneal layer with higher accuracy than a pathologist.

[0089] Regarding FIG. 10, the image 1000 can be an image of a patient's tissue. The patient's tissue can include a dermis layer, an epidermis layer, a corneal layer, and nerves. The image 1000 can also include a background. A pathologist can use software to label the image 1000 to generate the image 1002. The image 1002 can have labels assigned to the dermis layer, epidermis layer, corneal layer, nerves, and background.

[0090] Furthermore, the image 1000 can be labeled by the model 104. The model 104 can label each pixel of the image 1000 as part of the dermis layer, epidermis layer, corneal layer, nerve, or background to generate the image 1004. The model 104 was able to identify the size and shape of the corneal layer with higher accuracy than a pathologist.

[0091] Regarding FIG. 11, the image 1100 may be an image of a patient's tissue. The patient's tissue can include a dermis layer, an epidermis layer, a corneal layer, and nerves. The image 1100 may also include a background. A pathologist can use software to label the image 1100 to generate the image 1102. The image 1102 can have labels assigned to the dermis layer, epidermis layer, corneal layer, nerves, and background.

[0092] Furthermore, the image 1100 can be labeled by the model 104. The model 104 can label each pixel of the image 1100 as part of the dermis layer, epidermis layer, corneal layer, nerve, or background to generate the image 1104. The model 104 was able to identify the size and shape of the dermis layer with higher accuracy than a pathologist.

[0093] With respect to FIG. 12, the image 1200 may be an image of a patient's tissue. The patient's tissue may include a dermis layer, an epidermis layer, a corneal layer, and nerves. The image 1200 may also include a background. A pathologist may use software to label the image 1200 to generate the image 1202. The image 1202 may have labels assigned to the dermis layer, epidermis layer, corneal layer, nerves, and background.

[0094] Furthermore, the image 1200 may be labeled by the model 104. The model 104 may label each pixel of the image 1200 as part of the dermis layer, epidermis layer, corneal layer, nerves, or background to generate the image 1204. The model 104 was able to identify the sizes and shapes of the epidermis and dermis layers with higher accuracy than the pathologist.

[0095] FIG. 13 illustrates an image of skin tissue showing nerve fiber crossings from the dermis layer to the epidermis layer. The number of nerve fiber crossings is determined using the IENF counting rule in FIG. 13.

[0096] The IENF counting rule is a standard counting rule used to count nerve fibers. The IENF counting rule counts the nerves that cross the dermis layer and reach the epidermis layer. The total number of IENF counts can be normalized by dividing the IENF counts by the length of the boundary between the dermis layer and the epidermis layer. The normalized IENF counts can be used to determine the nerve state.

[0097] The image 1300 shows an image of skin tissue including an epidermis layer, a dermis layer, and nerves. The image 1302 shows a diagram of the nerves of the skin tissue shown in the image 1300 and the boundary (BM) between the epidermis and the dermis layer. The image 1300 includes nerve fibers 1304-1310.

[0098] In Image 1302, the nerves in the dermis layer are labeled as nerve fibers a - i. Based on the IENF counting rule, when a nerve fiber originates from the dermis layer and crosses into the epidermis layer, that nerve fiber is counted as a nerve fiber intersection. If a nerve fiber branches into multiple different fibers in the dermis layer or at the boundary (BM) before crossing into the epidermis layer, that nerve fiber is counted as two nerve fiber intersections. If a nerve fiber branches into multiple nerve fibers in the epidermis layer after the boundary (BM), it is counted as a single nerve fiber intersection. If a nerve fiber in the epidermis layer has a severed part from the original nerve fiber in the dermis layer, that nerve fiber is not counted as a nerve fiber intersection. If a nerve fiber does not cross into the epidermis layer, that nerve fiber is not counted as a nerve fiber intersection.

[0099] In Image 1302, after entering the epidermis layer, nerve fiber a branches into two nerve fibers. Therefore, nerve fiber a is counted as one nerve fiber intersection. Nerve fiber b branches into two nerve fibers after the boundary (BM) and in the epidermis layer. Therefore, nerve fiber b is counted as one nerve fiber intersection. Nerve fiber c branches into two nerve fibers before crossing into the epidermis layer (on the boundary (BM)). As a result, nerve fiber c is counted as two nerve fiber intersections. Nerve fiber d branches into two nerve fibers before crossing into the epidermis layer (dermis layer). Therefore, nerve fiber d is counted as two nerve fiber intersections. Nerve fiber e is a single nerve fiber that crosses into the epidermis layer from the dermis layer. Therefore, nerve fiber e is counted as one nerve fiber intersection. Nerve fiber f does not cross into the epidermis layer. As a result, nerve fiber f is not counted as a nerve fiber intersection. Nerve fiber g is a single nerve fiber that crosses into the epidermis layer from the dermis layer. Therefore, nerve fiber g is counted as one nerve fiber intersection. Nerve fiber h branches into two nerve fibers before crossing into the epidermis layer (on the boundary (BM)). Therefore, nerve fiber h is counted as two nerve fiber intersections. The branches of nerve fiber i are severed before crossing into the epidermis layer. Therefore, nerve fiber i is not counted as a nerve fiber intersection. Based on the IENF counting rule, Image 1302 shows 10 nerve fiber intersections.

[0100] Figure 14 shows an image labeled by model 104 for determining nerve density and nerve fiber crossing according to some embodiments. Figure 14 will be described with reference to Figure 1.

[0101] Model 104 may label each pixel of an image of a patient's tissue to generate an image such as image 1400. Analysis engine 102 may determine nerve density using the labels. For example, analysis engine 102 may determine the number of pixels in image 1400 labeled as nerves. Analysis engine 102 may further determine the number of pixels in image 1400 labeled as tissue type. Analysis engine 102 may ignore pixels labeled as belonging to the background.

[0102] Analysis engine 102 can calculate nerve density based on the ratio between the number of pixels labeled as nerves in the epidermis and the number of pixels labeled as tissue in the epidermis. For example, analysis engine 102 may divide the number of pixels labeled as nerves by the number of pixels labeled as tissue to determine nerve density. In the specific example of Figure 14, analysis engine 102 can determine that 8.76% of the tissue region determined by analysis engine 102 to include the epidermis is covered by nerves in image 1400. Analysis engine 102 may use nerve density to determine the patient's nerve condition. The lower the nerve density, the higher the likelihood that the patient is diagnosed with peripheral neuropathy.

[0103] Additionally or alternatively, as illustrated in Image 1402, the analysis engine 102 may evaluate the likelihood of peripheral neuropathy by counting the nerve fiber crossings from the dermis to the epidermis. For example, the analysis engine 102 may define a boundary 1404 between the epidermis and the dermis layer based on Image 1402 as labeled by the model 104. The analysis engine 102 can identify a group of pixels labeled as nerves that cross the boundary 1404 between the epidermis and the dermis layer. The number of identified crossings can be normalized by dividing the number of identified crossings over a particular boundary length by the number of pixels at that particular boundary length.

[0104] As a non-limiting example, the analysis engine 102 can determine that the length of the crossing boundary 1404 contains 4,401 pixels. In Figure 14, although represented by an orange region for visibility, the crossing boundary 1404 includes a line with a single pixel width. Further, the analysis engine 102 can determine that Image 1402 may contain 36 crossings over the length of the crossing boundary 1404. The number of crossings can be normalized by dividing the number of crossings (e.g., 36) by the number of pixels at the length of the crossing boundary 1404 (in this example, 4,401). The fewer the number of nerve fiber crossings, the higher the likelihood that the patient will be diagnosed with peripheral neuropathy.

[0105] FIG. 15 includes a chart showing exemplary test data associated with the epidermal nerve density determined for mice undergoing treatment with paclitaxel, according to some embodiments, using the analysis engine 102 and the model 104. Specifically, chart 1500 shows the differences in epidermal nerve density among mice undergoing various treatments, determined using the analysis engine 102 and the model 104. Such information can be used, for example, by the analysis engine 102 to determine the presence or progression of peripheral neuropathy resulting from a treatment plan such as cancer treatment. This information can also be used, for example, by the analysis engine 102 to determine the effectiveness of various neuropathy treatments (such as local injection of corticosteroids, lidocaine, or botulinum toxin) using the data reflected in chart 1500. The example of referring to FIG. 15 is particularly related to paclitaxel, but the nerve analysis system and method of the present application, particularly the system described with respect to FIG. 1, can be used to measure the effects of various drugs that may induce neuropathy, such as drugs in chemotherapy, heart and blood pressure treatment (such as amiodarone), drugs for infectious diseases (such as chloroquine), drugs for autoimmune disease treatment (such as infliximab), anti-epileptic drugs (such as phenytoin), etc.

[0106] In this exemplary test, the analysis engine 102 used images of the mouse tissue labeled by the model 104 to determine the nerve density of mice administered a neuropathy-inducing drug. The mice were grouped based on the treatment administered to each mouse. The treatments included saline i.v. for the control group and a neuropathy-inducing drug, in this case paclitaxel (PCTX) 25 mg / kg treatment, for the test group. PCTX 25 mg / kg represents the treatment where paclitaxel was intravenously administered to the mice. Paclitaxel is an anti-tumor drug commonly used to treat cancer and may induce neuropathy as a side effect. As shown, two separate batches of the control group and the PCTX group were analyzed.

[0107] In this exemplary test, the analysis engine 102 determined the nerve density by dividing the number of nerve pixels in the image of the mouse tissue by the number of pixels in the epidermal layer pixels.

[0108] Chart 1500 shows the nerve area over the epidermal area (nerve density) of the mice being treated for neuropathy, as determined by the analysis engine using images of mouse tissue labeled by model 104. In these two batches, the mice were grouped based on the treatment administered to each mouse. The treatments included a control and PCTX 25 mg / kg treatment. In this exemplary test, the analysis engine 102 determined the nerve density by dividing the number of nerve pixels in the image of the mouse tissue by the number of pixels in the epidermal layer pixels.

[0109] Chart 1500 includes two batches, Batch 1 and Batch 2. Each batch includes its respective control and PCTX 25 mg / kg IV treatment.

[0110] Chart 1502 shows the nerve density of the mice being treated for neuropathy in Batch 1 and Batch 2, as determined by an automated system such as the analysis engine 102 described above. Chart 1502 reflects the data of Chart 1500 in bar graph form. Comparing Charts 1500 and 1502 shows that an automated system trained to determine the nerve fiber ratio can adequately distinguish the effectiveness of different treatments even when the differences between them are small. This is because, as a result of the automated analysis, the values are distributed in a concentrated manner, that is, the within-group variance is small.

[0111] Chart 1504 shows the epidermal nerve density of the mice in Batch 1 as determined by a pathologist. As can be seen by comparison, the comparative output of the automated system shown in Chart 1502 is confirmed by the manual pathologist evaluation shown in Chart 1504. The comparison results between the control and paclitaxel for Batch 1 are similar between the two charts, but the manual evaluation in Chart 1504 reports a higher nerve fiber ratio over a larger distribution than the automated system in Chart 1502. This could be caused by pathologist bias. Nevertheless, the comparison supports the result that the automated system can accurately identify and distinguish treatment outcomes with lower within-group variability compared to manual evaluation.

[0112] Figure 16 shows a significant correlation between nerve conduction and determination of nerve fiber density by the test model shown above. Nerve response amplitude and latency are conventional methods for detecting nerve damage. For example, stimulation and recording electrodes may be applied to the skin, and nerve response amplitude and latency are determined after calibrated electrical stimulation. Accordingly, to further validate the methods and systems for determining nerve fiber density described herein, specific members of both the control group and the PCTX group for Batches 1 and 2 referenced with respect to Figure 15 were analyzed for nerve response amplitude and latency. Chart 1600 shows a high positive correlation between nerve response amplitude and nerve fiber density for both control and PCTX treatment measured in accordance with the present disclosure, whereas Chart 1602 shows a high negative correlation between nerve response latency and nerve fiber density for both control and PCTX treatment measured in accordance with the present disclosure, thus supporting the veracity of the nerve disorder detection method described herein.

[0113] FIG. 17 is a flowchart illustrating an exemplary process for determining a neurological state using image analysis according to one embodiment. Method 1700 may be implemented by processing logic that may include hardware (e.g., circuitry, dedicated logic, programmable logic, microcode, etc.), software (e.g., instructions executed on a processing device), or a combination thereof. It should be understood that not all things are required to implement the disclosure provided herein. Further, as will be understood by those skilled in the art, some of the steps may be performed simultaneously or in an order different from that shown in FIG. 17.

[0114] Method 1700 is to be described with reference to FIG. 1. However, method 1700 is not limited to its exemplary embodiments.

[0115] In operation 1702, an image of patient tissue is received. For example, server 100 may receive an image of patient tissue. The image displays the nerves and layers of the patient tissue. The patient tissue may be skin tissue. The image may be received with a request to determine the neurological state of the patient's tissue.

[0116] In operation 1704, each nerve and layer of the patient tissue in the image is labeled. For example, server 100 may use model 104 to label each nerve and each layer of the patient tissue in the image. Model 104 is trained to identify differences between nerves and tissue. Specifically, model 104 is trained to label each pixel to correspond to a given tissue layer, nerve, or image background.

[0117] In operation 1706, the amount of neural area relative to the tissue area is determined based on the labels. For example, server 100 may use analysis engine 102 to determine the amount of neural area relative to the layers of the tissue area (e.g., epidermis) based on the labels assigned to the nerves and layers of the patient tissue in the image. Analysis engine 102 can separate the pixels associated with the nerves based on the labeled nerves in the image. Further, analysis engine 102 can identify the total number of pixels labeled to correspond to the layers of the tissue. That is, analysis engine 102 can exclude the pixels labeled as background. Analysis engine 102 can determine the number of pixels covering the tissue area.

[0118] In operation 1708, the neural state is determined. For example, server 100 may use analysis engine 102 to determine the neural state based on the amount of neural area relative to the area of the tissue area, e.g., epidermis. The neural area relative to the tissue area may be referred to as neural density. In some embodiments, if the neural density is lower than a threshold amount, it may indicate peripheral neuropathy. Alternatively, if the neural density is lower than a previously determined neural density, it may indicate peripheral neuropathy. Additionally, analysis engine 102 can use the amount of neural area relative to the tissue area to determine the effectiveness of ongoing treatment for a patient already diagnosed with neuropathy.

[0119] In some embodiments, analysis engine 102 can generate the boundary between the epidermis and the dermis layer in the image based on the labels assigned by model 104. Analysis engine 102 can determine the number of nerve fibers crossing the boundary. If the number of nerve fiber crossings is less than a threshold amount, it may indicate peripheral neuropathy. Additionally, analysis engine 102 can use the number of nerve fiber crossings to determine the effectiveness of ongoing treatment for a patient already diagnosed with neuropathy.

[0120] Various embodiments may be implemented using one or more computer systems, such as computer system 1800 shown in FIG. 18. Computer system 1800 may be used, for example, to implement method 1700 of FIG. 17. Further, computer system 1800 may be at least part of server 100, client device 110, and data storage device 120, as shown in FIG. 1. For example, computer system 1800 routes communications to various applications. Computer system 1800 can be any computer capable of performing the functions described herein.

[0121] Computer system 1800 can be any well-known computer capable of performing the functions described herein.

[0122] Computer system 1800 includes one or more processors (also referred to as central processing units or CPUs), such as processor 1804. Processor 1804 is connected to a communication infrastructure or bus 1806.

[0123] Each of the one or more processors 1804 can be a graphics processing unit (GPU). In one embodiment, the GPU is a processor that is a dedicated electronic circuit designed to process applications that use a lot of computing. The GPU can have a parallel structure that is efficient for parallel processing of large data blocks, such as data that uses a lot of computing common to computer graphics applications, images, videos, etc.

[0124] Computer system 1800 also includes user input / output devices 1803, such as a monitor, keyboard, pointing device, etc., that communicate with communication infrastructure 1806 via user input / output interface 1802.

[0125] Computer system 1800 also includes main or primary memory 1908, such as random access memory (RAM). The main memory 1808 can include one or more levels of cache. The main memory 1808 stores control logic (i.e., computer software) and / or data therein.

[0126] Computer system 1800 can also include one or more secondary storage devices or memories 1810. The secondary memory 1810 can include, for example, hard disk drive 1812 and / or removable storage device or drive 1814. The removable storage drive 1814 can be a floppy (registered trademark) disk drive, magnetic tape drive, compact disk drive, optical storage device, tape backup device, and / or any other storage device / drive.

[0127] The removable storage drive 1814 can interact with a removable storage unit 1818. The removable storage unit 1818 includes a computer-usable or readable storage device on which computer software (control logic) and / or data is stored. The removable storage unit 1818 can be a floppy (registered trademark) disk, magnetic tape, compact disk, DVD, optical storage disk, and / or any other computer data storage device. The removable storage drive 1814 reads from and / or writes to the removable storage unit 1818 in well-known ways.

[0128] According to an exemplary embodiment, the secondary memory 1810 can include other means, methods, or other techniques for enabling a computer program and / or other instructions and / or data to be accessed by the computer system 1800. Such means, methods, or other techniques can include, for example, a removable storage unit 1822 and an interface 1820. Examples of the removable storage unit 1822 and the interface 1820 can include a program cartridge and a cartridge interface (such as those found in video game devices), a removable memory chip (such as an EPROM or PROM) and associated socket, a memory stick and a USB port, a memory card and associated memory card slot, and / or any other removable storage unit and associated interface.

[0129] The computer system 1800 can further include a communication or network interface 1824. The communication interface 1824 enables the computer system 1800 to communicate and interact with any combination of remote devices, remote networks, remote entities, etc. (collectively and individually referred to by reference numeral 1828). For example, the communication interface 1824 can enable the computer system 1900 to communicate with a remote device 1828 via a communication path 1826, which can be wired and / or wireless and can include any combination of a LAN, a WAN, the Internet, etc. Control logic and / or data can be transmitted between the computer system 1800 and via the communication path 1826.

[0130] In one embodiment, a tangible non-transitory device or article that includes a tangible non-transitory computer-usable or readable medium having stored control logic (software) is also referred to herein as a computer program product or program storage device. This includes, but is not limited to, a tangible manufactured article embodying the computer system 1800, main memory 1808, secondary memory 1810, and removable storage units 1818 and 1822, and any combination of the foregoing. Such control logic, when executed by one or more data processing devices (such as computer system 1800), causes such data processing devices to operate as described herein.

[0131] Based on the teachings contained in this disclosure, it will be apparent to those of ordinary skill in the art how to make and use embodiments of this disclosure using data processing devices, computer systems, and / or computer architectures other than those shown in FIG. 18. In particular, embodiments can operate with software, hardware, and / or operating system implementations other than those described herein.

[0132] It should be understood that the detailed description section, rather than other sections, is intended to be used to interpret the claims. The other sections can set forth example embodiments that are one or more, but not all, contemplated by the inventor, and thus are not intended to limit this disclosure or the appended claims in any way.

[0133] The present disclosure describes exemplary embodiments for exemplary fields and applications, but it should be understood that the present disclosure is not limited thereto. Other embodiments and modifications thereto are possible and are within the scope and spirit of the present disclosure. For example, without limiting the generality of this paragraph, embodiments are not limited to the software, hardware, firmware, and / or entities illustrated and / or described herein. Further, embodiments (whether or not explicitly described herein) have significant utility for fields and applications other than those exemplified herein.

[0134] Embodiments are described herein using functional building blocks that represent the implementation of specified functions and their relationships. The boundaries of these functional building blocks are arbitrarily defined herein for the sake of explanation. Alternative boundaries can be defined as long as the specified functions and relationships (or their equivalents) are appropriately implemented. Also, alternative embodiments can execute functional blocks, steps, operations, methods, etc. using an order different from the order described herein.

[0135] References to "one embodiment", "an embodiment", "exemplary embodiment", or similar phrases in this specification indicate that the described embodiment can include a particular feature, structure, or characteristic, but not all embodiments necessarily include the particular feature, structure, or characteristic. Further, such phrases do not necessarily refer to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an embodiment, it will be within the knowledge of those skilled in the art to incorporate such feature, structure, or characteristic into other embodiments, whether or not explicitly recited or described herein. Additionally, some embodiments can be described using the expressions "coupled" and "connected" along with their derivatives. These terms are not necessarily intended to be synonyms of each other. For example, some embodiments can be described using the terms "connected" and / or "coupled" to indicate that two or more elements are in direct physical or electrical contact with each other. However, the term "coupled" can also mean that two or more elements are not in direct contact with each other but still cooperate or interact with each other.

[0136] The breadth and scope of the present disclosure should not be limited by any of the above exemplary embodiments, but should be defined only in accordance with the following claims and their equivalents.

Claims

1. A diagnostic method for determining a nerve state, the method comprising: receiving, by a processor, an image of a patient's skin, the image displaying a plurality of nerves and a plurality of layers of the patient's skin; using a trained model executed by the processor and trained to identify differences between nerves of skin tissue and the plurality of layers to label each nerve among the plurality of nerves of the patient's skin in the image and each layer among the plurality of layers; determining, by the processor, an amount of nerve area relative to a layer of a patient's skin tissue region based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin; determining a nerve state based on the amount of the nerve area relative to the layer of the patient's skin tissue region.

2. The method according to claim 1, wherein the amount of the nerve area relative to the patient's skin tissue region corresponds to nerve density.

3. The method according to claim 2, wherein determining the nerve state includes determining the presence of peripheral neuropathy when the nerve density is lower than a threshold amount.

4. further comprising: receiving, by the processor, a second image of the patient's skin, the second image displaying a plurality of nerves and a plurality of layers of the patient's skin; using the trained model executed by the processor to label each nerve among the plurality of nerves of the patient's skin in the second image and each layer among the plurality of layers; determining, by the processor, a second amount of nerve area relative to a skin tissue region based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin; wherein the amount of the nerve area is a first amount, and determining the nerve state includes determining a difference between the first amount and the second amount corresponding to a difference in nerve density, and determining the presence of peripheral neuropathy when the difference in nerve density is greater than a threshold amount.

5. wherein labeling each nerve among the plurality of nerves and each layer among the plurality of layers in the image of the patient's skin is Using the model executed by the processor, determining whether a pixel among a plurality of pixels in the image corresponds to a nerve in the plurality of nerves of the patient's skin or each layer among the plurality of layers. The method according to claim 1, further comprising, by the processor, assigning to each pixel among the plurality of pixels a respective label indicating whether each respective pixel corresponds to a nerve of the patient's skin or each layer among the plurality of layers. **Claim 6** Further comprising training the model, wherein the training of the model comprises, by the processor, assigning a training label to each nerve of the tissue and each layer among the plurality of layers in each training image among a plurality of training images. comprises, by the processor, inputting the plurality of training images into a convolutional neural network to label the nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images. The method according to claim 1, further comprising, by the processor, verifying each respective label of the nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images assigned by the convolutional neural network using the respective training labels corresponding to the nerve and each layer among the plurality of labels. **Claim 7** The assignment of the training labels to the nerve of the tissue and each layer among the plurality of layers in each image among the plurality of training images The method according to claim 6, comprising, by the processor, applying one or more filters to each image among the plurality of training images to identify the nerve in each training image among the plurality of training images. **Claim 8** The training of the model The method according to claim 6, comprising, by the processor, inputting a mirrored or rotated version of each training image among the plurality of training images into the convolutional neural network to label the nerve of the tissue and each layer among the plurality of layers in the mirrored or rotated version of each training image among the plurality of training images. **Claim 9** The method according to claim 6, wherein the set of images among the plurality of training images includes a duplicate of a single image, and each of the duplicates includes a contrast, brightness, crop, distortion, or rotation that varies from the single image.

10. The method according to claim 1, wherein the layer corresponds to the epidermis.

11. A diagnostic system for determining a neurological state, the system comprising: a memory including stored instructions; a processor coupled to the memory, the instructions, when executed by the processor, causing the processor to: receive an image of a patient's skin, the image displaying a plurality of nerves and a plurality of layers of the patient's skin; using a trained model trained to identify differences between nerves of skin tissue and the plurality of layers, label each nerve among the plurality of nerves of the patient's skin and each layer among the plurality of layers in the image; determine an amount of nerve area relative to a layer of a patient skin tissue region based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin; A diagnostic system that determines a neurological state based on the amount of the nerve area relative to the layer of the tissue region.

12. The system according to claim 11, wherein the amount of the nerve area relative to the tissue region corresponds to nerve density.

13. The system according to claim 12, wherein determining the neurological state includes determining the presence of peripheral neuropathy when the nerve density is lower than a threshold amount.

14. When the instructions are executed by the processor, the processor is further caused to: receive a second image of a patient's skin, the second image displaying a plurality of nerves and a plurality of layers of the patient's skin; using the trained model, label each nerve among the plurality of nerves of the patient's skin and each layer among the plurality of layers in the second image; determine a second amount of nerve area relative to a skin tissue region based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin; wherein the amount of the nerve area is a first amount, and determining the neurological state includes determining a difference between the first amount and the second amount corresponding to a difference in nerve density, and determining the presence of peripheral neuropathy when the difference in nerve density is greater than a threshold amount. The system according to claim 12.

15. Labeling each nerve among the plurality of nerves and each layer among the plurality of layers in the image of the patient's skin, Using the model to determine whether a pixel among the plurality of pixels in the image corresponds to a nerve among the plurality of nerves of the patient's skin or to each layer among the plurality of layers, Assigning to each pixel among the plurality of pixels a respective label indicating whether each respective pixel corresponds to a nerve of the patient's skin or to each layer among the plurality of layers, the system according to claim 11.

16. When executed, causing the processor to train the model, the training of the model comprising: Assigning a training label to each nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images, Inputting the plurality of training images into a convolutional neural network to label the nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images, Verifying each respective label of the nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images assigned by the convolutional neural network using the respective training labels corresponding to the nerve and each layer among the plurality of labels, the system according to claim 11.

17. The assignment of the training labels to the nerve of the tissue and each layer among the plurality of layers in each image among the plurality of training images comprises: Applying one or more filters to each image among the plurality of training images to identify the nerve in each training image among the plurality of training images, the system according to claim 16.

18. The training of the model comprises: Inputting a mirrored or rotated version of each training image among the plurality of training images into the convolutional neural network to label the nerve of the tissue and each layer among the plurality of layers in the mirrored or rotated version of each training image among the plurality of training images, the system according to claim 16.

19. The system of claim 16, wherein the set of images among the plurality of training images includes a copy of a single image, and each of the copies includes a contrast, brightness, crop, distortion, or rotation that varies from the single image.

20. The system of claim 11, wherein the layer corresponds to the epidermis.

21. A non-transitory machine-readable medium having instructions stored thereon, which when executed by at least one computing device, cause the at least one computing device to perform operations, the operations being Receiving an image of a patient's skin, the image displaying a plurality of nerves and a plurality of layers of the patient's skin; Using a trained model trained to identify differences between nerves of skin tissue and the plurality of layers, to label each nerve among the plurality of nerves of the patient's skin in the image and each layer among the plurality of layers; Determining an amount of a nerve region relative to a layer of a patient's skin tissue region based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin; Determining a nerve condition based on the amount of the nerve region relative to the layer of the patient's skin tissue region. A non-transitory machine-readable medium comprising.

22. The non-transitory machine-readable medium of claim 21, wherein the amount of the nerve region relative to the patient's skin tissue region corresponds to nerve density.

23. The non-transitory machine-readable medium of claim 22, wherein determining the nerve condition includes determining the presence of peripheral neuropathy when the nerve density is lower than a threshold amount.

24. The operations are Receiving a second image of a patient's skin, the second image displaying a plurality of nerves and a plurality of layers of the patient's skin; Using the trained model to label each nerve among the plurality of nerves of the patient's skin in the second image and each layer among the plurality of layers; Further comprising determining a second amount of a nerve region relative to a skin tissue region based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin. The amount of the nerve area is the first amount, and determining the nerve state includes determining a difference between the first amount and the second amount corresponding to a difference in nerve density, and determining the presence of peripheral neuropathy when the difference in nerve density is greater than a threshold amount. The non-transitory machine-readable medium according to claim 22.

25. Labeling each nerve among the plurality of nerves and each layer among the plurality of layers in the image of the patient's skin Using the model to determine whether a pixel among the plurality of pixels in the image corresponds to a nerve among the plurality of nerves of the patient's skin or a respective layer among the plurality of layers Assigning to each pixel among the plurality of pixels a respective label indicating whether each respective pixel corresponds to a nerve of the patient's skin or a respective layer among the plurality of layers. The non-transitory machine-readable medium according to claim 21.

26. The operation further includes training the model, and the training of the model Assigning a training label to each nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images Inputting the plurality of training images into a convolutional neural network to label the nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images Verifying each respective label of the nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images assigned by the convolutional neural network using the respective training labels corresponding to the nerve and each layer among the plurality of labels. The non-transitory machine-readable medium according to claim 21.

27. The assignment of the training label to each nerve of the tissue and each layer among the plurality of layers in each image among the plurality of training images Includes applying one or more filters to each image among the plurality of training images to identify the nerve in each training image among the plurality of training images. The non-transitory machine-readable medium according to claim 26.

28. The training of the model Inputting a mirrored or rotated version of each training image among the plurality of training images into the convolutional neural network, and labeling the nerves of the tissue and each layer among the plurality of layers in the mirrored or rotated version of each training image among the plurality of training images. The non - transitory machine - readable medium according to claim 26.

29. The set of images among the plurality of training images includes replicas of a single image, and each of the replicas includes a contrast, brightness, crop, distortion, or rotation that varies from the single image. The non - transitory machine - readable medium according to claim 26.

30. The non - transitory machine - readable medium according to claim 21, wherein the layer corresponds to the epidermis.

31. A diagnostic method for determining a nerve state, the method comprising: Receiving, by a processor, an image of a patient's skin, the image displaying a plurality of nerves and a plurality of layers of the patient's skin; Using a trained model, executed by the processor and trained to identify differences between nerves of skin tissue and the plurality of layers, to label each nerve among the plurality of nerves of the patient's skin and each layer among the plurality of layers in the image; Determining, by the processor, the number of nerve crossings between adjacent layers of a skin tissue region based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin; Determining a nerve state based on the number of nerve crossings. A diagnostic method.

32. The method according to claim 31, wherein the determining the nerve state includes determining the presence of peripheral neuropathy when the number of nerve crossings is lower than a threshold amount.

33. Receiving, by the processor, a second image of the patient's skin, the second image displaying a plurality of nerves and a plurality of layers of the patient's skin; Using the trained model executed by the processor to label each nerve among the plurality of nerves of the patient's skin and each layer among the plurality of layers in the second image; Further comprising determining, by the processor, a second number of nerve crossings between adjacent layers of a skin tissue region in the second image based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin. The number of the nerve crossings is a first number, and determining the nerve state includes determining a difference between the first number and the second number, and determining the presence of peripheral neuropathy when the difference is greater than a threshold amount. The method according to claim 31.

34. Labeling each nerve among the plurality of nerves and each layer among the plurality of layers in the image of the patient's skin, Using the model executed by the processor to determine whether a pixel among a plurality of pixels in the image corresponds to a nerve among the plurality of nerves of the patient's skin or a respective layer among the plurality of layers, The method according to claim 31, including the processor assigning to each pixel among the plurality of pixels a respective label indicating whether each respective pixel corresponds to a nerve of the patient's skin or a respective layer among the plurality of layers.

35. Further including training the model, and the training of the model The processor assigning a training label to each nerve of the tissue and each layer among a plurality of layers in each training image among a plurality of training images, The processor inputting the plurality of training images into a convolutional neural network to label the nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images, The method according to claim 31, including the processor verifying each respective label of the nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images assigned by the convolutional neural network using the respective training labels corresponding to the nerve and each layer among the plurality of labels.

36. The assignment of the training labels to each nerve of the tissue and each layer among the plurality of layers in each image among the plurality of training images The method according to claim 35, including the processor applying one or more filters to each image among the plurality of training images to identify the nerve in each training image among the plurality of training images.

37. The training of the model is The method according to claim 35, comprising: inputting, by the processor, a mirrored or rotated version of each training image among the plurality of training images into the convolutional neural network, and labeling the nerves of the tissue and each layer among the plurality of layers in the mirrored or rotated version of each training image among the plurality of training images.

38. The method according to claim 35, wherein the set of images among the plurality of training images includes replicas of a single image, and each of the replicas includes a contrast, brightness, crop, distortion, or rotation that varies from the single image.

39. The method according to claim 31, wherein the layer corresponds to the epidermis.

40. A diagnostic system for determining a nerve condition, the system comprising: a memory including stored instructions; a processor coupled to the memory, the instructions, when executed by the processor, causing the processor to: receive an image of a patient's skin, the image displaying a plurality of nerves and a plurality of layers of the patient's skin; label each nerve among the plurality of nerves of the patient's skin and each layer among the plurality of layers in the image using a trained model trained to identify differences between nerves of skin tissue and the plurality of layers; determine a number of nerve crossings between adjacent layers of a skin tissue region based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin; a diagnostic system for determining a nerve condition based on the number of nerve crossings.

41. The system according to claim 40, wherein determining the nerve condition includes determining the presence of peripheral neuropathy when the number of nerve crossings is lower than a threshold amount.

42. The instructions, when executed by the processor, further cause the processor to: receive a second image of the patient's skin, the second image displaying a plurality of nerves and a plurality of layers of the patient's skin; label each nerve among the plurality of nerves of the patient's skin and each layer among the plurality of layers in the second image using the trained model; determine a second number of nerve crossings between adjacent layers of a skin tissue region in the second image based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin; The number of the nerve crossings is a first number, and determining the nerve state includes determining a difference between the first number and the second number, and determining the presence of peripheral neuropathy when the difference is greater than a threshold amount. The system according to claim 40.

43. Labeling each nerve among the plurality of nerves and each layer among the plurality of layers in the image of the patient's skin includes Using the model to determine whether a pixel among the plurality of pixels in the image corresponds to a nerve among the plurality of nerves of the patient's skin or a respective layer among the plurality of layers, and Assigning to each pixel among the plurality of pixels a respective label indicating whether each respective pixel corresponds to a nerve of the patient's skin or a respective layer among the plurality of layers. The system according to claim 40.

44. When executed, the instruction causes the processor to train the model, and the training of the model includes Assigning training labels to each nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images, Inputting the plurality of training images into a convolutional neural network to label each nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images, and Verifying each respective label of each nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images assigned by the convolutional neural network using the respective training labels corresponding to the nerve and each layer among the plurality of labels. The system according to claim 40.

45. The assignment of the training labels to each nerve of the tissue and each layer among the plurality of layers in each image among the plurality of training images includes Applying one or more filters to each image among the plurality of training images to identify the nerves in each training image among the plurality of training images. The system according to claim 44.

46. The training of the model includes Inputting a mirrored or rotated version of each of the plurality of training images into the convolutional neural network, and labeling the nerves of the tissue and each of the plurality of layers in the mirrored or rotated version of each of the plurality of training images, the system according to claim 44.

47. The system according to claim 44, wherein the set of images among the plurality of training images includes replicas of a single image, and each of the replicas includes a contrast, brightness, crop, distortion, or rotation that varies from the single image.

48. The system according to claim 40, wherein the layer corresponds to the epidermis.

49. A non-transitory machine-readable medium having instructions stored therein, which when executed by at least one computing device, cause the at least one computing device to perform operations, the operations being Receiving an image of a patient's skin, the image displaying a plurality of nerves and a plurality of layers of the patient's skin, Using a trained model trained to identify differences between nerves of skin tissue and the plurality of layers to label each nerve among the plurality of nerves of the patient's skin and each of the plurality of layers in the image, Determining the number of nerve crossings between adjacent layers of a skin tissue region based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin, Determining a nerve state based on the number of nerve crossings, a non-transitory machine-readable medium.

50. The non-transitory machine-readable medium according to claim 49, wherein the determining the nerve state includes determining the presence of peripheral neuropathy when the number of nerve crossings is lower than a threshold amount.

51. The operations are Receiving a second image of a patient's skin, the second image displaying a plurality of nerves and a plurality of layers of the patient's skin, Using the trained model to label each nerve among the plurality of nerves of the patient's skin and each of the plurality of layers in the second image, Further including determining a second number of nerve crossings between adjacent layers of a skin tissue region in the second image based on the labels assigned to the plurality of nerves and the plurality of layers of the patient's skin, The number of the nerve crossings is a first number, and the determining of the nerve state includes determining a difference between the first number and the second number, and determining the presence of peripheral neuropathy when the difference is greater than a threshold amount. The non-transitory machine-readable medium according to claim 49.

52. The labeling of each nerve among the plurality of nerves and each layer among the plurality of layers in the image of the patient's skin includes using the model to determine whether a pixel among a plurality of pixels in the image corresponds to a nerve among the plurality of nerves of the patient's skin or a respective layer among the plurality of layers, and assigning to each pixel among the plurality of pixels a respective label indicating whether each respective pixel corresponds to a nerve of the patient's skin or a respective layer among the plurality of layers. The non-transitory machine-readable medium according to claim 49.

53. The operation further includes training the model, and the training of the model includes assigning a training label to each nerve of the tissue and each layer among a plurality of layers in each training image among a plurality of training images, inputting the plurality of training images into a convolutional neural network to label the nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images, and verifying each respective label of the nerve of the tissue and each layer among the plurality of layers in each training image among the plurality of training images assigned by the convolutional neural network using the respective training labels corresponding to the nerve and each layer among the plurality of labels. The non-transitory machine-readable medium according to claim 49.

54. The assignment of the training labels to each nerve of the tissue and each layer among the plurality of layers in each image among the plurality of training images includes applying one or more filters to each image among the plurality of training images to identify the nerve in each training image among the plurality of training images. The non-transitory machine-readable medium according to claim 53.

55. The training of the model includes Inputting a mirrored or rotated version of each of the plurality of training images into the convolutional neural network to label the nerves of the tissue and each of the plurality of layers in the mirrored or rotated version of each of the plurality of training images, the non-transitory machine-readable medium according to claim 53.

56. The set of images among the plurality of training images includes replicas of a single image, each of the replicas including a contrast, brightness, crop, distortion, or rotation that varies from the single image, the non-transitory machine-readable medium according to claim 53.

57. The non-transitory machine-readable medium according to claim 49, wherein the layer corresponds to the epidermis.

Images