Assessment of facial paralysis and gaze deviation

Abstract

Images of the individual may be acquired and analyzed to determine an amount of facial paralysis in the individual. The images may be analyzed to determine an amount of gaze deviation in the individual. The amount of facial paralysis in the individual and / or the amount of gaze deviation in the individual may be used to determine a probability that the individual is suffering from a biological condition.

Description

[Background technology]

[0001] Various biological conditions can result in facial paralysis and / or gaze deviation. For example, cerebrovascular accidents, brain lesions, and nerve paralysis can result in gaze deviation. Furthermore, Bell's palsy, brain tumors, tick bites, or seizures can cause facial paralysis. Regarding stroke, intracranial blood vessels supply nutrients and oxygen to the brain, including the cerebrum, cerebellum, and brainstem. Some arteries supply blood to the anterior part of the brain, while others supply blood to the posterior part. Interruption of blood flow to any part of the brain can have serious consequences. Blood flow to the brain can be blocked by narrowing and / or occlusion of blood vessels supplying the brain. Interruption of blood flow to a part of the brain can reduce brain function and lead to numbness, weakness, or paralysis in a body part. When blood supply to a part of the brain is interrupted, a stroke can occur. Early detection and treatment of stroke can minimize damage to the part of the brain where blood supply is interrupted and minimize the sequelae of the stroke. [Brief explanation of the drawings]

[0002] [Figure 1] FIG. 1 is a schematic diagram of an architecture for determining measurements of gaze deviation and facial paralysis in accordance with one or more exemplary embodiments. [Figure 2] FIG. 1 is a schematic diagram of an architecture for determining gaze deviation of an individual based on image data in accordance with one or more exemplary embodiments. [Figure 3] FIG. 1 is a schematic diagram of an architecture for determining the amount of facial paralysis in an individual based on image data in accordance with one or more exemplary embodiments. [Figure 4] 1 is a flowchart illustrating example operations of a process for analyzing an individual's amount of gaze deviation and facial paralysis to identify the probability that the individual has a biological condition, according to one or more exemplary embodiments. [Figure 5] It includes a plurality of images showing individuals without facial paralysis and individuals with facial paralysis. [Figure 6] It includes images rendered using computed tomography data of an individual's face, different images corresponding to different perspectives of the individual's face. [Figure 7] It includes images showing the location of occluded vessels in individuals who have had a stroke and the correlation between gaze deviation and the location of occluded vessels. [Figure 8] FIG. 1 is a block diagram illustrating components of a machine in the form of a computer system that may read and execute instructions from one or more machine-readable media to perform any one or more methods described herein, in accordance with one or more exemplary embodiments. [Figure 9] FIG. 1 is a block diagram illustrating a representative software architecture that may be used in conjunction with one or more hardware architectures in accordance with one or more exemplary embodiments. DETAILED DESCRIPTION OF THE INVENTION

[0003] In the drawings, which are not necessarily drawn to scale, like numerals may describe like components in different figures. To easily identify any particular element or description of an operation, the most significant digit(s) in a reference number refers to the figure number in which that element is first introduced. Some implementations are shown by way of example, and not by way of limitation.

[0004] Various imaging techniques are employed to detect the presence (current or past) of a biological condition that may result in facial paralysis and / or stroke. In one or more examples, the biological condition may include a neurological condition. In one or more instances, the neurological condition may include a stroke in an individual.

[0005] In one or more embodiments, computed tomography (CT) imaging techniques can be used to obtain images of the blood vessels supplying blood to an individual's brain. In addition, magnetic resonance (MR) imaging techniques can also be used to obtain images of the blood vessels in an individual's brain. Imaging techniques used to generate images of the blood vessels supplying blood to an individual's brain can include perfusion and diffusion techniques. Images of the blood vessels in an individual's brain generated by one or more imaging techniques can be analyzed to identify areas of the individual's brain that have blocked blood vessels. Images of an individual's brain generated by one or more imaging techniques can also be analyzed to identify areas of the individual's brain that have been damaged by a lack of blood supply due to blocked blood vessels.

[0006] In addition to imaging-based techniques, clinicians can provide a diagnosis of stroke based on findings obtained during their own examination of an individual. In situations where a blocked blood vessel is the cause of a stroke, an individual's electrophysiology may not function properly, in addition to possible tissue damage due to lack of blood flow. Among various examples, gaze deviation can be an indicator of a stroke. Gaze deviation can be detected when the iris of at least one of an individual's eyes is misaligned relative to a reference position. The direction of the gaze deviation can indicate the location of the stroke within the individual's brain. For example, the direction in which an individual's gaze is directed may be in the same direction as the side of the brain where the blocked blood vessel is located. By way of example, a gaze deviation to the right (from a viewpoint facing the individual) may indicate a blocked blood vessel on the right side of the individual's brain, while a gaze deviation to the left (from a viewpoint facing the individual) may indicate a blocked blood vessel on the left side of the individual's brain.

[0007] Additionally, clinicians may observe facial paralysis in individuals experiencing a stroke. In some circumstances, vascular occlusion in the brain may cause paralysis of the individual's facial muscles, thereby altering the appearance of one or more areas of the individual's face. Facial drooping may occur as a result of changes in the individual's facial expression during a stroke. In various instances, muscles on the opposite side of the face may compensate for the drooping, thereby also altering the appearance of the opposite side of the face. The side of the face on which the drooping occurs may indicate that the location of the stroke, resulting from the occlusion of a large blood vessel, is in the opposite hemisphere of the individual's brain. For example, facial drooping on the right side of an individual's face (as viewed from a vantage point opposite the individual) may indicate the presence of an occluded blood vessel on the left side of the individual's brain, while facial drooping on the left side of an individual's face (as viewed from a vantage point opposite the individual) may indicate the presence of an occluded blood vessel on the right side of the individual's brain.

[0008] Typically, images of an individual's brain are analyzed by a trained professional, such as a radiologist, to identify blocked blood vessels and / or locate damage to brain tissue. In some instances, the radiologist receives medical records from clinicians who have examined the individual, but these medical records often do not indicate the individual's physical symptoms. In addition, radiologists typically cannot physically examine the individual whose images they are examining. The accuracy and speed with which a stroke is detected can affect the treatment administered to the individual as well as the individual's prognosis for recovery. Therefore, tools, techniques, systems, and methods that can provide additional information that can improve the accuracy of stroke detection can improve the individual's treatment and aid in the recovery of individuals who have suffered or are currently suffering from a stroke.

[0009] The techniques, systems, processes, and methods described herein relate to identifying an individual's gaze deviation and facial paralysis based on image data. In various examples, image data captured for an individual can be analyzed with respect to reference image data or training image data. Gaze deviation metrics can be determined based on image data indicative of the individual's degree of gaze deviation. Additionally, facial paralysis metrics can be determined based on image data indicative of the individual's degree of facial paralysis. In various examples, gaze deviation metrics and / or facial paralysis metrics can be used to identify the likelihood that an individual has suffered a stroke. Identifying metrics indicative of the degree of gaze deviation and facial paralysis can help improve the accuracy of a radiologist's assessment of the presence or absence of blocked blood vessels in the individual's brain. In one or more examples, metrics indicative of the degree of gaze deviation and facial paralysis can be used to capture image information that can indicate a blockage of the individual's blood vessels, such as images of the blood vessels in the individual's brain.

[0010] The image data used to determine the gaze deviation metrics and facial paresis metrics can be captured by one or more image data sources. For example, the image data can be captured by one or more cameras on a computing device, such as a mobile phone, smartphone, tablet computing device, wearable device, or laptop computing device. Additionally, the image data can be captured by a CT imaging device or an MR imaging device. When the image data is captured by a CT imaging device or an MR imaging device, the techniques, systems, processes, and methods described herein provide 3D renderings corresponding to external features of an individual's head, such as the individual's facial paresis. Therefore, in contrast to techniques performed by existing systems that render portions of the CT or MR image data corresponding to the interior of an individual's brain, such as blood vessels and other tissues, the techniques, systems, methods, and processes described herein are directed to rendering the external features of an individual. In this way, radiologists can obtain visual information about an individual almost as if the patient were in front of them, which is not possible with existing systems that simply provide internal images of an individual's brain. In various examples, combining an external image of an individual's face and metrics derived from the external image of the individual's face with an internal image of the individual's brain can also increase the accuracy of diagnosing an individual as to whether or not they have had a stroke. Because the techniques, systems, methods, and processes described herein can be performed using a mobile computing device, an initial assessment of a patient's facial paralysis and / or gaze deviation can be made by the paramedic treating the patient. In this way, an initial assessment of a patient's condition can be made before the patient is seen by a trained specialist, such as a neurologist or emergency physician.

[0011] 1 is a schematic diagram of an architecture 100 for determining the degree of gaze deviation and facial paralysis based on image data in accordance with one or more exemplary embodiments. The architecture 100 may include an image processing system 102. The image processing system 102 may be implemented by one or more computing devices 104. The one or more computing devices 104 may include one or more server computing devices, one or more desktop computing devices, one or more mobile computing devices, or a combination thereof. In certain embodiments, at least a portion of the one or more computing devices 104 may be implemented within a distributed computing environment. For example, at least a portion of the one or more computing devices 104 may be implemented within a cloud computing architecture.

[0012] The image processing system 102 can obtain image data 106 captured regarding the individual 108. The image data 106 can include one or more data files containing data regarding images captured by one or more image data sources. In one or more examples, the image data 106 can be formatted according to one or more imaging data formats, such as the Digital Imaging and Communication in Medicine (DICOM) format. In one or more additional examples, the image data 106 can be formatted according to at least one of the Portable Network Graphics (PNG) format, the Joint Photographic Experts Group (JPEG) format, or the High Efficient Image File (HEIF). The image data 106 can be rendered by the image processing system 102 to generate one or more images that can be displayed on a display device. In various examples, the image data 106 can include a series of images of the individual 108 captured by one or more imaging data sources. In one or more instances, the imaging data 106 may correspond to one or more features of the head of the individual 108. In one or more situations, the imaging data 106 may be rendered to show external features of the head of the individual 108, such as at least a portion of the face of the individual 108. In one or more other instances, the image data 106 may be rendered to show internal features of the head of the individual 108, such as blood vessels or brain tissue within the head of the individual 108.

[0013] The image data 106 can be captured by the imaging device 110. The imaging device 106 can capture the image 104 using one or more imaging techniques. In one or more examples, the imaging device 110 can implement a computed tomography (CT) imaging technique. In one or more other examples, the imaging device 110 can implement a magnetic resonance (MR) imaging technique. The MR imaging technique implemented by the imaging device 110 can include a perfusion MR imaging technique or a diffusion MR imaging technique. When the imaging device 110 implements a CT-based imaging technique or an MR-based imaging technique, the image data 106 can include thin-slice volumetric data.

[0014] The one or more image data sources may also include one or more computing devices 112. The one or more computing devices 112 may include at least one of a mobile computing device, a smartphone, a wearable device, a tablet computing device, or a laptop computing device. In one or more examples, the one or more computing devices 112 may generate a structured light pattern over at least a portion of a face of the individual 108 and capture one or more images of the individual 108 in response to the structured light pattern. In these examples, the image data 106 may include a point cloud. In one or more additional examples, the one or more computing devices 112 may include one or more laser-based time-of-flight cameras. In these situations, the one or more computing devices 112 may include a light detecting and ranging (LiDAR) system. In one or more other examples, the one or more computing devices 112 may include multiple cameras, and the image data 106 may correspond to stereoscopic images. In one or more examples, the image data may also include a random point cloud or a sinusoidal pattern.

[0015] The image processing system 102 can include a gaze deviation analysis system 114. The gaze deviation analysis system 114 can analyze the image data 106 to determine whether the gaze of the individual 108 differs from the gaze of an individual who has not suffered a stroke. For example, the gaze deviation analysis system 114 can determine gaze deviation metrics 116 for the individual 108. The gaze deviation metrics 116 can indicate a deviation in the gaze of at least one eye of the individual 108. The deviation in the gaze of one eye of the individual 108 can be determined by analyzing the position of the iris of the eye of the individual 108 relative to an expected position of that iris. The gaze deviation metrics 116 can indicate the amount by which the position of the iris of the eye of the individual 108 differs from the expected position of that iris. In various examples, the gaze deviation analysis system 114 can identify the angular offset of the iris of the individual 108 in degrees, such as 0.5 degrees, 1 degree, 2 degrees, 5 degrees, 10 degrees, 15 degrees, etc. The gaze deviation analysis system 114 can also identify a category that indicates the amount of gaze deviation for the individual 108. In one or more instances, the gaze deviation metrics 116 can indicate categories such as "no deviation," "slight deviation," "significant deviation," etc. The gaze deviation metrics 116 can also include a directional component such as "left" and "right" that indicates which direction the iris of the individual 108 is pointing relative to an expected orientation. In one or more instances, the gaze deviation metrics 116 can indicate categories such as "no deviation," "slight deviation," "significant deviation," etc. In addition to the gaze deviation metrics 116 being classified as categorical data, the gaze deviation metrics can also be characterized as continuous or ordinal data.

[0016] The image processing system 102 can also include a facial paralysis analysis system 118. The facial paralysis analysis system 118 can analyze the imaging data 106 to determine whether the individual 108 is experiencing facial paralysis. In various examples, the facial paralysis analysis system 118 can analyze the facial appearance of the individual 108 with respect to the expected facial appearance of an individual who has had a stroke and the expected facial appearance of an individual who has not had a stroke. In various examples, the facial paralysis analysis system 118 can determine a degree of similarity of the facial appearance of the individual 108 with respect to the expected facial appearance of an individual who has had a stroke and the expected facial appearance of an individual who has not had a stroke. In one or more instances, the facial paralysis analysis system 118 can analyze target regions on the face of the individual 108 to generate facial paralysis metrics 120. The facial paralysis metric 120 can indicate whether the characteristics of the target region on the face of the individual 108 are closer to the characteristics of the target region on the face of an individual without a stroke or closer to the characteristics of the target region on the face of an individual with a stroke. The facial paralysis metric 120 can include a numerical metric that indicates the amount of difference between the characteristics of the target region on the face of the individual 108 and the characteristics of the target region on the face of an individual with a stroke. In various examples, the facial paralysis metric 120 can indicate a categorical value that indicates the amount of difference between the characteristics of the target region on the face of the individual 108 and the characteristics of the target region on the face of an individual with a stroke. For example, the facial paralysis metric 120 can include "no paralysis," "minor paralysis," or "significant paralysis." The facial paralysis metric 120 can also include a directional component, such as "left" and "right," that indicates the side of the face of the individual 108 on which facial paralysis may be present. In these situations, the facial paralysis metric may indicate "minor paralysis, left," "minor paralysis, right," etc.

[0017] The gaze deviation metrics 116 and the facial paralysis metrics 120 can be provided to an evaluation system 122 of the image processing system 102. The evaluation system 122 can analyze the gaze deviation metrics 116 and the facial paralysis metrics 120 to determine a probability that the individual 108 is suffering from a biological condition, such as a stroke or Bell's palsy. The probability that the individual 108 is suffering from a biological condition can be higher if the gaze deviation metrics 116 or the facial paralysis metrics 120 have a relatively high value. Additionally, the probability that the individual 108 is suffering from a biological condition can be lower if the gaze deviation metrics 116 or the facial paralysis metrics 120 have a relatively low value.

[0018] In one or more examples, the image processing system 102 can generate a system output 124. The system output 124 can include assessment metrics 126 corresponding to the probability that the individual 108 has a biological condition. As used herein, a biological condition can refer to an abnormality in an individual's function and / or structure to a degree that results in or threatens to result in a detectable characteristic of the abnormality. A biological condition can be characterized by external and / or internal features, signs, and / or symptoms indicative of a deviation from biological norms in one or more populations. The biological condition can include at least one of one or more diseases, one or more disorders, one or more infections, one or more isolated symptoms, or other abnormal changes in the individual's biological structure and / or function. In one or more instances, the one or more biological conditions can include one or more neurological conditions. In various examples, the system output 124 generated by the image processing system 102 can include user interface data corresponding to a user interface that includes the assessment metrics 126. The evaluation metric 126 may indicate a numerical value of the probability that the biological condition exists in the individual 108. In one or more additional examples, the evaluation metric 126 may indicate a categorical value of the probability that the biological condition exists in the individual 108. For example, the evaluation metric 126 may include "no biological condition," "minimal probability of biological condition," or "significant probability of biological condition." In one or more examples, the evaluation metric 126 may include "no stroke," "minimal probability of stroke," or "significant probability of stroke."

[0019] The system output 124 may also include one or more rendered images 128. The one or more rendered images 128 may include one or more images of external features of the individual 108. For example, the one or more rendered images 128 may include a first image 130 of the face of the individual 108. Additionally, the one or more rendered images 128 may include one or more images of internal features of the individual 108. By way of example, the one or more rendered images 128 may include a second image 132 of the blood vessels of the brain of the individual 108. In one or more instances, the system output 124 may include a user interface that includes the first image 130 or the second image 132. In one or more additional examples, the system output 124 may include a user interface that includes both the first image 130 and the second image 132.

[0020] Additionally, in one or more other situations, image processing system 102 can determine assessment metrics 126 based on an analysis of external features of the head of individual 108 and internal features of the head of individual 108. For example, image processing system 102 can analyze vascular features of the brain of individual 108 to determine the probability that individual 108 has had a stroke. Image processing system 102 can combine the probability that individual 108 has had a stroke based on the analysis of the individual's cerebral vasculature with the probability that individual 108 has had a stroke determined by assessment system 122. In these examples, the output generated by assessment system 122 can complement the image processing system's analysis of the cerebral vasculature of individual 108. In various examples, the accuracy of the image processing system 102 in determining the probability that an individual 108 has a stroke can be increased by combining the probability that an individual 108 has a stroke determined by the evaluation system 122 with the probability that an individual 108 has a stroke determined based on an analysis of the blood vessels in the brain of the individual 108.

[0021] 1 shows image processing system 102 including both gaze deviation analysis system 114 and facial paralysis analysis system 118, in one or more additional embodiments, image processing system 102 can include gaze deviation analysis system 114 or facial paralysis system 118. If image processing system 102 includes gaze deviation analysis system 114 but not facial paralysis analysis system 118, evaluation system 122 can use gaze deviation metrics 116 rather than facial paralysis metrics 120 to determine evaluation metrics 126. In additional cases where image processing system 102 includes facial paralysis analysis system 118 but not gaze deviation analysis system 114, evaluation system 122 can use facial paralysis metrics 120 rather than gaze deviation metrics 116 to determine evaluation metrics 126.

[0022] 2 is a schematic diagram of an architecture 200 for determining an amount of gaze deviation of an individual based on image data in accordance with one or more exemplary embodiments. The architecture 200 may include a gaze deviation analysis system 114. The gaze deviation analysis system 114 may include one or more gaze deviation models 202. The one or more gaze deviation models 202 may be implemented to generate the gaze deviation metrics 116. The one or more gaze deviation models 202 may be generated using one or more machine learning techniques. The one or more machine learning techniques may include one or more classification machine learning techniques. In various examples, the one or more machine learning techniques may include a supervised classification machine learning technique. The one or more supervised classification machine learning techniques may include at least one of one or more logistic regression algorithms, one or more naive Bayes algorithms, one or more K-nearest neighbor algorithms, or one or more support vector machine algorithms. In one or more additional examples, the one or more machine learning techniques used to generate the one or more gaze deviation models 202 can include one or more residual neural networks. In one or more other examples, the one or more machine learning techniques used to generate the one or more gaze deviation models 202 can include at least one of one or more u-net convolutional networks, one or more v-net convolutional neural networks, or one or more residual neural networks (ResNet). In one or more instances, a combination of machine learning techniques can be used to generate the one or more gaze deviation models 202. For example, one or more residual neural networks and one or more u-net convolutional networks can be used to generate the one or more gaze deviation models 202.

[0023] In one or more embodiments, one or more gaze detection models 202 can be trained using training data 204. The training data 204 can include image training data 206. In various examples, the image training data 206 can include facial images of a large number of different individuals. At least a portion of the image training data 206 can include images captured from individuals experiencing a biological condition. Additionally, at least a portion of the image training data 206 can include images captured from individuals not experiencing a biological condition. In one or more instances, the training data 206 can include a series of images of a large number of individuals moving their eyes in different directions and to different degrees. The training data 204 can also include classification data 208 indicative of each image included in the image training data 206 corresponding to individuals experiencing a biological condition and each additional image included in the image training data 206 corresponding to individuals not experiencing a biological condition. Additionally, the training data 204 can include a plurality of images that can be used to validate one or more gaze deviation models 202. In one or more instances, a convolutional neural network split 80 / 20 using 5-fold validation can be trained using the training data 206 to generate one or more gaze deviation models 202.

[0024] In various examples, one or more gaze deviation models 202 can be trained to identify images in which an individual's gaze deviates with respect to an expected gaze. In one or more examples, the expected gaze may correspond to the gaze of an individual without a stroke when the individual is looking straight ahead at an object. The one or more gaze deviation models 202 can identify an individual's expected gaze by analyzing images included in the image training data 206 of individuals without a biological condition looking straight ahead at an object to identify the individual's eye characteristics. In one or more instances, the one or more gaze deviation models 202 can identify an expected gaze of an individual without a biological condition as the gaze when the individual's irises are aligned with their respective vertical axes. FIG. 2 includes an exemplary expected gaze 210 in which a first iris 212 and a second iris 214 are aligned with a first vertical axis 216 and a second vertical axis 218, respectively.

[0025] The one or more gaze deviation models 202 can also be trained to identify ocular characteristics of an individual with gaze deviation. Gaze deviation can be identified as an individual's gaze having at least a threshold difference amount from an expected gaze. The threshold difference amount can include at least a minimum angular offset with respect to each vertical axis. The minimum angular offset can be at least about 0.5 degrees, at least about 1 degree, at least about 2 degrees, at least about 3 degrees, at least about 4 degrees, or at least about 5 degrees. In one or more examples, the minimum angular offset can be between about 0.5 degrees and about 5 degrees, between about 1 degree and about 3 degrees, or between about 0.5 degrees and about 2 degrees. After training and validation of the one or more gaze deviation models 202, the one or more gaze deviation models 202 can be implemented to classify image data 106 of an individual 108.

[0026] The one or more gaze deviation models 202 can generate gaze deviation metrics 116 based on the image data 106. The example in FIG. 2 includes an example gaze 220 obtained from the image data 106. The example gaze 220 includes a third iris 220 and a fourth iris 222. The example gaze 220 shows that the third iris 220 is misaligned with the first vertical axis 216 by an angular offset 224. The example gaze 220 also shows that the fourth iris 222 is aligned with the second vertical axis 222 without any angular offset. The one or more gaze deviation models 202 can determine whether the angular offset 224 is greater than a minimum angular offset. If the angular offset 224 is greater than the minimum angular offset, the one or more gaze deviation models 202 can classify the individual 108 as having a gaze deviation corresponding to at least a minimum probability that the individual 108 has suffered a stroke. In various examples, the degree of angular offset 224 above the minimum angular offset can indicate the magnitude of the probability that a stroke has occurred to the individual 108. For example, as angular offset 224 increases, the probability that the individual 108 has suffered a stroke can also increase. Additionally, as angular offset 224 increases, the severity of the stroke can also increase. While the example in FIG. 2 shows angular offset 224 for the third iris 220, in other examples, there can also be an additional angular offset for the fourth iris 222. Additionally, both the third iris 220 and the fourth iris 222 can be misaligned. In one or more examples, the additional angular offset of the fourth iris 222 can be the same as angular offset 224. In one or more other examples, the angular offset of the fourth iris 222 can be different from angular offset 224.

[0027] In one or more additional examples, the one or more gaze deviation models 202 can determine an amount of gaze deviation for the individual 108 based on the reference image data 210. In one or more examples, the reference image data 210 can include one or more images of the individual 108 captured during a period when the biological condition was not occurring for the individual 108. In these situations, the one or more gaze deviation models 202 can analyze the image data 106 with respect to the reference image data 210 to determine an amount of difference between the gaze of the individual 108 included in the reference image data 210 and the gaze of the individual 108 included in the image data 106. In various examples, the expected gaze 210 can correspond to the gaze of the individual 108 determined based on the reference image data 210.

[0028] In one or more examples, the training data 204 and image data 106 can be preprocessed by the gaze deviation analysis system 114 or the image processing system 102 of FIG. 1 before being processed by the one or more gaze deviation models 202. For example, one or more computational techniques can be applied to the training data 204 and the image data 106 to generate a modified data set, which is then processed according to one or more machine learning techniques implemented by the one or more gaze deviation models 202. Three-dimensional (3D) rendering can be performed when at least one of the training data 204 or the image data 106 includes three-dimensional volume data from a CT-based or MR-based imaging device. 3D rendering can also be performed on data captured via a LiDAR system, a time-of-flight system, and / or a point cloud / structured light system. 3D rendering can include volume rendering or surface rendering. In one or more instances, a binary mask can be generated from an image included in at least one of the training data 204 or the image data 106. The binary mask can represent voxels of the image, which are foreground and background voxels. Additionally, the binary mask can be used to generate a polygon mesh corresponding to the surfaces of objects contained in the image, such as the face of an individual corresponding to the image. The polygon mesh can be rendered to generate an image of the individual's face that can be analyzed by one or more gaze deviation models 202.

[0029] Images obtained from the training data 204 and the image data 106 can also be processed to identify the eyes of an individual. In one or more examples, the classification data 208 can identify the eyes of an individual included in images in the image training data 206. In this manner, one or more gaze deviation models 202 can be trained to identify the eyes of an individual. In one or more additional examples, a filter can be generated and used to identify the eyes of an individual. The filter for identifying the eyes of an individual can be generated based on the training data 204. In various examples, the gaze deviation analysis system 114 may determine that at least one eye of the individual 108 cannot be identified. In such a situation, the gaze deviation analysis system 114 can provide an indication that the gaze deviation analysis system 112 cannot analyze the image data 106. In one or more examples, at least one eye of the individual 108 may be blocked by an object such as sunglasses, hair, or the like.

[0030] Additionally, at least one of the training data 204 or the image data 106 can be augmented before being processed by the one or more gaze deviation models 202. In one or more examples, volumetric data obtained from an MR-based imaging device, a CT-based imaging device, a LiDAR system, a time-of-flight system, a structured light system, or images captured by a mobile device camera from many different viewpoints can be used to generate multiple images corresponding to multiple viewpoints. In this manner, an individual's gaze and facial palsy can be analyzed from multiple viewpoints. In one or more additional examples, one or more viewpoints can be identified, which provides an image of the individual's gaze that can be processed by the one or more gaze deviation models 202 to improve the accuracy of the gaze deviation metrics 116. For example, one or more images associated with a viewpoint showing the individual facing forward and looking ahead can be selected, rather than images from a viewpoint where the individual is looking to one side or not looking ahead, to be processed by the one or more gaze deviation models. In one or more other examples, at least one of the training data 204 or the image data 106 may include multiple images corresponding to different viewpoints of an individual. Illustratively, one or more cameras of a computing device may be used to capture images of the individual from various viewpoints (relative to the orientation of the computing device), such as one or more right-hand viewpoints, one or more left-hand viewpoints, or a front-facing viewpoint. In these examples, images corresponding to the front-facing viewpoints may be selected for analysis by one or more gaze deviation models 202.

[0031] Although an individual's gaze deviation has been described with reference to FIG. 2 in terms of the individual's iris, gaze deviation can also be determined based on the position of the individual's pupil. 3 is a schematic diagram of an architecture 300 for determining the amount of facial paralysis in an individual in accordance with one or more exemplary embodiments. The architecture 300 may include a facial paralysis analysis system 118. The facial paralysis analysis system 118 may include one or more facial paralysis models 302. The one or more facial paralysis models 302 may be implemented to generate the facial paralysis metrics 120. The one or more facial paralysis models 302 may be generated using one or more machine learning techniques. The one or more machine learning techniques may include one or more classification machine learning techniques. In various examples, the one or more machine learning techniques may include a supervised classification machine learning technique. The one or more supervised classification machine learning techniques may include at least one of one or more logistic regression algorithms, one or more naive Bayes algorithms, one or more K-nearest neighbor algorithms, or one or more support vector machine algorithms. In one or more additional examples, the one or more machine learning techniques used to generate the one or more facial paralysis models 302 may include one or more residual neural networks. In one or more other examples, the one or more machine learning techniques used to generate the one or more facial paralysis models 302 may include one or more u-net convolutional networks, one or more v-net convolutional neural networks, or one or more residual neural networks (ResNet). In one or more instances, a combination of machine learning techniques may be used to generate the one or more facial paralysis models 302. Illustratively, the one or more facial paralysis models 302 may be generated using one or more residual neural networks and one or more u-net convolutional networks.

[0032] In one or more embodiments, one or more facial paralysis models 302 can be trained using training data 304. The training data 304 can include image training data 306. In various examples, the image training data 306 can include images of a large number of different people's faces. At least a portion of the image training data 306 can include images captured from individuals experiencing a biological condition. Additionally, at least a portion of the image training data 306 can include images captured from individuals not experiencing a biological condition. The training data 304 can also include classification data 308 indicative of each of the images included in the image training data 306 corresponding to individuals experiencing a biological condition and additional images included in the image training data 306 corresponding to individuals not experiencing a biological condition. Additionally, the training data 304 can include a plurality of images that can be used to validate the one or more facial paralysis models 302. In one or more examples, a convolutional neural network split 80 / 20 using 5-fold validation can be trained to generate one or more facial paralysis models 302 using training data 306.

[0033] In various examples, one or more facial paralysis models 302 can be trained to identify target regions on an individual's face. The target regions can be regions of the individual's face that may exhibit facial paralysis. The one or more facial paralysis models 302 can be trained to identify features of the target regions that are indicative of facial paralysis. In one or more examples, the one or more facial paralysis models can be trained to identify the amount of facial droop on one side of the individual's face. The one or more facial paralysis models 302 can analyze images of individuals without a stroke and images of individuals with a stroke to identify differences between the features of the target regions in individuals with a stroke and those without a stroke. In one or more circumstances, the features of the target regions in individuals with a stroke may be shifted relative to the features of the target regions in individuals without a stroke. The facial paralysis metrics 120 may indicate the amount of similarity between features of a target region of an individual included in an image being analyzed by the facial paralysis analysis system 118 and features of a target region of an individual having a stroke. In one or more instances, the facial paralysis metrics 120 may indicate a degree of facial paralysis of the individual, e.g., a numerical or categorical value, and / or a combination thereof.

[0034] In this manner, the one or more facial paralysis models 302 can analyze the image data 106 to determine whether features of target regions on the face of the individual 108 correspond to features of target regions of an individual experiencing a biological condition or features of target regions of an individual not experiencing a biological condition. In the example of FIG. 3 , an image 312 can be rendered based on the image data 106. The one or more facial paralysis models 302 can analyze features of multiple frontal regions on the face of the individual 108, including a first target region 314, a second target region 316, a third target region 318, a fourth target region 320, and a fifth target region 322. The one or more facial paralysis models 302 can analyze the target regions 314, 316, 318, 320, 322 to determine one or more features of the target regions 314, 316, 318, 320, 322 of the individual 108. Additionally, the one or more facial paralysis models 302 may determine facial paralysis metrics 120 for the individual 108 based on characteristics of the target regions 314 , 316 , 318 , 320 , 322 of the individual 108 .

[0035] In one or more additional examples, the one or more facial paralysis models 302 can identify an amount of facial paralysis in the individual 108 based on the reference image data 310. In one or more examples, the reference image data 310 can include one or more images of the individual 108 captured during a period in which the individual 108 was not experiencing a stroke. In these circumstances, the one or more facial paralysis models 302 can analyze the image data 106 with respect to the reference image data 310 to identify an amount of difference between the features of the target regions 314, 316, 318, 320, 322 of the individual 108 included in the reference image data 310 and the features of the target regions 314, 316, 318, 320, 322 of the individual 108 identified based on the image data 106.

[0036] Further, the one or more facial paralysis models 302 may determine an amount of facial paralysis of the individual 108 by analyzing characteristics of at least some of the target regions 314, 316, 318, 320, and 322 on either side of the face of the individual 108. For example, the one or more facial paralysis models 302 may analyze characteristics of the first target region 314 with respect to characteristics of the fifth target region 322. Additionally, the one or more facial paralysis models 302 may analyze characteristics of the second target region 316 with respect to characteristics of the fourth target region 320. The one or more facial paralysis models 302 may determine an amount of facial paralysis of the individual 308 based on an amount of difference between the characteristics of the first target region 314 and the fifth target region 322 and / or an amount of difference between the characteristics of the second target region 316 and the fourth target region 320.

[0037] In one or more examples, the training data 304 and the image data 106 can be preprocessed by the facial paralysis analysis system 118 or the image processing system 102 of FIG. 1 before being processed by the one or more facial paralysis models 302. For example, one or more computational techniques can be applied to the training data 304 and the image data 106 to generate a modified database, which is then processed according to one or more machine learning techniques implemented by the one or more facial paralysis models 302. If at least one of the training data 304 or the image data 106 includes three-dimensional volumetric data from a CT-based imaging device or an MR-based imaging device, three-dimensional rendering can be performed. In one or more instances, a binary mask can be generated from an image included in at least one of the training data 304 or the image data 106. The binary mask can indicate voxels of the image, which can be foreground voxels and background voxels. Additionally, the binary mask can be used to generate a mesh corresponding to the surface of an object included in the image, such as the face of an individual corresponding to the image. The mesh can be rendered to generate an image of the individual's face that can be analyzed by one or more facial paralysis models 302. In various examples, the mesh can include a polygonal mesh. Additionally, the 3D rendering can include volumetric and surface rendering techniques. In one or more additional examples, preprocessing can be performed without generating a binary mask.

[0038] The training data 304 and images obtained from the image data 106 can also be processed to identify target regions of the individual, such as target regions 314, 316, 318, 320, and 322. In one or more examples, the classification data 308 can identify the target regions. In this manner, one or more facial paralysis models 302 can be trained to identify target regions of the individual. In one or more additional examples, a filter can be generated and used to identify target regions of the individual. The filter that identifies the target regions of the individual can be generated based on the training data 304. In various examples, the facial paralysis analysis system 118 can identify that the target regions 314, 316, 318, 320, and 322 of the individual 108 are indistinguishable or distorted. In such a situation, the facial paralysis analysis system 118 can provide an indication that the facial paralysis analysis system 118 is unable to analyze the image data 106. In one or more instances, at least one eye of the individual 108 may be blocked by an oxygen mask, glass, or other object.

[0039] Additionally, at least one of the training data 304 or the image data 106 can be augmented before being processed by the one or more facial paralysis models 302. In one or more examples, volumetric data obtained from an MR-based imaging device or a CT-based imaging device can be used to generate multiple images corresponding to multiple viewpoints. In this manner, features of a target region of an individual can be analyzed from multiple viewpoints. In one or more additional examples, one or more viewpoints providing images of at least a portion of a target region of an individual can be identified and processed by the one or more facial paralysis models 302 to improve the accuracy of the facial paralysis metrics 120. For example, one or more images associated with a viewpoint showing an individual facing and looking forward can be selected for processing by the one or more facial paralysis models 302, rather than images at a viewpoint where the individual is looking to one side or where the individual is not looking forward. In one or more other examples, at least one of the training data 304 or the image data 106 can include multiple images of an individual corresponding to different viewpoints. By way of example, one or more cameras of a computing device may be used to capture images of an individual from various viewpoints (relative to the orientation of the computing device), such as one or more right-side viewpoints, one or more left-side viewpoints, or a forward-facing viewpoint. In these examples, images corresponding to viewpoints facing directly forward may be selected for analysis by one or more facial paralysis models 302. In various examples, a computing device used to capture image data 106 for an individual 108 and / or capture image training data 306 may include user interface guides, such as outlines, that allow a user of the computing device to align the individual's face with the user interface guides. In this manner, the image training data 306 and the image data 106 may be standardized, resulting in more accurate and efficient processing of the image data 106 by the facial paralysis analysis system 118.

[0040] FIG. 4 illustrates a flowchart of a process for analyzing an image of an individual's face to assess whether the individual has suffered a stroke. The process may be embodied in a computer-readable medium and executed by one or more processors, whereby the operations of the process may be performed, in whole or in part, by functional components of the image processing system 102. Accordingly, in certain circumstances, the process described below is illustrative thereof. However, in other embodiments, at least some of the operations of the process described with respect to FIG. 4 may be deployed in various other hardware configurations. The process described with respect to FIG. 4 is therefore not limited to the image processing system 102 and may be performed, in whole or in part, by one or more additional components. While the illustrated flowchart depicts operations as a sequential process, many of the operations may occur in parallel or simultaneously. Additionally, the order of operations may be reversed. A process concludes when the operations are completed. A process may correspond to a method, a procedure, an algorithm, or the like. The operations of the method may be performed, in whole or in part, with some or all of the operations of other methods and may be performed by many different systems, such as the systems described herein or any part thereof, such as a processor included in any of the systems.

[0041] FIG. 4 is a flowchart illustrating example operations of a process 400 for analyzing an individual's amount of gaze deviation and facial paralysis to identify a probability that the individual is suffering from a stroke, according to one or more exemplary embodiments. At operation 402, process 400 may include acquiring image data from one or more image sources. The image data may correspond to at least a portion of the individual's face. The one or more image sources may include an MR-based imaging device or a CT-based imaging device. The CT-based imaging device, in one or more examples, may include a CT angiogram imaging device or a CT perfusion imaging device. Additionally, the one or more image sources may include one or more computing devices, such as a smartphone, a wearable device, a mobile computing device, a laptop computing device, or a tablet computing device. In these situations, the image data may be captured by one or more cameras of the one or more computing devices. In various examples, the image data may be captured using a stereo camera system of the one or more computing devices. In one or more additional examples, the one or more computing devices may generate a structured light pattern and capture one or more images of the individual in response to the structured light pattern. The one or more image sources may also include a time-of-flight camera, a LiDAR system, and / or a stereo vision camera.

[0042] At operation 404, process 400 can include determining a gaze deviation metric for the individual based on the image data. The gaze deviation metric can indicate an offset of a position of at least one iris of the individual relative to an expected iris position. Additionally, process 400 can include determining a facial paralysis metric for the individual based on the image data at operation 406. The facial paralysis metric can indicate a difference in appearance of one or more target regions of the individual's face relative to an expected appearance of the one or more target regions.

[0043] In various examples, one or more machine learning techniques can be implemented to identify gaze deviation metrics and facial paralysis metrics. In one or more examples, one or more convolutional neural networks can be trained to identify gaze deviation metrics and facial paralysis metrics. The training data for the one or more convolutional neural networks can include a first set of images from first individuals experiencing a biological condition that caused an amount of gaze deviation. The training data for the one or more convolutional neural networks can also include a second set of images from second individuals experiencing a biological condition that caused an amount of facial paralysis. Furthermore, the training data for the one or more convolutional neural networks can include a third set of images from third individuals who did not experience the biological condition, did not experience the amount of gaze deviation, and did not experience the amount of facial paralysis. The training data can also include additional sets of images from an additional group of individuals experiencing both gaze deviation and facial paralysis. The training data may also include characteristics of the individuals, such as ethnic background, age, gender, whether or not they wear eyeglasses, whether or not they have facial hair, and so forth. In one or more other examples, the training data may be obtained from the same individual, captured at a time prior to the individual's development of a biological condition. In one or more examples, the training data may include classification data indicating images with gaze deviation, images with facial paralysis, and images with neither gaze deviation nor facial paralysis. In one or more examples, the output of the one or more convolutional networks may include multiple categories of gaze deviation and multiple categories of facial paralysis. In one or more other examples, the one or more convolutional neural networks may be trained to identify a feature set indicative of gaze deviation and a feature set indicative of facial paralysis. The feature sets may not include one or more target regions. In such a situation, one or more convolutional neural networks can analyze new patient images with respect to each feature set to determine the amount of similarity between the feature set and the new image, and thus the degree of gaze deviation and / or the degree of facial paralysis.

[0044] In one or more examples, training data can be identified from multiple viewpoints. The viewpoints can include images obtained from various positions on an individual. The number of viewpoints can also be identified in silico, such as by extracting MR or CT data for different viewpoints.

[0045] Process 400 may also include, at operation 408, analyzing at least one of the gaze deviation metric or the facial paralysis metric to identify an evaluation metric indicative of the probability of the presence of a biological condition for the individual. In various examples, an increase in the gaze deviation metric may increase the evaluation metric because the more pronounced the individual's gaze deviation, the higher the probability that the individual has suffered a stroke. Additionally, an increase in the facial paralysis metric may also increase the evaluation metric because an increase in the amount of facial paralysis increases the probability that the individual has a biological condition. The amount of gaze deviation and / or the amount of facial paralysis may also be an indicator of the severity of the biological condition suffered by the individual. In one or more examples, the biological condition may include a neurological condition, such as Bell's palsy or stroke.

[0046] Further, at operation 410, process 400 may include generating one or more user interfaces corresponding to the individual, including a user interface element that indicates a probability that the biological condition occurs for the individual. In one or more examples, the assessment metrics may be displayed as numerical probabilities within the user interface. In one or more additional examples, the assessment metrics may be displayed as categorical indicators within the user interface. The one or more user interfaces may also include an image of the individual's face, rendered based on the image data, for which the gaze deviation metrics and facial palsy metrics have been determined. In addition to the image of the individual's face, an image of the individual's brain may also be displayed. For example, an image showing the blood vessels of the individual's brain may also be displayed on the one or more user interfaces. In this manner, an image of an individual's external parts and an image of the individual's internal parts may be rendered from the same image data and used to determine a probability that the individual has a biological condition.

[0047] In one or more instances, images can be captured from an individual using a mobile computing device, such as an EMT or other medical professional. The screen of the mobile computing device can include an outline, and the image of the individual shown in the user interface can be aligned with the outline. In one or more instances, multiple images can be captured from different perspectives. In these situations, the image data captured by the mobile computing device can be analyzed to determine the amount of facial paralysis and / or the amount of gaze deviation of the individual. Additionally, an initial assessment of the individual can be determined based on the amount of gaze deviation and / or the amount of facial paralysis with respect to their physiological condition.

[0048] Figure 5 includes multiple images showing individuals with and without facial paralysis. Image A includes an image of the individual before the stroke, and image B includes an image of the same individual after the stroke, showing left-sided facial paralysis and a gaze shift of the left eye to the right. Image C includes a volumetric rendering of CT volumetric data from an 88-year-old individual with a right-sided proximal intracranial artery (ICA) occlusion, resulting in facial paralysis on the left side. Image D includes a volumetric rendering of CT volumetric data from an 82-year-old patient with an occlusion of the right upper M2 region of the middle cerebral artery (MCA), resulting in left-sided facial paralysis and a moderate conjugate gaze shift to the right. Image E includes a photograph of the individual's face superimposed on a surface rendering obtained with a 3D structured light camera.

[0049] Figure 6 contains images rendered using computed tomography data of an individual's face, with different images corresponding to different viewpoints of the individual's face. The images contained in Figure 6 are from nine different viewpoints and were rendered from CT volumetric data of an 80-year-old patient with an obstruction in the right distal portion of the ICA.

[0050] Figure 7 includes images showing the location of occluded vessels in individuals who have had a stroke and the correlation between gaze shift and the location of the occluded vessels. Image A illustrates the location of an exemplary patient whose stroke is in the right hemisphere of the brain and shows a corresponding conjugate gaze shift to the right. Image B includes a non-contrast CT image of an 80-year-old patient who has experienced a distal ICA occlusion, which causes a large intracerebral blood flow deficiency in the right middle cerebral artery territory, resulting in a conjugate gaze shift to the right. The gaze shift can be identified based on the non-contrast CT data. Image C includes an exemplary eye extracted by segmentation from the non-contrast CT image and the corresponding lens position data.

[0051] FIG. 8 is a block diagram illustrating components of a machine 800 according to some example embodiments that can read instructions from a machine-readable medium (e.g., a machine-readable storage medium) and perform any one or more of the methodologies discussed herein. Specifically, FIG. 8 shows a schematic diagram of machine 800, an example form of computer system, within which instructions 802 (e.g., software, programs, applications, applets, apps, or other executable code) can be executed to cause machine 800 to perform any one or more of the methodologies discussed herein. As such, instructions 802 can be used to implement modules or components described herein. Instructions 802 transform a general-purpose, unprogrammed machine 800 into a specific machine 800 programmed to perform the functions described and illustrated herein in the described manner. In alternative embodiments, machine 800 can operate as a standalone device or be coupled (e.g., networked) to other machines. In a networked deployment, machine 800 may operate as a server or client machine in a server-client network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. Machine 800 may include, but is not limited to, a server computer, a client computer, a personal computer (PC), a tablet computer, a laptop computer, a netbook, a set-top box (STB), a personal digital assistant (PDA), an entertainment media system, a mobile phone, a smartphone, a mobile device, a wearable device (e.g., a smart watch), a smart home device (e.g., a smart appliance), other smart devices, a web appliance, a network router, a network switch, a network bridge, or any machine capable of executing, sequentially or otherwise, instructions 802 specifying operations to be performed by machine 800. Furthermore, although only one machine 800 is shown, the term "machine" should be interpreted to include a collection of machines that individually or collectively execute instructions 802 to perform any one or more of the methods discussed herein.

[0052] Machine 800 may include processor 804, memory / storage 806, and I / O components 808, which may be configured to communicate with each other via, for example, bus 810. A “processor,” in this regard, refers to any circuit or virtual circuit (a physical circuit emulated by logic executing on actual processor 804) that manipulates data values ​​in accordance with control signals (e.g., “commands,” “instruction code,” “machine code,” etc.) and generates corresponding output signals that are applied to operate machine 800. In an exemplary embodiment, processor 804 (e.g., a central processing unit (CPU), a reduced instruction set computing (RISC) processor, a complex instruction set computing (CISC) processor, a graphics processing unit (GPU), a digital signal processor (DSP), an application specific integrated circuit (ASIC), a radio frequency integrated circuit (RFIC), other processor, or any suitable combination thereof) may include, for example, processor 812 and processor 814, which may execute instructions 802. The term "processor" is intended to include a multi-core processor 804, which may include two or more independent processors (sometimes referred to as "cores") that may simultaneously execute instructions 802. While Figure 8 shows multiple processors 804, machine 800 may include a single processor 812 with a single core, a single processor 812 with multiple cores (e.g., a multi-core processor), multiple processors 812, 814 with a single core, multiple processors 812, 814 with multiple cores, or any combination thereof.

[0053] Memory / storage 806 may include memory, such as main memory 816 or other memory storage, and storage unit 818, both of which are accessible to processor 804, for example, via bus 810. Storage unit 818 and main memory 816 store instructions 802 that embody any one or more of the methods or functions described herein. Instructions 802 may also reside, completely or partially, in main memory 816, in storage unit 818, in at least one of processors 804 (e.g., in a processor's cache memory), or any suitable combination thereof, during execution thereof by machine 800. Thus, main memory 816, storage unit 818, and memory of processor 804 are examples of machine-readable media. A "machine-readable medium," also referred to herein as a "computer-readable storage medium," in this regard, refers to a component, device, or other tangible medium capable of temporarily or permanently storing instructions 802 and data, and may include, but is not limited to, random access memory (RAM), read-only memory (ROM), buffer memory, flash memory, optical media, magnetic media, cache memory, other types of storage (e.g., erasable programmable read-only memory (EEPROM)), and / or any suitable combination thereof. The term "machine-readable medium" may be interpreted to include a medium or multiple media (e.g., centralized or distributed databases, or associated caches and servers) capable of storing instructions 802. The term "machine-readable medium" should also be interpreted to include any medium or combination of media capable of storing instructions 802 (e.g., code) for execution by machine 800, which, when executed by one or more processors 804 of machine 800, cause machine 800 to perform any one or more of the methods described herein. Thus, the term "machine-readable medium" refers to a single storage device or device, as well as a "cloud-based" storage system or storage network that includes multiple storage devices or devices. The term "machine-readable medium" does not include the signal itself.

[0054] I / O components 808 may include various components for receiving input, providing output, generating output, transmitting information, exchanging information, obtaining measurements, and the like. The specific I / O components 808 included in a particular machine 800 will depend on the type of machine. For example, a portable machine such as a mobile phone will likely include a touch-based input device or other such input mechanism, while a headless server machine will likely not include such a touch-based input device. It should be understood that I / O components 808 may include many other components not shown in FIG. 8 . I / O components 808 are categorized according to functionality solely to simplify the following description, and this categorization is not intended to be limiting in any way. In various exemplary embodiments, I / O components 808 may include a user output component 820 and a user input component 822. User output components 820 may include visual components (e.g., displays such as plasma display panels (PDPs), light-emitting diode (LED) displays, liquid crystal displays (LCDs), projectors, cathode ray tubes (CRTs), etc.), acoustic components (e.g., speakers), tactile components (e.g., vibration motors, resistance mechanisms), other signal generators, etc. User input components 822 may include alphanumeric input components (e.g., keyboards, touchscreens configured to receive alphanumeric input, optical keyboards, or other alphanumeric input components), pointer-type input components (e.g., mice, touchpads, trackballs, joysticks, motion sensors, or other pointing instructions), tactile input components (e.g., physical buttons, touchscreens that provide the position or force of touch or touch gestures, or other tactile input components), audio input components (e.g., microphones), etc.

[0055] In other exemplary embodiments, I / O component 808 may include a biometric component 824, a motion component 826, an environmental component 828, or a position component 830, among a wide range of other components. For example, biometric component 824 may detect facial expressions (e.g., hand expressions, facial expressions, vocal indicators, gestures, or eye tracking), measure biometric signals (e.g., blood pressure, heart rate, body temperature, sweat, or brain waves), authenticate a person (e.g., voice recognition, retinal recognition, facial recognition, fingerprint recognition, or electroencephalogram-based authentication), etc. Motion component 826 may include an acceleration sensor component (e.g., an accelerometer), a gravity sensor component, a rotation sensor component (e.g., a gyroscope), etc. The environmental components 828 may include, for example, a lighting sensor component (e.g., a light meter), a temperature sensor component (e.g., one or more thermometers that detect ambient temperature), a humidity sensor component, a pressure sensor component (e.g., a barometer), an acoustic sensor component (e.g., one or more microphones that detect background noise), a proximity sensor component (e.g., an infrared sensor that detects nearby objects), a gas sensor (e.g., a gas detection sensor that detects the concentration of toxic gases for safety purposes or measures pollutants in the air), or other components that may provide indicators, measurements, or signals corresponding to the surrounding physical environment. The location component 830 may include a location sensor component (e.g., a GPS receiver component), an altitude sensor component (e.g., an altimeter or barometer that can detect air pressure and derive altitude therefrom), a three-dimensional position and orientation sensor component (e.g., a magnetometer), etc.

[0056] Communications may be implemented using a variety of technologies. The I / O component 808 may include a communications component 832 operable to couple the machine 800 to a network 834 or a device 836. For example, the communications component 832 may include a network interface component or other device suitable for interfacing with the network 834. In another example, the communications component 832 may include a wired communications component, a wireless communications component, a cellular communications component, a near field communications (NFC) component, a Bluetooth® component (e.g., Bluetooth® Low Energy), a Wi-Fi® component, and other communications components that provide communications via other modalities. The device 836 may be another machine 800 or any of a wide range of peripheral devices (e.g., a peripheral device coupled via USB).

[0057] Further, communication component 832 may include a component that detects or is operable to detect an identifier. For example, communication component 832 may include a radio frequency identification (RFID) tag reader component, an NFC smart tag detection component, an optical reader component (e.g., an optical sensor that detects one-dimensional barcodes such as Universal Product Code (UPC) barcodes, multi-dimensional codes such as QR Code, Aztec Code, Data Matrix, Dataglyph, MaxiCode, PDF417, Ultra Code, UCC RSS-2D barcodes, and other optical codes), or an acoustic detection component (e.g., a microphone for identifying tagged audio signals). Additionally, various information may be derived via communication component 832, such as location via Internet Protocol (IP) geolocation, location via Wi-Fi signal triangulation, location via detection of NFC beacon signals that may indicate a specific location, etc.

[0058] As used herein, a "component" refers to a device, physical entity, or logic having boundaries defined by function or subroutine calls, branch points, APIs, or other techniques that provide partitioning or modularization of specific processing or control functions. A component may perform machine processing in combination with other components via their interfaces. A component may be a packaged functional hardware unit designed for use with other components and part of a program that performs specific functions, usually related functions. A component may constitute either a software component (e.g., code embedded on a machine-readable medium) or a hardware component. A "hardware component" is a tangible unit capable of performing specific operations and may be configured or arranged in a specific physical manner. In various exemplary embodiments, one or more computer systems (e.g., a stand-alone computer system, a client computer system, or a server computer system) or one or more hardware components of a computer system (e.g., a processor or processors) may be configured as a hardware component that operates by software (e.g., an application or application portion) to perform specific operations as described herein.

[0059] A hardware component may also be implemented mechanically, electronically, or any suitable combination thereof. For example, a hardware component may include dedicated circuitry or logic permanently configured to perform specific operations. A hardware component may be a special-purpose processor, such as a field programmable gate array (FPGA) or an ASIC. A hardware component may also include programmable logic or circuitry temporarily configured by software to perform specific operations. For example, a hardware component may include software executed by a general-purpose processor 804 or other programmable processor. Once configured by such software, the hardware component becomes a specific machine (or a specific component of machine 800) uniquely adapted to perform the function for which it is configured, and is no longer a general-purpose processor 804. It should be understood that the decision whether to implement a hardware component mechanically, with dedicated permanently configured circuitry, or with temporarily configured circuitry (e.g., configured by software) may be made based on cost and time considerations. Thus, the term "hardware component" (or "hardware-implemented component") should be understood to include a tangible entity that is physically constructed and permanently configured (e.g., hardwired) or temporarily configured (e.g., programmed) to operate in a particular manner or perform such particular operations as described herein. When contemplated are embodiments in which hardware components are temporarily configured (e.g., programmed), it is not necessary that each of the hardware components be configured or instantiated at any one time. For example, if the hardware components include a general-purpose processor 804 that is configured by software to be a special-purpose processor, then the general-purpose processor 804 may be configured as different special-purpose processors (e.g., including different hardware components) at different times.The software may thus configure a particular processor 812, 814, or processor 804, for example, to implement particular hardware components at one time and different hardware components at another time.

[0060] Hardware components can provide information to and receive information from other hardware components. Thus, the hardware components described herein may be considered to be communicatively coupled. When multiple hardware components are present simultaneously, communication may be achieved through signal transmission (e.g., using appropriate circuits and buses) between two or more of these hardware components. In embodiments in which multiple hardware components are configured or instantiated at different times, communication between such hardware components may be achieved, for example, through the storage and retrieval of information in memory structures accessible to the multiple hardware components. For example, one hardware component may perform an operation and store the output of that operation in a memory device to which it is communicatively coupled. Another hardware component may then access the memory device at a later time to retrieve and process the stored output.

[0061] Hardware components may also initiate communications with input or output devices and operate on resources (e.g., collect information). Various operations of the example methods described herein may be performed, at least in part, by one or more processors 804 that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors 804 may constitute processor-implemented components that operate to perform one or more operations or functions described herein. As used herein, "processor-implemented components" refers to hardware components implemented using one or more processors 804. Similarly, methods described herein may be at least partially processor-implemented, with a particular processor 812, 814, or processor 804 being an example of hardware. For example, at least some of the operations of a method may be performed by one or more processors 804 or processor-implemented components. Additionally, one or more processors 804 may also operate to assist in performing the relevant operations in a "cloud computing" environment or as "software as a service" (SaaS). For example, at least some of the operations may be performed by a cluster of computers (e.g., machine 800 including processor 804), and these operations may be performed via a network 834 (e.g., the Internet) and via one or more appropriate interfaces (e.g., APIs). Execution of certain of the operations may be distributed among processors, which may be located within one machine 800 as well as spread across multiple machines. In some exemplary embodiments, processor 804 or processor-implemented components may be located in one geographic location (e.g., in a home environment, an office environment, or a server farm). In other exemplary embodiments, processor 804 or processor-implemented components may be distributed across multiple geographic locations.

[0062] FIG. 9 is a block diagram illustrating a system 900 including an exemplary software architecture 902 that can be used with various hardware architectures described herein. FIG. 9 is a non-limiting example of a software architecture; many other architectures can be implemented to facilitate the functionality described herein. The software architecture 902 can be executed on hardware, such as machine 800 of FIG. 8, which can include, among other things, a processor 804, memory / storage 806, and input / output (I / O) components 808. A representative hardware layer 904 is illustrated and can represent, for example, machine 800 of FIG. 8. The representative hardware layer 904 includes a processing unit 906 having associated executable instructions 908. The executable instructions 908 represent executable instructions for the software architecture 902, including methods, components, and other implementations described herein. The hardware layer 904 also includes at least one memory / storage 910 of a memory or storage module, also having executable instructions 908. The hardware layer 904 can also include other hardware 912.

[0063] In the example architecture of FIG. 9 , software architecture 902 may be conceptualized as a layered structure, with each layer providing specific functionality. For example, software architecture 902 may include layers such as operating system 914, libraries 916, framework / middleware 918, application 920, and presentation layer 922. In operation, application 920 or other components within a layer may invoke API calls 924 and receive messages 926 in response to API calls 924 through the software stack. The illustrated layers are representative in nature, and not all software architectures have all layers. For example, some mobile or application-specific operating systems may not provide framework / middleware 918, while others may provide such a layer. Other software architectures may include additional or different layers.

[0064] The operating system 914 may manage hardware resources and provide common services. The operating system 914 may include, for example, a kernel 928, services 930, and drivers 932. The kernel 928 may act as an abstraction layer between the hardware and other software layers. For example, the kernel 928 may be responsible for memory management, processor management (e.g., scheduling), component management, networking, security configuration, and the like. The services 930 may provide other common services for the other software layers. The drivers 932 are responsible for controlling or interfacing with the underlying hardware. For example, the drivers 932 may include a display driver, a camera driver, a Bluetooth driver, a flash memory driver, a serial communications driver (e.g., a Universal Serial Bus (USB) driver), a Wi-Fi driver, an audio driver, a power management driver, and the like, depending on the hardware configuration.

[0065] Libraries 916 provide a common infrastructure used by applications 920 and / or other components or layers. Libraries 916 provide functionality that allows other software components to perform tasks more easily than if they directly interfaced with the underlying operating system 914 functionality (e.g., kernel 928, services 930, drivers 932). Libraries 916 may include system libraries 934 (e.g., standard C libraries), which may provide memory allocation functions, string manipulation functions, mathematical functions, and other functions. In addition, libraries 916 may include API libraries 936, such as a media library (e.g., a library that supports the presentation and manipulation of various media formats, such as MPEG4, H.264, MP3, AAC, AMR, JPG, PNG, etc.), a graphics library (e.g., an OpenGL framework that can be used to render graphical content on a display in two and three dimensions), a database library (e.g., SQLite, which provides various relational database functions), a web library (e.g., WebKit, which can provide web browsing functionality), and so forth. Libraries 916 may also include various other libraries 938, which provide many other APIs to applications 920 and other software components / modules.

[0066] Frameworks / middleware 918 (sometimes called middleware) provide a higher-level common infrastructure that can be used by applications 920 or other software components / modules. For example, frameworks / middleware 918 can provide various graphical user interface functionality, high-level resource management, high-level location services, and so on. Frameworks / middleware 918 can also provide a wide range of other APIs that can be utilized by applications 920 or other software components / modules, some of which may be specific to a particular operating system 914 or platform.

[0067] Applications 920 include built-in applications 940 and third-party applications 942. Examples of representative built-in applications 940 may include, but are not limited to, a contacts application, a browser application, a book reader application, a location application, a media application, a messaging application, or a game application. Third-party applications 942 may include applications developed using the Android® or IOS™ Software Development Kit (SDK) by an entity other than the specific platform vendor and may be mobile software running on a mobile operating system such as IOS™, Android®, WINDOWS® Phone, or other mobile operating system. Third-party applications 942 may invoke API calls 924 provided by the mobile operating system (e.g., operating system 914) to facilitate the functionality described herein.

[0068] Applications 920 may use built-in operating system functionality (e.g., kernel 928, services 930, drivers 932), libraries 916, and frameworks / middleware 918 to create a user interface for interacting with a user of the system. Alternatively, or additionally, in some systems, user interaction may occur through a presentation layer, such as presentation layer 922. In these systems, application / component "logic" can be separated from the aspects of the application / component that interact with the user.

[0069] Exemplary Aspects of the Disclosure Aspect 1. A method, comprising: by a computing system including one or more processing units and one or more memory devices, acquiring image data from one or more image data sources, the image data corresponding to at least a portion of an individual's face; determining, by the computing system, gaze deviation metrics for the individual based on the image data, the gaze deviation metrics indicating a deviation of a position of at least one of an iris or a pupil of the individual from an expected iris or pupil position; and determining, by the computing system, facial paralysis metrics for the individual based on the image data. identifying facial paralysis metrics indicating a difference between an appearance of one or more target regions of the individual's face and an expected appearance of the one or more target regions; analyzing, with the computing system, at least one of the gaze deviation metrics or the facial paralysis metrics to identify an evaluation metric indicative of a probability that a biological condition exists for the individual; and generating, with the computing system, user interface data corresponding to one or more user interfaces including a user interface element indicative of the probability that the biological condition exists for the individual.

[0070] Aspect 2. The method of Aspect 1, wherein the biological condition includes a neurological condition, the method including acquiring, by a computing system, training data including image training data, the image training data including: a first set of images of first individuals experiencing a biological condition causing an amount of gaze deviation; a second set of images of second individuals experiencing a biological condition causing an amount of facial paralysis; a third set of images of third individuals experiencing gaze deviation and facial paralysis in response to the biological condition; and a fourth set of images of fourth individuals not experiencing a biological condition, not experiencing the amount of gaze deviation, and not experiencing the amount of facial paralysis.

[0071] Aspect 3. The method of aspect 2, comprising: analyzing, by the computing system, the second set of images and the third set of images to identify a plurality of regions of the faces of individuals having different characteristics between the second individuals in whom the biological condition occurred and the fourth individuals in whom the biological condition did not occur; and identifying, by the computing system, one or more target regions from the plurality of regions.

[0072] Aspect 4. The method of aspect 2, further comprising: analyzing, by the computing system, the first set of images and the third set of images to identify expected gazes corresponding to the third individuals in whom the biological condition did not occur; and analyzing, by the computing system, the second set of images and the third set of images to identify expected appearances of the one or more target regions for the third individuals in whom the biological condition did not occur.

[0073] Aspect 5. The method of Aspect 2, wherein the training data includes classification data, the classification data indicating the first set of images as having been obtained from the first individuals experiencing a biological condition causing a certain amount of gaze deviation, indicating the second set of images as having been obtained from second individuals experiencing a biological condition causing a certain amount of facial paralysis, indicating the third set of images as having been obtained from third individuals experiencing a biological condition causing facial paralysis and gaze deviation, and indicating the fourth set of images as having been obtained from fourth individuals not experiencing a biological condition.

[0074] Aspect 6. The method of aspect 2, comprising: training, by the computing system, one or more convolutional neural networks using the training data to identify the facial paralysis metrics and to identify the gaze deviation metrics.

[0075] Example 7. The method of example 6, wherein inputs to the one or more convolutional neural networks during training include the first set of images having classification data indicating that the first individuals experienced a biological condition causing a certain amount of gaze deviation, and the fourth set of images having additional classification data indicating that the fourth individuals did not experience a certain amount of gaze deviation, and the one or more convolutional neural networks are trained to provide outputs indicative of a plurality of categories, each category of the plurality of categories corresponding to a different amount of gaze deviation.

[0076] Example 8. The method of Example 6, wherein inputs to the one or more convolutional neural networks during training include the second set of images having classification data indicating that the second individuals experienced a biological condition causing an amount of facial paralysis, and the fourth set of images having additional classification data indicating that the fourth individuals did not experience an amount of facial paralysis, and wherein the one or more convolutional neural networks are trained to provide an output indicative of a plurality of categories, each category of the plurality of categories corresponding to a different amount of facial paralysis.

[0077] Aspect 9. The method of any one of aspects 1 to 8, further comprising: determining, by a computing system, the location of an occluded blood vessel in the individual based on at least one of the gaze deviation metrics or the facial paralysis metrics.

[0078] Aspect 10. The method of any one of Aspects 1-9, wherein the one or more image sources include a computed tomography-based imaging device or a magnetic resonance-based imaging device.

[0079] Aspect 11. The method of any one of aspects 1 to 10, wherein the one or more image sources include a mobile computing device, and the image data includes one or more images captured by the mobile computing device using a structured light pattern, using a time of flight (TOF) camera, or using a LiDAR system.

[0080] Aspect 12. A system, comprising: one or more hardware processors; and one or more non-transitory computer-readable storage media comprising computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations, including acquiring image data from one or more image sources, the image data comprising at least a portion of an individual's face; and determining gaze deviation metrics for the individual based on the image data, the gaze deviation metrics representing an expected iris or pupil position of at least one of the individual's iris or pupil. and determining facial paralysis metrics for the individual based on the image data, the facial paralysis metrics indicating a difference between an appearance of one or more target regions of the individual's face and an expected appearance of the one or more target regions; analyzing at least one of the gaze deviation metrics or the facial paralysis metrics to determine an evaluation metric indicative of a probability that the biological condition exists for the individual; and generating user interface data corresponding to one or more user interfaces including a user interface element indicative of the probability that the biological condition exists for the individual.

[0081] Aspect 13. The system of claim 12, wherein the image data is obtained from a magnetic resonance-based imaging device or a computed tomography-based imaging device, and wherein the one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations, the additional operations including generating a first image showing blood vessels of the individual's brain based on the image data and generating a second image including facial features of the individual based on the image data, and wherein a user interface of the one or more user interfaces includes the first image and the second image.

[0082] Aspect 14. The system of claim 12, wherein the one or more target regions include a first plurality of target regions on a first side of the individual's face and a second plurality of target regions on a second side of the individual's face, the first side of the face being opposite to the second side of the face, and the one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations, the additional operations including determining the expected appearance of the one or more target regions based on a first characteristic of the first plurality of target regions, and the facial paralysis metric is determined based on a difference between the first characteristic of the first plurality of target regions and a second characteristic of the second plurality of target regions.

[0083] Aspect 15. The system of claim 12, wherein the one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations, including obtaining a reference image of the individual captured prior to an incident suspected of causing a biological condition for the individual; determining the expected iris position based on a first position of a first iris or pupil of the individual in the reference image and a second position of a second iris or pupil of the individual in the reference image; and determining the expected appearance of the one or more target regions based on the appearance of one or more target regions in the reference image.

[0084] Aspect 16. The system of claim 12, wherein the one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations, the additional operations including determining an angular offset between a position of the individual's iris or pupil determined based on the image data and the expected position of the iris or pupil, and the gaze deviation metric is based on the angular offset.

[0085] Aspect 17. The system of claim 12, wherein the one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations, including generating multiple images of the individual's face from multiple viewpoints based on the image data, and at least one of the gaze deviation metrics or the facial paralysis metrics is determined using a first image of the multiple images corresponding to a first viewpoint and a second image of the multiple images corresponding to a second viewpoint.

[0086] Aspect 18. The system of claim 12, wherein the one or more non-transitory computer-readable sources include additional computer-readable instructions that, when executed by the one or more processing devices, cause the one or more processing devices to perform additional operations, the additional operations including identifying a binary mask indicating a first portion of the image data corresponding to a foreground and a second portion of the image data corresponding to a background, generating a mesh corresponding to a surface of the face of the individual, and rendering the mesh to generate an image of the face of the individual.

[0087] Aspect 19. The system of claim 12, wherein the one or more non-transitory computer-readable media include additional computer-readable instructions that, when executed by the one or more processing devices, cause the one or more processing devices to perform additional operations, the additional operations including performing one or more segmentation processes on the image data to identify the individual's eyes, determining that a viewpoint of an image included in the image data is different from a reference viewpoint, and correcting a first position of a first iris of the individual and a second position of a second iris of the individual to correct for a difference between the viewpoint of the image and the reference viewpoint.

[0088] Aspect 20. One or more non-transitory computer-readable media that, when executed by one or more processing devices, cause the one or more processing devices to perform operations, the operations including: acquiring image data from one or more imaging sources; the image data including at least a portion of an individual's face; determining gaze deviation metrics for the individual based on the image data; the gaze deviation metrics indicating an offset of a position of at least one iris or pupil of the individual relative to an expected position of the iris or pupil; and determining a gaze deviation metric for the individual based on the image data. determining the facial paralysis metrics, wherein the facial paralysis metrics indicate a difference in appearance of one or more target regions of the face of the individual from an expected appearance of the one or more target regions; analyzing at least one of the gaze deviation metrics or the facial paralysis metrics to determine an evaluation metric indicative of a probability that a biological condition exists for the individual; and generating user interface data corresponding to one or more user interfaces including a user interface element indicative of the probability that the biological condition exists for the individual.

[0089] Aspect 21. The one or more non-transitory computer-readable media of claim 18 storing additional computer-readable instructions that, when executed by the one or more processing devices, cause the one or more processing devices to perform additional operations, the additional operations including: identifying a binary mask indicating a first portion of the image data corresponding to a foreground and a second portion of the image data corresponding to a background; generating a mesh of polygons corresponding to a surface of the face of the individual; and rendering the mesh of polygons to generate an image of the face of the individual.

[0090] Aspect 22. One or more non-transitory computer-readable media as described in claim 18 storing additional computer-readable instructions that, when executed by the one or more processing devices, cause the one or more processing devices to perform additional operations, the additional operations including performing one or more segmentation processes on the image data to identify the eyes of the individual, determining that a viewpoint of an image included in the image data is different from a reference viewpoint, and correcting a first position of a first iris of the individual and a second position of a second iris of the individual to correct for the difference between the viewpoint of the image and the reference viewpoint.

[0091] Changes and modifications may be made to the disclosed embodiments without departing from the scope of the disclosure. These and other changes and modifications are intended to be included within the scope of the disclosure as set forth in the following claims.

Claims

1. 1. A method comprising: acquiring, by a computing system including one or more processing units and one or more memory devices, image data corresponding to at least a portion of the individual's face from one or more image data sources, the image data including three-dimensional volumetric data obtained from a magnetic resonance based imaging device or a computed tomography based imaging device; performing, with the computing system, a three-dimensional rendering of the image data by identifying a binary mask indicative of a first portion of the imaging data corresponding to a foreground and a second portion of the imaging data corresponding to a background, generating a mesh corresponding to the face of the individual based on the binary mask, and rendering the mesh to generate an image of the face of the individual; determining, by the computing system, gaze deviation metrics for the individual based on the image data, the gaze deviation metrics indicating a deviation of a position of at least one of an iris or a pupil of the individual from an expected iris or pupil position; determining, by the computing system, facial paralysis metrics for the individual based on the image data, the facial paralysis metrics indicating a difference between an appearance of one or more target regions of the individual's face and an expected appearance of the one or more target regions; analyzing, with the computing system, at least one of the gaze deviation metrics or the facial paralysis metrics to identify an assessment metric indicative of a probability that a biological condition exists for the individual; generating, by the computing system, user interface data corresponding to one or more user interfaces including a user interface element indicative of a probability of the biological condition being present for the individual; A method comprising:

2. 10. The method of claim 1, wherein the biological condition is a neurological condition, the method comprising: and acquiring, by the computing system, training data including image training data, the image training data comprising: a first set of images of a first individual experiencing the neurological condition causing an amount of gaze deviation; a second set of images of second individuals experiencing the neurological condition causing an amount of facial paralysis; a third set of images of a third individual experiencing the neurological condition causing an amount of facial paralysis and an amount of gaze deviation; a fourth set of images of a fourth individual who is not experiencing the neurological condition, who does not have a certain amount of gaze deviation, and who does not have a certain amount of facial paralysis; A method comprising:

3. 3. The method of claim 2, analyzing, with the computing system, the second set of images and the fourth set of images to identify regions of faces of individuals having different characteristics between the second individuals with the neurological condition and the fourth individuals without the neurological condition; identifying, by the computing system, one or more target regions from the plurality of regions; A method comprising:

4. 3. The method of claim 2, analyzing, with the computing system, the first set of images and the fourth set of images to identify expected gazes corresponding to the fourth individuals in whom the neurological condition did not occur; analyzing, with the computing system, the second set of images and the third set of images to identify an expected appearance of the one or more target regions for the fourth individuals who were not affected by the neurological condition; A method comprising:

5. 3. The method of claim 2, wherein the training data includes classification data that indicates the first set of images as being from the first individuals experiencing the neurological condition causing an amount of gaze deviation, the second set of images as being from the second individuals experiencing the neurological condition causing an amount of facial paralysis, the third set of images as being from the third individuals experiencing the neurological condition, and the fourth set of images as being from the fourth individuals not experiencing the neurological condition.

6. 3. The method of claim 2, training, by the computing system, one or more convolutional neural networks using the training data to identify the facial paralysis metrics and to identify the gaze deviation metrics.

7. 7. The method of claim 6, inputs to the one or more convolutional neural networks during training include the first set of images with classification data indicating that the neurological condition causing an amount of gaze deviation occurred in the first individuals, and the fourth set of images with additional classification data indicating that the third individuals did not have an amount of gaze deviation; 10. The method of claim 1, wherein the one or more convolutional neural networks are trained to provide outputs indicative of a plurality of categories, each category of the plurality of categories corresponding to a different amount of gaze deviation.

8. 7. The method of claim 6, inputs to the one or more convolutional neural networks during training include the second set of images with classification data indicative of the second individuals experiencing the neurological condition causing an amount of facial paralysis, and the third set of images with additional classification data indicative of the third individuals not experiencing an amount of facial paralysis; 10. A method according to claim 1, wherein the one or more convolutional neural networks are trained to provide an output indicative of a plurality of categories, each category of the plurality of categories corresponding to a different amount of facial paralysis.

9. 10. The method of claim 1, and determining, by the computing system, a location of an occluded blood vessel in the individual based on at least one of the gaze deviation metric or the facial palsy metric.

10. The method of claim 1 , wherein the one or more image sources include a computed tomography-based imaging device or a magnetic resonance-based imaging device.

11. 10. The method of claim 1, wherein the one or more image sources include a mobile computing device, and the image data includes one or more images captured by the mobile computing device using a structured light pattern, using a time-of-flight camera, or using a LiDAR system.

12. 1. A system comprising: one or more hardware processors; one or more non-transitory computer-readable storage media containing computer-readable instructions; Including, The computer-readable instructions, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations, such as: acquiring image data comprising at least a portion of an individual's face from one or more imaging sources, the image data including three-dimensional volumetric data obtained from a magnetic resonance based imaging device or a computed tomography based imaging device; performing a three-dimensional rendering of the image data by identifying a binary mask indicative of a first portion of the imaging data corresponding to a foreground and a second portion of the imaging data corresponding to a background, generating a mesh corresponding to the face of the individual based on the binary mask, and rendering the mesh to generate an image of the face of the individual; determining a gaze deviation metric for the individual based on the image data, the gaze deviation metric indicating an offset of a position of at least one of an iris or a pupil of the individual relative to an expected iris or pupil position; determining facial paralysis metrics for the individual based on the image data, the facial paralysis metrics indicating a difference between an appearance of one or more target regions of the individual's face and an expected appearance of the one or more target regions; analyzing at least one of the gaze deviation metrics or the facial paralysis metrics to identify an assessment metric indicative of a probability that a biological condition exists for the individual; generating user interface data corresponding to one or more user interfaces including a user interface element indicative of the probability of the biological condition being present in the individual; Including, the system.

13. 13. The system of claim 12, The one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations, such as: generating a first image indicative of the vasculature of the brain of the individual based on the image data; generating a second image including facial features of the individual based on the image data; Including, A user interface of the one or more user interfaces includes the first image and the second image.

14. 13. The system of claim 12, the one or more target areas include a first plurality of target areas on a first side of the face of the individual and a second plurality of target areas on a second side of the face of the individual, the first side of the face being opposite to the second side of the face; The one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations, such as: determining the expected appearance of the one or more target regions based on a first characteristic of the first plurality of target regions; The facial paralysis metric is determined based on a difference between the first feature of the first plurality of target regions and a second feature of the second plurality of target regions.

15. 13. The system of claim 12, wherein the one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations, the additional operations including: obtaining a reference image of the individual captured prior to the suspected incidence of the biological condition with respect to the individual; determining the expected iris location based on a first location of a first iris or pupil of the individual in the reference image and a second location of a second iris or pupil of the individual in the reference image; determining the expected appearance of the one or more target regions based on the appearance of the one or more target regions in the reference image; Including, the system.

16. 13. The system of claim 12, wherein the one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations, the additional operations including: determining an angular offset between the determined iris or pupil position of the individual based on the image data and the expected iris or pupil position; The gaze deviation metric is based on the angular offset.

17. 13. The system of claim 12, wherein the one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations, the additional operations including: generating a plurality of images of the face of the individual from a plurality of viewpoints based on the image data; At least one of the gaze deviation metrics or the facial paralysis metrics is determined using a first image of the plurality of images corresponding to a first viewpoint and a second image of the plurality of images corresponding to a second viewpoint.

18. 13. The system of claim 12, wherein the one or more non-transitory computer-readable media include additional computer-readable instructions that, when executed by the one or more processing devices, cause the one or more processing devices to perform additional operations, the additional operations including: performing one or more segmentation processes on the image data to identify the eyes of the individual; Identifying that a viewpoint of an image included in the image data is different from a reference viewpoint; modifying a first position of a first iris of the individual and a second position of a second iris of the individual to compensate for a difference between the viewpoint of the image and the reference viewpoint; Including, the system.

Images