System and method for diagnosing disease

Abstract

The present disclosure provides systems and methods for diagnosing disease. In some aspects, an imaging system is provided that includes a light source configured to illuminate a retina of an eye with light, one or more imaging devices configured to receive light returned from the retina and generate one or more spatio-spectral images of the retina, and a computing device configured to receive the one or more spatio-spectral images of the retina, evaluate the one or more spatio-spectral images, and identify one or more biomarkers indicative of neurodegenerative pathology.

Description

[Technical Field]

[0001] Related Applications This application claims the benefit of and priority to U.S. Provisional Application No. 62 / 965,080, filed January 23, 2020, and U.S. Provisional Application No. 63 / 059,349, filed July 31, 2020, the entire contents of which are incorporated herein by reference.

[0002] The present disclosure relates to systems and methods for diagnosing diseases, such as Alzheimer's disease, by using optical techniques to measure and analyze characteristics of the eye. [Background technology]

[0003] Alzheimer's disease (AD) is a debilitating and fatal neurodegenerative disorder. Confirmation of the disease is generally performed postmortem. Several existing diagnostic systems involve highly invasive procedures and imaging devices, which are often inaccessible or unsuitable due to cost, complexity, or the use of harmful radioactive tracers.

[0004] There is a need for a non-invasive detection system that is easily operated and accessible by clinicians to screen patient populations for the early detection, diagnosis, and tracking of patient response to preventative or therapeutic interventions for AD-related pathology.

[0005] The optic nerve and retina are outgrowths of the developing brain, and many conditions that affect the brain, such as amyloid-beta (Aβ) protein accumulation, changes to the structure of retinal layers, and other changes in chemical composition, structure, and function, also manifest in these structures. Optical analysis is well suited to this need because the eye can be easily examined using a variety of non-invasive light-based techniques to identify these physical changes. Summary of the Invention

[0006] The present disclosure provides systems and methods for diagnosing disease. In some aspects, an imaging system is provided that includes a light source configured to illuminate a retina of an eye with light, one or more imaging devices configured to receive light returned from the retina and generate one or more spatio-spectral images of the retina, and a neuroimaging system configured to receive the one or more spatio-spectral images of the retina, evaluate the one or more spatio-spectral images, and generate one or more neuroimaging images. origin and a computing device configured to identify one or more biomarkers indicative of a pathology. In some embodiments, the one or more imaging devices include a spectral sensor. The spectral sensor may be, for example, a hyperspectral sensor or a multispectral sensor.

[0007] In some embodiments, the nerve origin Pathology (neurology origin The pathology is selected from the group consisting of Alzheimer's disease, Parkinson's disease, amyotrophic lateral sclerosis, multiple sclerosis, prion diseases, motor neurone diseases (MND), Huntington's disease (HD), spinocerebellar ataxia (SCA), spinal muscular atrophy (SMA), and cerebral amyloid angiopathy (CAA). In some embodiments, the biomarker comprises amyloid formation or tau protein formation.

[0008] In some embodiments, the system may further include a retinal observation device, wherein the one or more imaging devices and the light source are integrated into the retinal observation device. The retinal observation device is, for example, a fundus camera. In some embodiments, the one or more spatio-spectral images include spectral images of multiple retinal regions. In some embodiments, for each of the multiple retinal regions, the one or more spatio-spectral images include spectral images in multiple wavelength ranges. The spectral images include spatial information about the corresponding retinal region. The spatial information may include texture, formations, and patterns in the corresponding retinal region. In some embodiments, evaluation of the one or more spatio-spectral images uses pixel-by-pixel analysis of the images.

[0009] In some embodiments, an imaging system is provided, comprising a light source configured to illuminate a retina of an eye with light, one or more imaging devices configured to receive light returned from the retina and generate at least one spectral image and at least one spatial image of the retina, and a neuroimaging system configured to receive the at least one spectral image and at least one spatial image of the retina, evaluate the images, and generate a neuroimaging signal. origin and a computing device configured to identify one or more biomarkers indicative of the disease. In some embodiments, the one or more imaging devices include a spatial camera configured to generate at least one spatial image and a hyperspectral camera configured to generate at least one spectral image. In some embodiments, the one or more imaging devices include a hyperspectral camera configured to generate images including a spectral image and a spatial image. In some embodiments, the biomarkers include amyloid formation or tau protein formation.

[0010] In some embodiments, the neuro originA method for diagnosing a disease is provided, the method including: illuminating a retina of an eye with light using a light source; generating at least one spatio-spectral image with one or more imaging devices, the one or more imaging devices configured to receive light returned from the retina; evaluating the at least one spatio-spectral image, the image being evaluated by a computing device configured to receive the at least one spatio-spectral image from the one or more imaging devices; and evaluating a neurological function. origin and identifying one or more biomarkers indicative of the disease.

[0011] In some embodiments, evaluating the at least one spatio-spectral image uses pixel-by-pixel analysis of the image. In some embodiments, the one or more imaging devices include a spectral sensor. In some embodiments, the spectral sensor includes a hyperspectral sensor or a multispectral sensor. In some embodiments, the neurosurgery device includes a neural network. origin The disease is selected from the group consisting of Alzheimer's disease, Parkinson's disease, amyotrophic lateral sclerosis, multiple sclerosis, prion disease, motor neuron disease (MND), Huntington's disease (HD), spinocerebellar ataxia (SCA), spinal muscular atrophy (SMA), and cerebral amyloid angiopathy (CAA). In some embodiments, the biomarker comprises amyloid formation or tau protein formation.

[0012] In some embodiments, the neuro origin A method for diagnosing a disease is provided, the method including: obtaining, by a computing device, a retinal image mosaic comprising a plurality of spatio-spectral images from one or more regions of the retina; analyzing, by the computing device, the plurality of spatio-spectral images to identify neural defects; origin The method includes identifying one or more biomarkers indicative of pathology, and generating, by a computing device, a digital representation indicative of the presence or absence of the biomarkers in one or more regions of the retina.

[0013] In some embodiments, the method comprises: origin In some embodiments, the method further includes predicting a probability of pathology. In some embodiments, the digital representation is a heat map overlaid on the retinal image mosaic. In some embodiments, analyzing the plurality of spatio-spectral images includes using a pixel-by-pixel analysis of each image. In some embodiments, the plurality of spatio-spectral images are generated by one or more imaging devices including a spectral sensor. The spectral sensor may comprise, for example, a hyperspectral sensor or a multispectral sensor. ... origin The pathology is selected from the group consisting of Alzheimer's disease, Parkinson's disease, amyotrophic lateral sclerosis, multiple sclerosis, prion disease, motor neuron disease (MND), Huntington's disease (HD), spinocerebellar ataxia (SCA), spinal muscular atrophy (SMA), cerebral amyloid angiopathy (CAA). In some embodiments, the biomarker comprises amyloid formation or tau protein formation.

[0014] In some embodiments, an imaging system is provided that includes a light source configured to illuminate a retina of an eye with light, one or more imaging devices configured to receive light returned from the retina and generate one or more spatio-spectral images of the retina, and a computing device, wherein the computing device obtains a retinal image mosaic including one or more spatio-spectral images from one or more regions of the retina, and analyzes the one or more spatio-spectral images to determine neural activity. origin It may be configured to identify one or more biomarkers indicative of pathology and generate a digital representation indicative of the presence or absence of the biomarkers in one or more regions of the retina.

[0015] In some embodiments, any of the systems or methods may be used to segment one or more blood vessels in the retina and identify biomarkers along one or more blood vessels or blood vessel walls.

[0016] The present disclosure will be further described in the detailed description that follows, by way of non-limiting examples of illustrative embodiments, with reference to the indicated several drawings in which like reference numerals refer to like parts throughout the several views of the drawings. [Brief explanation of the drawings]

[0017] [Figure 1A] 1 is a schematic diagram of an exemplary embodiment of a system having a light source and an imaging device. [Figure 1B] 1 is a schematic diagram of an exemplary embodiment of a system having a light source and an imaging device included in a retinal viewing device. [Figure 1C] 1 is a schematic diagram of an exemplary embodiment of a system having a light source and an imaging device external to a retinal viewing device. [Figure 1D] FIG. 1 illustrates an exemplary embodiment of a system having an imaging device included in a retinal viewing device and a light source separate from the retinal viewing device. [Figure 1E] FIG. 2 is a more detailed schematic diagram of an exemplary embodiment of the system. [Figure 2A] FIG. 1 shows an exemplary fundus camera view acquired with a corresponding line scan from a spectrometer. [Figure 2B] FIG. 1 shows a diagram illustrating segmented regions of the optic disc. [Figure 3A] FIG. 1 illustrates an exemplary fundus camera. [Figure 3B] FIG. 1 illustrates an exemplary image from a spectrometer. [Figure 3C] FIG. 1 illustrates an exemplary image from a hyperspectral camera. [Figure 3D] FIG. 1 shows an exemplary image from a color fundus camera. [Figure 4] 1 is an exemplary flowchart of a method for processing a monospectral signal. [Figure 5A-B]FIG. 5A is an exemplary plot of optical density spectra of the discs of a subject with amyloid status positive and a subject with amyloid status negative, and FIG. 5B is an exemplary plot of optical density spectra of the fovea of ​​a subject with amyloid status positive and a subject with amyloid status negative. [Figure 5C] 10 is an exemplary plot of average results from the temporal disc region. [Figure 6] 10 is an exemplary plot of wavelength significance of reflectance signals in the temporal zone by amyloid status. [Figure 7A-B] FIG. 7A is an exemplary graph of wavelength significance of reflectance signal ratios in temporal zones by amyloid status, and FIG. 7B is an exemplary graph of wavelength significance of optical density signal ratios in temporal zones by amyloid status. [Figure 8] 1 is an exemplary flowchart of a wavelength signature identification method. [Figure 9A] FIG. 1 illustrates an example of an ensemble prediction model that uses various features from retinal imaging. [Figure 9B] FIG. 1 illustrates an example of an ensemble prediction model using various features from retinal imaging. [Figure 10A-B] FIG. 10A is an exemplary plot of probabilistic amyloid status calculated from the ensemble model, and FIG. 10B is an exemplary plot of a receiver operating characteristic curve for classification of amyloid status from the predictive model. [Figure 11] FIG. 1 illustrates an exemplary embodiment of a method for processing a set of hyperspectral or multispectral images. [Figure 12] FIG. 1 illustrates one embodiment of a representation of multiple 3D images taken at multiple wavelengths from a single location on the retina. [Figure 13] FIG. 1 shows a stack of images taken at six different locations on the retina that can be used as CNN input. [Figure 14]FIG. 10 shows an exemplary representation of images taken at six different locations of a heat map superimposed on an image of the retina. [Figure 15A-B] FIG. 15A shows amyloid and tau proteins identified in a representative histological slide from a subject with Alzheimer's disease, and FIG. 15B shows the absence of amyloid and / or tau proteins in a representative histological slide from a healthy subject. [Figure 16] FIG. 15B shows spectral transmission results of a brain slice from the same subject as in FIGS. 15A and 15B. [Figure 17A] 1 is an exemplary flowchart illustrating a method for analyzing retinal blood vessels. [Figure 17B] 17B shows an exemplary representation of an image associated with the method of FIG. 17A. [Figure 17C] FIG. 1 illustrates an exemplary representation of segmentation of blood vessels in the retina using AI. [Figure 18] FIG. 1 is a block diagram of an exemplary computer-based system and platform. [Figure 19] FIG. 1 is a block diagram of an exemplary computer-based system and platform. [Figure 20] 1 is a schematic diagram of an exemplary implementation of a cloud computing architecture. [Figure 21] 1 is a schematic diagram of an exemplary implementation of a cloud computing architecture. DETAILED DESCRIPTION OF THE INVENTION

[0018] While the above-identified drawings depict presently disclosed embodiments, other embodiments are contemplated, as noted in the discussion. The present disclosure presents exemplary embodiments by way of representation, not limitation. Numerous other modifications and embodiments may be devised by those skilled in the art that fall within the scope and spirit of the principles of the presently disclosed embodiments.

[0019] The following description provides only exemplary embodiments and is not intended to limit the scope, applicability, and configuration of the present disclosure. Rather, the following description of exemplary embodiments will provide those skilled in the art with an enabling description for implementing one or more exemplary embodiments. It will be understood that various changes may be made in the function and arrangement of elements without departing from the spirit and scope of the presently disclosed embodiments. Example embodiments are described below with reference to the figures. Identical, similar, and similarly acting elements in various figures are identified with the same reference numerals, and repeated descriptions of these elements are partially omitted to avoid redundancy.

[0020] The methods and systems of the present invention are directed to one or more pathologies associated with Alzheimer's disease (AD) and other neurological disorders. origin The systems and methods of the present disclosure may be used to detect the presence of one or more pathologies associated with diseases, such as Parkinson's disease (PD), amyotrophic lateral sclerosis (ALS), multiple sclerosis, prion diseases, motor neuron diseases (MND), Huntington's disease (HD), spinocerebellar ataxia (SCA), spinal muscular atrophy (SMA), cerebral amyloid angiopathy (CAA), other forms of dementia, and similar disorders of the brain or nervous system. origin The system can be used to detect protein aggregates of Aβ, tau, phosphorylated tau, and other neuronal proteins, which are indicative of disease, particularly Alzheimer's disease. In some embodiments, the detected protein aggregates can include at least one of tau neurofibrillary tangles, amyloid-β deposits or plagues, soluble amyloid-β aggregates, or amyloid precursor protein. These detected proteins can be correlated with cerebral amyloid and / or cerebral tau and thus indicate pathology in the brain. The system enables identification of at-risk populations, diagnosis, and tracking of patient response to treatment.

[0021] In some embodiments, the disclosed systems and methods may detect biomarkers indicative of tau pathology or tauopathy, including, but not limited to, total tau (T-tau), tau PET, and phosphorylated tau (P-tau). In some embodiments, biomarkers indicative of tauopathy include, but are not limited to, phosphorylated paired helical filament tau (pTau), early tau phosphorylation, late tau phosphorylation, pTau181, pTau217, pTau231, total tau, plasma Aβ42 / 40, neurofibrillary tangles (NFTs), and misfolded tau protein aggregates. In some embodiments, neurofilament light protein (NFL), neurofilaments (NF), or abnormal / elevated neurofilament light protein (NFL) concentrations may be detected. In some embodiments, surrogate markers of neurodegenerative disease or nerve damage may be detected, such as volume loss or other changes in the retina or optic nerve, degeneration within the neurosensory retina, and optic disc axonal damage. In some embodiments, an inflammatory response or neuroinflammation may be detected, and neuronal damage may be detected. originIn some embodiments, such inflammatory responses may be detected in retinal tissue. Examples of such responses include, but are not limited to, retinal microglial activation, degenerating ganglion cells (ganglionoid degeneration), or astrocyte activation. Other protein aggregates or biomarkers useful in the methods and systems of the present disclosure include alpha-synuclein and TDP43 (TAR DNA binding protein-43), as well as others described, for example, in Biomarkers for Tau Pathology (Molecular and Cellular Neuroscience, Volume 97, June 2019, Pages 18-33), which is incorporated herein by reference in its entirety. In some embodiments, the systems and methods of the present disclosure detect one or more neuronal aggregates in a patient's eye tissue, brain tissue, central nervous system tissue, peripheral nervous system tissue, cerebrospinal fluid (CSF), or any other tissue in which such formations or their biomarkers occur. origin In some embodiments, the disclosed systems and methods can be used to detect the presence or absence of protein aggregates or biomarkers indicative of disease. origin Detect protein aggregates or biomarkers indicative of disease. In some embodiments, dyes or ligands may be used to support the methods and systems disclosed herein. In some embodiments, the results of the optical examination may be confirmed using anatomical MRI, FDG PET, plasma testing, and / or CSF total tau.

[0022] In some embodiments, a nerve in the eye originA non-invasive optical-based ocular detection system is provided for detecting disease-related pathologies. The system can be used for optical examination of a portion of the fundus, such as the retina, to search for signs of AD-related pathologies in a patient's eye tissue, brain tissue, central nervous system tissue, cerebrospinal fluid (CSF), or any other tissue where the formation of AD-related pathologies and their biomarkers occurs. The device is an optical-based tool that provides an accessible, non-invasive procedure for identifying populations at risk for Alzheimer's disease, diagnosing, and tracking the effectiveness of treatments and interventions. In some embodiments, disease can be detected by first determining specific regions of tissue to be probed using a first imaging modality. This detected information can be used to guide a second imaging modality to determine the presence of one or more AD-related pathologies, including, but not limited to, protein aggregates, including at least one of tau neurofibrillary tangles, amyloid-β deposits, soluble amyloid-β aggregates, or amyloid precursor protein. For example, a first imaging modality may be used to identify the location of blood vessels within the retina, and a second imaging modality may then be used to analyze the spectral characteristics of the vessels where AD-related pathology may be more pronounced. The system allows for the identification, diagnosis, and tracking of patient response to treatment of at-risk populations.

[0023] Imaging system

[0024] In some embodiments, a system for diagnosing a disease such as Alzheimer's disease includes a retinal observation device, a light source that illuminates the retina, and an imaging device that generates measurements or images of the retina from light reflected by the retina and received by the retinal observation device (reflection of light from the light source). The retinal observation device may be any optical assembly, such as one or more lenses, configured to collect light reflected from the retina and / or other portions of the fundus. In some embodiments, the system may further include a computer system that analyzes the imaging data.

[0025] A retinal imaging system configured to capture one or more images of the retina for pathology detection may include an imaging device 102 and a light source 103, as shown in FIG. 1A. In some embodiments, the imaging device may be in the form of a hyperspectral or multispectral camera configured to capture one or more spatio-spectral images of the retina, as described in more detail below. In some embodiments, the imaging device and light source may be disposed within a housing 115 having optical elements 104 configured to direct light from the light source to the retina of the eye 105 and direct light reflected from the retina to the imaging device. In some embodiments, the housing 115 may be a handheld housing. In some embodiments, the imaging device 102, the light source 103, or both may be in communication with a computing device 106 for acquiring and analyzing pathology data.

[0026] 1B , the imaging device 102 and the light source 103 may be integrated (with or without a housing 115) into a retinal observation device 101 (e.g., a fundus camera) having optical elements 104 configured to direct light from the light source to the retina of the eye 105 and direct light reflected from the retina to the imaging device. Output from the imaging device is coupled to a computing device 106. The computing device may be configured to control settings of one or more of the imaging devices, including image settings and scanning and positioning settings. In some embodiments, the imaging device 102 and the light source 103 may be integrated into a retinal observation device 101, such as a fundus camera.

[0027] Figure 1C shows an embodiment in which the imaging device 102 and light source 103 are external to the retinal viewing device 101 and the optical element 104 is configured to be coupled to the imaging device and light source through an external port. Figure 1D shows an embodiment in which the light source 103 is separate from the retinal viewing device 101.

[0028] FIG. 1E is a more detailed schematic diagram of an embodiment in which a spatial spectrum from the fundus can be acquired by a retinal observation device 101, such as a fundus camera with an external camera port (i.e., a Topcon NW8, EX, or DX). Illumination is provided by a light source 103, which may be a xenon flash lamp integrated within the fundus camera. The light is directed toward the eye by optical elements 104, including a fundus camera objective lens. The optical elements 104 receive light reflected from the eye and direct it toward the external camera port. A field lens 107 is attached to the fundus camera's external camera port. Light exiting the field lens is split into two beam paths by a beam splitter 108. The relative intensities of the beam paths from the beam splitter can be 50 / 50, 30 / 70, or other combinations to optimize the signal collected by each imaging device. One beam path from the beam splitter passes through a 60 mm focal length macro lens 110 before reaching the imaging device 113. The imaging device 113 may be an optical spectrometer 111, such as a Specim VIR V10E or Specim VNIR V10E, having a camera 112, such as a PCO pixelfly camera. The other beam path of the beam splitter 108 passes through a 60 mm focal length macro lens 109 and is connected to the imaging device 102, such as a digital single-lens reflex camera (e.g., Canon E90) or a snapshot hyperspectral imager (e.g., Ximea MQ022HG-IM-SM4X4-VIS). Outputs from the imaging devices 102 and 113 are coupled to the computing device 106. A trigger mechanism 114 may be coupled to the imaging devices to synchronize their timing, or the trigger mechanism may be coupled to the computing device to control a trigger signal. The trigger mechanism may further control the timing of a light source, such as a xenon lamp flash; alternatively, in other embodiments, a xenon flash lamp in a fundus camera may trigger image capture at the imaging device, such that the imaging device captures an image of the eye when illuminated by the xenon flash. Using a fundus camera, various views of the fundus can be obtained, as shown in FIG. 2A.The acquired fundus camera view is shown in Figure 2A along with the corresponding line scan from the spectrometer. Figure 3A shows an exemplary fundus camera, and Figures 3B-3D show exemplary images from the spectrometer, hyperspectral camera, and color fundus camera.

[0029] The light source may be a broadband light source emitting light in a wide spectrum, e.g., UV (ultraviolet), visible, near-infrared, and / or infrared wavelength ranges, or a narrowband light source emitting light in a narrow spectrum or single wavelength. The light source may emit light in a single continuous spectrum or multiple discontinuous spectra. The light source may emit light having a fixed wavelength range and intensity, or the wavelength range and intensity may be adjustable. The light source may consist of a single light source or a combination of multiple light sources of the same or different types described above. The light source may be a xenon lamp, a mercury lamp, an LED, or any other light source. The light source may direct light to the retina through a retinal observation device using the same optical assembly configured to collect light reflected from the retina, or the light source may direct light to the retina via different optical paths.

[0030] In some embodiments, one or more optical filters may be used to filter the output of the light source, filter light reflected from the retina, or filter light entering one or more imaging devices, so that only light of selected wavelengths may be received by the one or more imaging devices. The optical filters may be adjustable or switchable (i.e., filter wheels or acousto-optic tunable filters), so that one or more wavelengths or wavelength ranges of light may be selected to pass through the filter simultaneously or sequentially.

[0031] The imaging device may generate a spatial image, a spectral measurement, or a spatio-spectral image. A spatial image is a two-dimensional image with a single light intensity value for each image pixel, such as produced by a monochrome (grayscale) camera. Three-dimensional spatial images may be generated using imaging devices such as optical coherence tomography (OCT), confocal microscopy, etc. Spectral measurements are measurements of the wavelength content of light, such as produced by an optical spectrometer. Spectral measurements may be of a narrow or broad range of wavelengths, or a set of continuous or discontinuous wavelengths or wavelength ranges (spectral bands). Spectral measurements may be based on absorbance, reflectance, wavelength shift (such as Raman spectroscopy, which measures wavelength shift relative to a narrowband light source such as a laser), or other spectral characteristics. The imaging device may be one or more of a camera, a spectrometer, a hyperspectral device, a multispectral device, an RGB camera, a microscope (conventional or confocal), or an optical coherence tomography system, where the optical coherence tomography system includes an imaging device (such as a camera) configured to receive images and communicate with a computer to transmit the images for analysis. The imaging device may be in the form of a stand-alone device or may be in the form of a sensor incorporated into a retinal viewing device or similar device.

[0032] A spatial-spectral image (or simply "spatial spectra" or "spectral image") is a two-dimensional image with spectral measurements of light at each image pixel. For example, a spatial-spectral image can be viewed as a 3D image, with the third dimension being the spectrum. In some embodiments, a spatial-spectral image can include spatial information that can be related to anatomical location. In some embodiments, a spatial-spectral image can include information related to patterns, formations, and / or textures of the imaged area, which can be observed based on the different wavelengths at which the image was captured. For example, a particular pathological formation may be observed at one or more wavelengths, but not all wavelengths. Spatial-spectral images can be generated by hyperspectral cameras, multispectral cameras, or by scanning a point spectrometer in two dimensions or a line spectrometer in one dimension (also known as a push-broom imager or whisk-broom imager). Line spectrometers can also generate one-dimensional spatial spectral images (i.e., 1 × n) with spectral measurements for each pixel along a line without scanning, and point spectrometers can generate point "images" (i.e., 1 × 1) without scanning. Some imaging techniques also allow for the generation of three-dimensional spectral images, in which a spectral image is generated for each pixel within a three-dimensional volume. In general, multispectral cameras tend to generate images with high spatial resolution but low spectral resolution, such as color (RGB) cameras (which measure three wavelength components, while other multispectral cameras can measure 10 or more wavelength components or spectral bands), and hyperspectral cameras tend to generate images with high spectral resolution but low spatial resolution. Scanning spectrometers can generate images with both high spatial and high spectral resolution, but require additional scanning optics and software.Spatio-spectral images can also be generated using a monochrome camera by sequentially measuring different wavelengths by illuminating the target with light of different wavelengths (by changing the light source and / or the wavelength of light emitted by the light source and / or by changing the optical filter used at the output of the light source) or by placing an optical filter at any point in the optical path between the light source, target, and camera to filter the light received by the camera.

[0033] A pure spatial image can also be generated from a spatio-spectral image by combining measurements of individual wavelength components at each pixel into a single intensity value for that pixel, and a hyperspectral camera can be used to generate both spatio-spectral and spatial images. In some embodiments, the light source and / or imaging device can be included in the same housing as the retinal observation device, attached to the retinal observation device, or separate from the retinal observation device. For example, a fundus camera can be used to image the fundus by providing broadband illumination and imaging optics, including an integrated or external camera, for capturing images of the fundus. In some embodiments, two or more imaging devices can be used to capture images simultaneously or sequentially. This can be done to obtain both high spatial and high spectral resolution images using different imaging devices. This can also be done to analyze images from a first imaging device to identify spatial and / or spectral features to determine which second imaging device to use and / or which location or portion of the retina to image with the second imaging device. In some cases, instead of using a second imaging device, the first imaging device can be used in a different setting to capture a second image of the retina having different spatial and / or spectral characteristics. In some embodiments, a first imaging device can be coupled to a retinal viewing device to generate a first image, and then a second imaging device can be coupled to the retinal viewing device to generate a second image.

[0034] In some embodiments, a beam splitter (such as a partially reflecting mirror) may be coupled to the retinal observation device and used to simultaneously split (at any intensity ratio) light collected by the retinal observation device among two or more imaging devices coupled to the beam splitter. In some embodiments, a beam redirection device (such as a movable mirror) may be coupled to the retinal observation device and used to sequentially redirect (redirect) light collected by the retinal observation device to two or more imaging devices (at any time ratio). Two or more beam splitters and / or two or more beam redirectors may also be used in series to split and / or redirect light among three or more imaging devices. In general, the system may comprise a fundus camera (using an internal integrated camera) that provides illumination and images the posterior segment of the eye, a beam splitter coupled to the external camera port of the fundus camera, a hyperspectral or multispectral camera coupled to a first output of the beam splitter to acquire spatial-spectral images, and a spectrometer coupled to a second output of the beam splitter to acquire high-resolution spectral measurements of points on the posterior surface of the fundus or a line across the posterior surface of the fundus.

[0035] In some embodiments, the output of the imaging device can be coupled to a computing device such as a computer (PC or laptop). In some embodiments, the spectral image can be read from the spatial spectrum imager and sent to the computing device, where wavelength calibration can be performed using previously acquired spectra of mercury lamps or other light sources with well-defined spectral characteristics. The wavelength locations of the peaks in the mercury lamp spectrum have well-defined and characterized wavelengths according to NIST standards. By comparing the known wavelengths and peak locations of the mercury lamp spectrum with the spectrum measured by the spectral imaging device and the pixels in the measured spectrum where those wavelengths and peak locations appear, a pixel-to-wavelength mapping can be calculated for the image, allowing the wavelengths of light in subsequent images to be known. Pixels in the spectral image where mercury lamp peaks are measured can be assigned to the known wavelengths of those peaks. By noting the pixels where each known mercury lamp peak is measured, an interpolation function can be calculated that maps each spatial pixel to a wavelength value, and this interpolation function can be used to correctly assign each pixel's wavelength value in subsequent spectral measurements.

[0036] In some embodiments, registration may be performed between different spectral or spatial images to ensure spatial alignment between the images. This may be done by identifying corresponding spatial features in two or more images and shifting (translating and / or rotating) the image positions so that the spatial features overlap in a common, registered coordinate system. The calculated shift for each image relative to the common, registered coordinate system may then be used to shift subsequent images. In some embodiments, when two or more imaging devices are used simultaneously, triggering of image capture may be performed to ensure temporal alignment. This may be done by connecting a common trigger line to all of the imaging devices and sending a trigger signal manually or from a computer to capture time-synchronized images.

[0037] The spectral image may be regionally segmented to identify pixels within various components of the eye, including the optic disc (nerve head), retina, and fovea, as shown in FIG. 2A. Furthermore, regions within the optic disc may be further segmented into more specific components, including the temporal rim, nasal rim, inferior rim, superior rim, and cup regions, as shown in FIG. 2B. Segmentation of the imaged eye components may be performed in various ways. In some embodiments, segmentation may be performed manually. In some embodiments, segmentation may be performed by an automated segmentation algorithm. In some embodiments, pixels within the temporal region of the optic disc are used in subsequent calculations to determine metrics indicative of the amyloid and tau status of individual pixels. The values ​​extracted from other regions may also include information related to an individual's amyloid or tau status, as well as information related to other ocular or systemic pathologies. For example, the values ​​extracted from other regions may include amyloid or tau protein deposits measurable by spectral imaging, or may indicate the effects of these proteins on the tissue. Furthermore, the values ​​extracted from other regions may include information related to other pathologies of the fundus, such as macular degeneration and diabetic retinopathy, which may also be measurable by spectral imaging.

[0038] Single Spectrum Analysis

[0039] 4 illustrates an exemplary embodiment of a method for processing a single-spectrum signal. In step 300, spectral information is obtained from one or more regions of the retina of one or both eyes of a patient. The spectral information associated with the image may be evaluated in step 302 to identify one or more patterns indicative of pathology. In step 304, the significance of the identified patterns may be determined, and one or more pathologies may be diagnosed in step 306.

[0040] In some embodiments, the optical density and / or reflectance of the acquired spectrum can be used to determine the presence or absence of amyloid formation or tau formation. Methods for determining optical density and reflectance are described below. Additionally, the present disclosure provides methods for identifying wavelength range(s) of optical density and reflectance, and optical density and reflectance ratios, that have significance for the presence or absence of amyloid formation or tau formation.

[0041] The spectral values ​​in each region (disc region, retina, fovea, etc.) are spatially averaged to obtain the average spectrum S ave is generated, where S ave is the average of all pixel values ​​contained in the region.

[0042] The average spectrum from each region is then multiplied by a previously acquired white reference spectrum S to calculate the average optical density (OD). ref Used in conjunction with the white reference spectrum S ref is obtained by imaging a diffuse broadband reflective standard target, e.g., a Spectralon target, and calculating the average spectrum of the acquired image. OD is a measure of how well a material absorbs light, with higher OD values ​​corresponding to higher levels of absorption by the material. The optical density of each region is given by OD=log(S ref / S ave ) It can be calculated as:

[0043] The OD spectrum is normalized, which can be done by dividing the spectrum by a value between 700 and 800 nm or by other wavelength values, or by other signal normalization methods such as standard normal variate (SNV) normalization.

[0044] In some embodiments, a reflectance spectrum R may be calculated. Reflectance is a measure of the light reflectance of the imaged material, with higher reflectance values ​​indicating a material with more reflective properties. In some embodiments, the reflectance R is calculated as: R=log(S ave / S ref ) It can be calculated as:

[0045] In some embodiments, the reflectance spectrum is also normalized to wavelengths between 700 nm and 850 nm. Alternatively, the reflectance spectrum can be normalized by standard normal variate (SNV), minimum maximum, or other normalization methods.

[0046] Example OD spectra are shown in Figures 5A, 5B, and 5C for the total disc, fovea, and temporal disc regions of an amyloid-status positive and an amyloid-status negative subject. In Figure 5A, the disc optical density spectrum 310 of the amyloid-status negative subject is distinguished from the disc optical density spectrum 312 of the amyloid-status positive subject. In Figure 5B, the foveal optical density spectrum 314 of the amyloid-status negative subject is distinguished from the foveal optical density spectrum 316 of the amyloid-status positive subject. In FIG. 5C, the optical density spectrum of the temporal disc region of the subject with negative amyloid status is shown bounded by one positive standard deviation 318 and one negative standard deviation 320, and is distinct from the optical density spectrum of the temporal disc region of the subject with positive amyloid status, which is shown bounded by one positive standard deviation 322 and one negative standard deviation 324.

[0047] To assess which wavelengths or wavelength ranges of OD and R values ​​significantly correlate with a subject's amyloid and / or tau status, statistical significance tests (e.g., Student's t-test) can be performed on the R and OD values ​​at each wavelength for spectra acquired from a group of subjects with negative amyloid and / or tau status and a group of subjects with positive amyloid and / or tau status. The statistical significance test identifies whether the values ​​at each wavelength significantly correlate with the subject's amyloid and / or tau status. Examples of significance tests include t-tests, Pearson correlation, Spearman correlation, chi-square, and ANOVA. Using a significance plot, such as the plot shown in Figure 6, wavelengths or wavelength ranges where the values ​​are significant can be identified. In this example, significance is defined as having a p-value less than 0.05, which is evident in the 600-700 nm wavelength range of the OD spectra. Once significant wavelengths or wavelength ranges are identified, the system can be optimized (for cost, size, speed, and other factors) by selecting or designing the imaging device, light source, and / or filters to measure only light at those wavelengths or wavelength ranges.

[0048] Such an approach can be used to assess the significance of spectral values ​​for any status of the sample being measured, and is not limited to human amyloid or tau status. For example, such an approach can be extended to measure other intraocular pathologies and other tissues. Ocular pathologies include, for example, macular degeneration, diabetic retinopathy, and glaucoma, and measurable tissues include, for example, skin, muscle, tendons, and blood vessels. Furthermore, such an approach can be applied to tissues of other organisms. While such an approach can be used similarly for these different pathologies and / or tissue types, the spectral values ​​determined to be significant may be different in each case. In some cases, it may be desirable to analyze a sample for more than one pathology or disease state, for example, to identify and diagnose subjects with more than one condition, or to identify subjects with a first disease state and exclude them from the analysis of a second disease state, where the presence of the first disease state is known to affect the outcome of the analysis for the second disease state.

[0049] In addition to the significance of R and OD values, the significance of the ratios of R and OD values ​​at various wavelengths can also be evaluated. Each wavelength-dependent spectral value of R and OD is divided by every other R and OD value, and all ratios can be evaluated for statistical significance. These significance ratio results can be plotted as a 2D image with the wavelength numerator (numerator) and denominator (denominator) as the X and Y axes. An example of a 2D significance ratio plot for an individual's amyloid status is shown in Figures 7A-7B. In this example, significance is defined as having a p-value less than 0.05, which is clearly visible within the contour area of ​​each image. In the example shown in Figures 7A-7B, the identified ratio regions (F1, F2, F3) exhibit significance and are used as features for model development.

[0050] A general approach to assessing the significance of values ​​at each wavelength and ratios of values ​​at each wavelength is shown in FIG. 8 . This process can be generalized to any spectral type of data to assess the significance of the signal for a parameter of interest, in this case, an individual's amyloid status and / or tau status. As shown in the exemplary flowchart of FIG. 8 , a wavelength feature identification method for values ​​at each wavelength and ratios of values ​​at each wavelength can include measuring spectral signals and correcting and / or calibrating the spectral signals (steps 400, 402). The significance of the spectral value at each wavelength relative to the parameter is calculated (step 404), and if the calculated value is found to be significant, the associated wavelength is recorded as significant (step 406). If not found to be significant, the ratio of the value at each wavelength relative to all other wavelengths is calculated (step 408), and then the significance of the spectral value at each wavelength ratio relative to the parameter is calculated (step 410). If the calculation is found to be significant, the wavelength ratio is recorded as significant (step 412). If a wavelength or wavelength ratio is found to be significant, the scan / image associated with that individual may be compared to a control image at the significant wavelength or wavelength ratio (step 414).

[0051] This approach can be used to evaluate the significance of a spectral value ratio for any status of the measured sample, and is not limited to amyloid or tau status in humans, but can be extended to, for example, other intraocular pathologies. Furthermore, this approach can also be applied to tissues of other organisms. This approach can be used in the same way for these different pathologies and / or tissue types, although the spectral value ratio that is determined to be significant may be different in each case.

[0052] In some embodiments, the system may include spectral measurements of interest in the mid-infrared wavelength range, specifically the 5900 nm to 6207 nm and / or 6038 nm to 6135 nm ranges, particularly wavelengths at 6053 nm and 6105 nm. The amyloid-β accumulation process can take years, during which amyloid-β exists in both soluble and plaque forms and folds into α-helical and β-sheet structures, the relative concentrations of which change over time. These structures and their concentration ratios are a function of the progression of the accumulation process and are important biomarkers for clinical assessment of the presence and progression of AD. It is known that differences in the protein's folding structure result in different spectral absorbances and reflectances; the α-helical structure of amyloid has a peak at 6053 nm, while the β-sheet structure of the protein has a peak at 6105 nm. Therefore, the peak absorbance and reflectance observed in the retina are indicative of concentration ratios; for example, a peak absorbance near 6079 nm may be evidence of a balanced mixture. The more the ratio tends toward the peak absorbance of one of the structures, the higher the concentration of the structure with that peak absorbance. Spectral imagers that specifically view the 6038-6135 nm range can be used to measure these biomarkers. Other important wavelengths are 5900 nm, 6060 nm, 6150 nm, and 6207 nm, which are associated with clinically important structures (β-hairpins, β-sheets, amyloid-β fibrils, and Tyr and Phe amino acids).

[0053] The feature identification method disclosed herein can be extended to identify wavelength ranges of spectral values ​​and spectral value ratios for the analysis of any sample of interest, and is not limited to spectroscopy of biological tissues. For example, this method can be extended for pharmaceutical process monitoring, industrial process monitoring, hazardous materials identification, explosives identification, and food process monitoring applications. This approach (method) can be extended to other optical and spectroscopic modalities to investigate significant features of interest in a sample. For example, this method can be used with various optical modalities, including Raman spectroscopy, fluorescence spectroscopy, laser-induced breakdown spectroscopy, and other optical modalities.

[0054] Significant wavelength regions of both the R and OD spectra and the ratio of the R and OD spectra can be identified as significant spectral features and used as inputs to machine learning (ML) or artificial intelligence (AI) algorithms to predict an individual's amyloid or tau status based on a model trained on spectra obtained from individuals with known status. ML models can include methods such as logistic regression, decision trees, random forests, linear discriminant analysis, neural networks (including convolutional neural networks), naive Bayes classifiers, nearest neighbor classifiers, or other ML or AI techniques.

[0055] In some embodiments, an ensemble approach may be employed, as shown in the examples of Figures 9A and 9B, where the features identified by the statistical significance tests described above along with other features extracted from retinal imaging (such as vessel tortuosity) are used as features to train an individual's ML model, and the model's output is considered in combination with various weights applied to the model's output to calculate a combined output to predict the individual's status.

[0056] Significant wavelength regions of both the R and OD spectra and the ratio of the R and OD spectra can be identified as significant spectral features and used as inputs to machine learning (ML) or artificial intelligence (AI) algorithms to predict an individual's amyloid or tau status based on models trained on spectra obtained from individuals with known status. ML models can include methods such as logistic regression, decision trees, random forests, linear discriminant analysis, neural networks (including convolutional neural networks), naive Bayes classifiers, nearest neighbor classifiers, or other ML or AI techniques. Other ML approaches, such as ensemble techniques, can also be employed, as shown in the example of Figures 9A-9B. Here, the features identified by the statistical significance tests described above, along with other features extracted from retinal imaging (e.g., vascular tortuosity), are used as features to train an individual's ML model, and the outputs of these models are considered in combination with various weights applied to the model's outputs to calculate a combined output for predicting the individual's status.

[0057] Such ensemble models can include features identified from hyperspectral, multispectral, and spectroscopy imaging, as well as features extracted from retinal images, including, but not limited to, tissue / vascular oxygenation, vascular tortuosity, cup-to-disc ratio, retinal nerve fiber layer thickness, and image texture metrics. Additionally, demographic and other medical information about individuals can also be used as inputs to such ensemble models, including, but not limited to, age, gender, intraocular pathology, comorbidities, lens status (natural vs. prosthetic), recent ocular surgery, whether dilating drops were used during imaging, and any other demographic or health information.

[0058] In ensemble models, different data types (i.e., spectral, spatial, demographic, etc.) can be processed independently by different machine learning or artificial intelligence algorithms and the output of those algorithms used as input to further algorithms, or composite data types (i.e., combined spatio-spectral data) can be used directly in the same algorithm to generate an output based on evaluation of multiple data domains. In some cases, the latter approach is used because it can better capture correlations in the data across multiple domains, which may be lost if the analysis combining the data types is only performed using extracted outputs from independent algorithms rather than the entire data set.

[0059] In an ensemble model using spatial retinal images, the analysis algorithm may use only the entire retinal image, a predetermined portion of the retinal image, or a segmented portion (section) of the image. This type of ensemble model can be used even if features such as the optic disc, which are often used as a basis for performing segmentation, are not present in the image. This type of model may analyze the entire image, for example, when the relevant information is not localized and appears over a wide area of ​​the image, or may analyze only a portion of the image or a segmented portion of the image, for example, when the relevant information is highly localized.

[0060] In some embodiments, machine learning or artificial intelligence algorithms can be designed to perform pixel-wise predictions. Pixel-wise predictions are based on the relationship between each pixel's spectral data and the spectral data from two or more adjacent or nearby pixels, rather than only data from a single pixel or the entire set of pixels combined, as is the case with channel-wise predictions. Existing convolutional neural networks often use channel-wise pooling as the last layer or network to create feature vectors. In these embodiments, channel-wise pooling is replaced with pixel-wise pooling followed by max pooling. Pixel-wise predictions are used because they prevent algorithms from relying on multiple pieces of information that are spatially separated from each other, which is important when relevant information is local, reducing model overfitting and improving performance. Pixel-wise predictions also allow for improved testing, validation, and explanation of algorithm outputs. The reason is that, unlike algorithms that rely on attention maps, input distortion, or other similar methods, this method allows us to verify that the location of the signal in the image corresponds to the correct area where the signal is expected to be present, thereby opening up the "black-box" of the algorithm and making the output prediction more transparent.

[0061] An example of prediction of an individual's amyloid status based on an ensemble model using features from the R ratio and OD ratio features from the eyes is shown in Figure 10A for 10 individuals with amyloid status negative and 10 individuals with amyloid status positive. The results of this model were evaluated using the receiver operating characteristic curve (ROC) in Figure 10B, providing an area under the curve (AUC) of greater than 0.9, indicating a high predictive ability of the developed model for amyloid status.

[0062] Hyperspectral analysis

[0063] 11 illustrates an exemplary embodiment of a method for processing a set of hyperspectral or multispectral images. In step 500, a retinal image mosaic is provided from images acquired from a patient. A spectral-spatial CNN (step 502) is used to create a heat map describing disease signal prediction probability (step 504). In step 506, the created heat map is evaluated and a final score is output that can indicate pathology.

[0064] For a given patient, various sources of information may be available to aid in predicting amyloid or tau status. For example, a patient may have spectral data from a subset of regions including the temporal, nasal, inferior, and superior rims of the optic disc, the cup, and the fovea, along with various other spatial regions within the retina. Individual models may be developed using data from each of these regions. Additional models may also be developed using data from vascular tortuosity (determined from color or hyperspectral images), nerve fiber thickness (determined from OCT), vascular oxygenation, pupil dilation, inflammatory response, demographic data, etc. Each individual model outputs a probability that a given subject's amyloid or tau status is positive or negative. Once all these models are trained, a prediction can be generated from an ensemble model, which may be a combination of all models for which data is available for a given individual. The final prediction may be a weighted combination of the outputs from the available models. The weight given to each model may be based on how significantly the predictions from that model correlate with the amyloid or tau status of the subjects in the training set. These weights may be adjusted as additional data becomes available. The rationale behind ensemble models is to address the reality that the combination of data available to make predictions will be different for each individual (e.g., some subjects may have data from the temporal limbus but not from the inferior limbus, or vice versa). Furthermore, different machine learning algorithms may be used to generate predictions for the various data sources available. The choice of algorithm may be based on the size and nature of the data.

[0065] In some embodiments, the algorithm design can build on state-of-the-art algorithms from the convolutional neural network (CNN) family (EfficientNet, ResNeXt, ViT, Scaled-YOLOv4), modifying the architecture to accept hyperspectral images instead of color (RGB) images, adapting layers to support the spatio-spectral requirements of the analysis, and varying the width, depth, and length of the network according to the capabilities required to detect signals in multispectral and / or hyperspectral retinal images. Each model can be further adapted to replace the last layer (close to the output) so that the final feature tensors of the network are pooled in a pixel-wise manner instead of a channel-wise manner.

[0066] The input to the neural network is a collection of HSIs from both eyes of each patient. For example, for each eye, a collection of up to seven images centered at different anatomical locations on the retina may be used. The locations may include one or more of the optic disc, center of the retina, fovea, superior, inferior, temporal, and nasal. In some embodiments, for each location, line spectrography data (line spectrography data) may optionally be acquired, which intersects the center of the image on a horizontal line. In some embodiments, a color fundus image is taken for each eye and may be used by AI to gain deeper insights and allow ophthalmologists to identify diagnoses and pathologies such as retinopathy, macular degeneration, glaucoma, cataracts, hypertension, etc.

[0067] FIG. 12 shows one embodiment of a representation of multiple 3D images acquired from a single location on the retina at multiple wavelengths. Images acquired at different wavelengths show different textures and structures. Each set of images from a single location becomes a stack of images that form a portion of the image for the CNN. Input to the CNN can include multiple stacks of HSI acquired at multiple locations on the retina. As shown in FIG. 12, different channels can show different textures and structures in the image. As shown, for example, some blood vessels and other structures or textures are present in wavelength band 2 but not in wavelength band 10, and vice versa. For example, FIG. 13 shows a stack of images acquired at six different locations on the retina that can be used as CNN input. Output from the CNN can include a probability score related to the likelihood of disease presence, and in some embodiments, the images can include those shown in FIG. 14. FIG. 14 shows an example set of images with a generated heat map overlaid on the images shown in FIG. 13. In the background is a representation of the original retinal image, and the heat map is color overlaid on the background. The heatmap shows one or more hotspots 510, which are spatio-spectral patterns indicative of protein formation. The overlay of the heatmap allows the user to see where the hotspots fall within the original retinal image. These heatmap images can be used as input for AI.

[0068] In some embodiments, the algorithm can be trained using a database of corresponding data from patients with known disease states. Multiple patients and a control set of healthy individuals can be used. Data can be acquired from each patient, and the collected data and / or images can be preprocessed. Low-quality images (following several criteria already specified by the inventors in the patent) can be eliminated, and images can be normalized. Data can be divided into three sets: training, validation, and test sets using multiple folds in a cross-validation method, and different models from different folds can be ensembled together using the training data before testing on the test set. In some embodiments, the training set is presented to the AI ​​during training and used for actual learning, the test set is used frequently during the training process to evaluate the model's performance on unseen data, and the validation set is provided completely separately from the AI ​​developer and used only once to validate the model on new data. Sometimes, training at an image-by-image level can be difficult. This is because relevant information may not be evident or present in all images acquired from a single subject, for example, when information is evident in the optic disc image but not in the fovea image, or when information is evident in the left eye but not in the right eye. In such cases, training at the image-by-image level becomes difficult because some labels (positive vs. negative) may mislead the AI. To address this issue, in some embodiments, some or all of a single subject's images are concatenated into a single mosaic of images and analyzed as a single ensemble. In this way, even if a signal is evident in only one of the images, the training label assigned to this mosaic will be correct and will not mislead the AI. In some embodiments, the final algorithm may be an ensemble of multiple algorithms, which is common practice in ML / AI.Images are taken of the patient, poor quality images are filtered out, and the images are compiled into the same mosaic as in the training dataset before being used by the algorithm to generate the final score. When using AI for prediction, some patients can be excluded based on certain clinical criteria. For example, if certain pathologies such as glaucoma or certain ethnicities are underrepresented in the training and validation sets, the AI ​​may be less likely to perform as expected on them. It is possible to run the AI ​​on all images and obtain clinical data from the patient.

[0069] Before using any of the collected or calculated spatial or spectral data, quality assurance criteria can be applied to ensure that the data is of sufficient quality to generate reliable predicted outputs from machine learning or artificial intelligence algorithms. For example, a spectral dynamic range, defined as the difference between the highest and lowest spectral bands at a particular pixel, can be calculated for each pixel in hyperspectral, multispectral, or spectroscopic data, and pixels with a spectral dynamic range below a preset percentile threshold can be rejected. This is because data with a low spectral dynamic range may be less informative and unusable. As another example, a saturation ratio, defined as the number of saturated data points (i.e., the maximum value of the measurement device) divided by the total number of data points in the input image, can be calculated, and images with saturation above a preset threshold can be rejected because excessive saturation results in a loss of information. A data point can be the overall intensity value measured at a pixel in an image, or the intensity of only one or more spectral (wavelength) components at that pixel. However, not all saturation reduces the predictive power of an algorithm; some saturation below a preset threshold can, in some cases, further improve predictive power to some extent. This is because saturation in a few data points indicates that the measured signal is reaching the maximum range of the imaging device, which means that the imaged subject is likely illuminated and highly reflective, such that the non-saturated data points are likely to produce signals with good intensity and dynamic range that can provide clear images and accurate results. If the data points are not saturated, it is possible that the imaged subject is not well illuminated and the dynamic range of the imaging device is not being fully utilized.As further examples of quality assurance criteria, image blurriness or image sharpness may be calculated based on the variation in intensity between adjacent image pixels, and images or portions of images with blurriness above a preset threshold may be rejected, or image homogeneity may be calculated based on the variation in intensity across all image pixels, and images or portions of images with too much or too little homogeneity may be rejected.

[0070] For each "quality assurance" criterion, a threshold may be set at which data will be accepted or rejected. Spectral dynamic range percentiles may vary, but in some embodiments, the range is 20% spectral dynamic range for the 5th percentile of pixels. In some embodiments, a maximum of 5% saturated pixels may be tolerated. Blur may be measured with a score of 0 to 1, with 1 being completely blurred. In some embodiments, a maximum blur of 0.2 may be tolerated. Homogeneity may be measured by examining histograms in subsections of the image and calculating the entropy across the different histograms. In some embodiments, a threshold is set to reject images with an entropy of 1.3 or greater. Based on the quality assurance criteria, light source and / or imaging device settings may be adjusted, for example, by increasing or decreasing the intensity of the light source to improve the saturation ratio, and this procedure may be repeated.

[0071] In Alzheimer's disease, both amyloid and tau levels in the brain are elevated prior to the onset of symptoms. Amyloid and tau levels correlate, with subjects who develop AD tending to have biomarker evidence of elevated amyloid deposition (detected by abnormal amyloid PET scans or low CSF Ab42 or Ab42 / Ab40 ratios) as the first identifiable abnormality, followed by biomarker evidence of pathological tau (detected via CSF phosphorylated tau and tau PET). This is thought to be because amyloid pathology induces altered release of soluble tau, leading to subsequent tau aggregation. These methods for predicting amyloid status shown in Figure 8 and the predictive capabilities of the developed model for amyloid status may also be effective for predicting an individual's tau status, since tau and amyloid levels are correlated.

[0072] The ability of hyperspectral imaging to detect tau protein in tissue is demonstrated in Figures 15A, 15B, and 16. Figure 15A shows an example histological slide of brain tissue from an individual with Alzheimer's disease stained with Congo red. It has discernible amyloid and tau protein deposits throughout the tissue, consistent with a diagnosis of Alzheimer's disease. Figure 15B shows an example histological slide of brain tissue from a healthy individual, also stained with Congo red. There are no discernible amyloid or tau protein deposits within the tissue, consistent with a negative diagnosis of Alzheimer's disease. Figure 16 shows the corresponding hyperspectral measurements of a brain section from the same individual histologically examined in Figure 15A, which has amyloid and tau deposits, compared to a healthy control sample in Figure 15B, which shows no amyloid or tau deposits. The difference in the spectra measured for tissue with and without tau protein is evident in Figure 16; in particular, a decrease in light transmittance is measured through tissue sections from individuals with tau deposits; this decrease in light transmittance is theorized to be due to the absorption of light at these wavelengths by tau protein. The target line 602 is the average for individuals with Alzheimer's disease, and the control line 600 is the average for healthy individuals. The decrease in transmittance in individuals with tau and amyloid proteins (i.e., Alzheimer's disease patients) indicates that healthy tissue can be distinguished from tissue with tau and amyloid proteins. These results demonstrate that spectral measurements can be used to distinguish between healthy tissue and tissue containing tau protein.

[0073] Because the eye is an extension of the central nervous system and directly linked to the brain by the optic nerve, proteins produced in the brain as part of the progression of Alzheimer's disease, such as beta-amyloid and tau, migrate from the brain to the retina. Therefore, detecting these proteins in the eye can indicate the presence or absence of these proteins in the brain and the risk of developing Alzheimer's disease. The ability to measure tau in brain tissue as demonstrated herein further indicates that it is possible to similarly measure tau in intraocular tissues, such as the retina and optic disc, and use these measurements as proxies for tau levels in the brain.

[0074] In some embodiments, the methods described herein can be used to analyze retinal blood vessels and vessel walls, as shown in Figures 17A-17B. Algorithms can be used to extract accurate vessel segmentations from captured images, and AI can be used to accurately locate amyloid markers along the vessels, i.e., vascular amyloidosis, covering a wide range of vascular changes and other vascular pathologies that may be associated with amyloid-related diseases, such as intraocular disease and cerebral amyloid angiopathy (CAA). As shown in Figure 17B, a model can be used to segment the vessels (step 700), and a projection can be developed (step 702). The vessel segmentation is extracted (step 704), and AI can be used to identify amyloid markers along the vessels (step 706).

[0075] CNN segmentation AI has been developed to automatically extract blood vessels from images such as HSI images. The HSI image is input to the AI, which outputs two probability maps showing the segmentation of blood vessels, as shown in Figure 17C. One map is for arteries, and the other is for veins. These maps are their own images, with higher brightness for each pixel where the AI ​​predicts the presence of a blood vessel. These maps are binarized using a configurable threshold, usually equal to 0.5. The output binary segmentation of arteries and blood vessels is sent to another AI that processes the vascular structure and spectral features along the vessels to detect disease-related biomarkers.

[0076] In some aspects, the present disclosure provides a system for measuring optical properties of an eye, the system including a retinal observation device, a light source configured to illuminate a retina of the eye observed by the retinal observation device, one or more imaging devices configured to receive light from the light source reflected by the retina and generate one or more spatial or spectral images of the retina, and a computing device configured to receive the images generated by the one or more imaging devices, calculate one or more metrics indicative of a disease state based on the images, and make a determination of the disease state. In some embodiments, such determination may be made using machine learning or artificial intelligence algorithms to compare one or more of the metrics indicative of the disease state to a database of corresponding values ​​measured from subjects with and without known disease states.

[0077] In some embodiments, the one or more imaging devices include at least one hyperspectral or multispectral imager and at least one optical spectrometer. In some embodiments, the at least one spatial image and the at least one spectral image are generated by the same imaging device or by a computing device from the output of the same imaging device. In some embodiments, the system further includes one or more optical elements that enable two or more imaging devices to receive light reflected by the retina and simultaneously or sequentially generate images of the retina. In some embodiments, the system further includes one or more optical filters to limit the wavelengths of light emitted by the light source or received by one or more of the imaging devices. In some embodiments, one or more metrics indicative of a disease state are calculated based on discrete wavelengths or wavelength ranges in the spectral images. In some embodiments, the discrete wavelengths or wavelength ranges are selected by using machine learning or artificial intelligence algorithms to compare a database of spectral images measured from subjects with known positive and negative disease states and determine wavelengths associated with one or more metrics indicative of the disease state.

[0078] In some embodiments, the light source is configured to emit only light of wavelengths appropriate for calculating a metric indicative of a disease state, or the system further includes an optical filter to limit the wavelengths of light received by the one or more imaging devices to wavelengths appropriate for calculating a metric indicative of a disease state. In some embodiments, the system further includes a trigger controller for synchronizing one or more of the imaging devices or the light sources. In some embodiments, the system further includes a wavelength calibration source, and the computing device is further configured to receive wavelength calibration signals measured by one or more of the imaging devices from the wavelength calibration source and calculate pixel-to-wavelength conversions for the one or more spectral images from the corresponding wavelength calibration signals.

[0079] In some embodiments, at least one metric indicative of a disease state is calculated for each pixel of at least one of the spectral image or spatial image. In some embodiments, the computing device may be further configured to segment the spectral image into various components of the eye and calculate a metric indicative of a disease state for each component based on averaged spectral data for that component. In some embodiments, the segmentation of the spectral image into multiple components is performed by an automatic segmentation algorithm. In some embodiments, the metric indicative of a disease state is calculated based on a ratio of disease state metrics between different components. In some embodiments, the system further includes a white reference light source and a target to generate a white reference spectrum for the system. In some embodiments, the metric indicative of a disease state is based on an optical density or reflectance of the retina calculated against a white reference spectrum at one or more wavelengths or wavelength ranges. In some embodiments, the metric indicative of a disease state is based on a ratio of optical density or reflectance at two or more different wavelengths or wavelength ranges. In some embodiments, the wavelength or wavelength range used to calculate the ratio is determined based on a statistical significance test performed on a database of corresponding values ​​measured from subjects positive for a known disease state and subjects negative for a known disease state. In some embodiments, the computing device is further configured to extract features from the one or more spatial images and calculate one or more spatial data metrics indicative of a disease state based on the extracted features alone or a combination of the extracted features and the spectral data metrics. In some embodiments, the computing device is configured to use the first image to determine the most appropriate imaging device settings for generating a second image from the same or a different imaging device.In some embodiments, the computing device is further configured to receive demographic or other medical information and calculate one or more additional metrics of the disease state based on the demographic or other medical information in combination with one or more spectral or spatial data metrics.

[0080] In some embodiments, the computing device is further configured to calculate one or more quality assurance metrics for the spatial or spectral data and reject data or portions of data for which the quality assurance metrics are above or below preset quality assurance thresholds. In some embodiments, the system further includes one or more additional imaging devices, such as optical coherence tomography, Raman spectroscopy, or other devices that measure the fundus in one or more dimensions, and the computing device is further configured to calculate one or more additional metrics indicative of a disease state based on the output of the imaging devices alone or in combination with other calculated metrics of the disease state. In some embodiments, the disease state metric is indicative of the amyloid or tau status of the eye, or the absence, presence, progression, or risk of developing Alzheimer's disease. In some embodiments, the machine learning or artificial intelligence algorithm is based on one or more of logistic regression, decision trees, random forests, linear discriminant analysis, neural networks (including convolutional neural networks), naive Bayes classifiers, nearest neighbor classifiers, or other ML or AI techniques.

[0081] FIG. 18 is a block diagram of an exemplary computer-based system and platform 800 in accordance with one or more embodiments of the present disclosure. However, not all of these components are necessarily required to implement one or more embodiments, and variations in the configuration and type of components may be made without departing from the spirit or scope of various embodiments of the present disclosure. In some embodiments, the exemplary computing devices and exemplary computing components of the exemplary computer-based system and platform 800 may be configured to manage multiple members and parallel transactions, as detailed herein. In some embodiments, the exemplary computer-based system and platform 800 may be based on a scalable computer and network architecture incorporating various strategies for data evaluation, caching, retrieval, and / or database connection pooling. One example of a scalable architecture is one capable of running multiple servers.

[0082] 18 , member computing device 802, member computing device 803, and member computing device 804 (e.g., clients) of exemplary computer-based system and platform 800 may include any computing devices capable of sending and receiving messages to and from other computing devices, such as servers 806 and 807, over a network (e.g., a cloud network), such as network 805. In some embodiments, member devices 802-804 may be personal computers, multiprocessor systems, microprocessor-based or programmable consumer electronics, network PCs, etc. In some embodiments, one or more member devices within member devices 802-804 may include computing devices that typically connect using wireless communication media, such as mobile phones, smartphones, pagers, walkie-talkies, radio frequency (RF) devices, infrared (IR) devices, CBs, integrated devices combining one or more of these devices, any mobile computing device, etc. In some embodiments, one or more of member devices 802-804 may be a device capable of connecting using a wired or wireless communication medium, such as a PDA, pocket PC, wearable computer, laptop, tablet, desktop computer, netbook, video game device, pager, smartphone, ultra-mobile personal computer (UMPC), and / or any other device equipped to communicate over a wired and / or wireless communication medium (e.g., NFC, RFID, NBIOT, 3G, 4G, 5G, GSM, GPRS, WiFi, WiMax, CDMA, satellite, ZigBee, etc.).In some embodiments, one or more of the member devices 802-804 may execute one or more applications, such as an internet browser, a mobile application, voice calls, video games, video conferencing, and email, among others. In some embodiments, one or more of the member devices 802-804 may be configured to receive and send web pages, etc. In some embodiments, a specially programmed exemplary browser application of the present disclosure may be configured to receive and display graphics, text, multimedia, etc. using any web-based language, including, but not limited to, Standard Generalized Markup Language (SMGL) such as HyperText Markup Language (HTML), wireless application protocol (WAP), Handheld Device Markup Language (HDML) such as Wireless Markup Language (WML), WMLScript, XML, JavaScript, etc. In some embodiments, one member device among the member devices 802-804 may be specially programmed in Java, .Net, QT, C, C++, and / or other suitable programming languages. In some embodiments, one or more member devices among the member devices 802-804 may include or be specially programmed to execute applications that perform a variety of possible tasks, including, but not limited to, messaging functions, browsing, searching, playing, and streaming various forms of content, including locally stored or uploaded messages, images and / or videos, and / or games.

[0083] In some embodiments, exemplary network 805 may provide network access, data transmission, and / or other services to any computing devices coupled to exemplary network 805. In some embodiments, exemplary network 805 may include or implement at least one dedicated network architecture, which may be based at least in part on one or more standards set by, for example, but not limited to, the Global System for Mobile communications (GSM) Association, the Internet Engineering Task Force (IETF), and the Worldwide Interoperability for Microwave Access (WiMAX) Forum. In some embodiments, exemplary network 805 may implement one or more of the GSM architecture, the General Packet Radio Service (GPRS) architecture, the Universal Mobile Telecommunications System (UMTS) architecture, and an evolution of UMTS called Long Term Evolution (LTE). In some embodiments, exemplary network 805 may include and implement the WiMAX architecture defined by the WiMAX Forum, as an alternative or in conjunction with one or more of the above.In some embodiments, and optionally in combination with any embodiment described above or below, exemplary network 805 may include, for example, at least one of a local area network (LAN), a wide area network (WAN), the Internet, a virtual LAN (VLAN), an enterprise LAN, a Layer 3 virtual private network (VPN), an enterprise IP network, or any combination thereof. In some embodiments, and optionally in combination with any embodiment described above or below, at least one computer network communication on exemplary network 805 may be transmitted based at least in part on one or more communication modes, such as, but not limited to, NFC, RFID, Narrow Band Internet of Things (NBIOT), ZigBee, 3G, 4G, 5G, GSM, GPRS, WiFi, WiMax, CDMA, satellite, and any combination thereof. In some embodiments, the exemplary network 805 may include mass storage such as network attached storage (NAS), a storage area network (SAN), a content delivery network (CDN), or other forms of computer- or machine-readable media.

[0084] In some embodiments, exemplary server 806 or exemplary server 807 may be a web server (or a series of servers) running a network operating system, examples of which may include, but are not limited to, Microsoft Windows Server, Novell NetWare, or Linux. In some embodiments, exemplary server 806 or exemplary server 807 may be used to provide cloud and / or network computing. Although not shown in FIG. 18 , in some embodiments, exemplary server 806 or exemplary server 807 may have connections to external systems, such as email, SMS, messaging, text messaging, advertising content providers, etc. Any of the features of exemplary server 806 may be implemented in exemplary server 807, and vice versa.

[0085] In some embodiments, one or both of exemplary servers 806 and 807 may be specially programmed to act as, by way of non-limiting example, an authentication server, a search server, an email server, a social networking server, an SMS server, an IM server, an MMS server, an exchange server, a photo sharing service server, an advertisement serving server, a financial / banking related service server, a travel service server, or any similarly suitable service-based server for users of member computing devices 801-804.

[0086] In some embodiments, and optionally in combination with any of the embodiments described above or below, for example, one or more of exemplary computing member devices 802-804, exemplary server 806, and / or exemplary server 807 may include specially programmed software modules that may be configured to send, process, and receive information using a scripting language, remote procedure calls, email, tweets, Short Message Service (SMS), Multimedia Message Service (MMS), instant messaging (IM), internet relay chat (IRC), mIRC, Jabber, an application programming interface, Simple Object Access Protocol (SOAP) methods, Common Object Request Broker Architecture (CORBA), HTTP (Hypertext Transfer Protocol), REST (Representational State Transfer), or any combination thereof.

[0087] FIG. 19 is a block diagram of another exemplary computer-based system and platform 900 in accordance with one or more embodiments of the present disclosure. However, not all of these components are necessarily required to practice one or more embodiments, and variations in the configuration and type of components may be made without departing from the spirit or scope of various embodiments of the present disclosure. In some embodiments, the illustrated member computing device 902a, member computing device 902b, through member computing device 902n each include at least a computer-readable medium, such as random access memory (RAM) 908 or FLASH memory, coupled to a processor 910. In some embodiments, the processor 910 may execute computer-executable program instructions stored in memory 908. In some embodiments, the processor 910 may include a microprocessor, an ASIC, and / or a state machine. In some embodiments, the processor 910 may include or be in communication with a medium, e.g., a computer-readable medium, that stores instructions that, when executed by the processor 910, cause the processor 910 to perform one or more steps described herein. In some embodiments, examples of computer-readable media may include, but are not limited to, electronic, optical, magnetic, or other storage or transmission devices capable of providing computer-readable instructions to a processor, such as processor 910 of client 902a. In some embodiments, other examples of suitable media may include, but are not limited to, floppy disks, CD-ROMs, DVDs, magnetic disks, memory chips, ROM, RAM, ASICs, configured processors, all optical media, all magnetic tape or other magnetic media, or any other medium from which a computer processor can read instructions. Similarly, various other forms of computer-readable media may transmit or carry instructions to a computer, including a router, private or public network, or other transmission device or channel (both wired and wireless).In some embodiments, the instructions may include code from any computer programming language, including, for example, C, C++, Visual Basic, Java, Python, Perl, JavaScript, and the like.

[0088] In some embodiments, member computing devices 902a-902n may include numerous external or internal devices, such as a mouse, CD-ROM, DVD, physical or virtual keyboard, display, or other input or output devices. In some embodiments, examples of member computing devices 902a-902n (e.g., clients) may be any type of processor-based platform connected to network 906, such as, but not limited to, personal computers, digital assistants, personal digital assistants, smartphones, pagers, digital tablets, laptop computers, Internet appliances, and other processor-based devices. In some embodiments, member computing devices 902a-902n may be specially programmed with one or more application programs according to one or more principles / methods detailed herein. In some embodiments, member computing devices 902a-902n may operate on any operating system capable of supporting a browser or browser-enabled applications, such as Microsoft®, Windows®, and / or Linux®. In some embodiments, the illustrated member computing devices 902a-902n may include personal computers running a browser application program such as, for example, Microsoft Corporation's Internet Explorer®, Apple Computer, Inc.'s Safari®, Mozilla Firefox, and / or Opera. In some embodiments, member computing client devices 902a-902n, users 912a, 912b-912n may communicate with each other and / or with other systems and / or devices coupled to network 906 through exemplary network 906. As shown in FIG. 19 , exemplary server devices 904 and 913 may include processor 905 and memory 917, processor 914 and memory 916, respectively.In some embodiments, server devices 904 and 913 may be coupled to network 906. In some embodiments, one or more of member computing devices 902a-902n may be mobile clients.

[0089] In some embodiments, at least one of the exemplary databases 907 and 915 may be any type of database, including a database managed by a database management system (DBMS). In some embodiments, the exemplary DBMS-managed database may be specially programmed as an engine that controls the organization, storage, management, and / or retrieval of data in the respective database. In some embodiments, the exemplary DBMS-managed database may be specially programmed to provide the ability to query, back up and replicate, enforce rules, provide security, perform calculations, change and access logging, and / or perform automatic optimizations. In some embodiments, the exemplary DBMS-managed database may be selected from Oracle Database, IBM DB2, Adaptive Server Enterprise, FileMaker, Microsoft Access, Microsoft SQL Server, MySQL, PostgreSQL, and NoSQL implementations. In some embodiments, the exemplary DBMS-managed database may be specially programmed to define each respective schema of each database within the exemplary DBMS according to a particular database model of the present disclosure, which may include a hierarchical model, a network model, a relational model, an object model, or any other suitable mechanism that can result in one or more applicable data structures that may include fields, records, files, and / or objects. In some embodiments, the exemplary DBMS-managed database may be specially programmed to include metadata about the data stored therein.

[0090] In some embodiments, the exemplary inventive computer-based systems / platforms, exemplary inventive computer-based devices, and / or exemplary inventive computer-based components of the present disclosure may be specially programmed to operate in a cloud computing / architecture 925, such as, but not limited to, infrastructure as a service (IaaS) 1110, platform as a service (PaaS) 1108, and / or software as a service (SaaS) 1106, using a web browser, mobile app, thin client, terminal emulator, or other endpoint 1104. Figures 20 and 21 show diagrams of exemplary implementations of cloud computing / architecture(s) in which the exemplary inventive computer-based systems / platforms, exemplary inventive computer-based devices, and / or exemplary inventive computer-based components of the present disclosure may be specially configured to operate.

[0091] In some embodiments, a patient's pathology information can be compared to the individual history of the same patient to determine progression (regression). A patient's progression (regression) can also be compared to other population cohorts and their past progression (regression).

[0092] The server may implement a machine learning algorithm using one or more neural networks. The machine learning algorithm may be a logistic regression, a variational autoencoding, a convolutional neural network, or a neural network. originOther statistical techniques may be used to identify and differentiate disease-related pathologies. Machine learning algorithms may also use a priori validated spectral scattering models, other scattering models, or optical physics models. The neural network may include multiple layers, some of which are defined and some of which are undefined (or hidden). The neural network is a supervised learning neural network.

[0093] In some examples, a neural network may include a neural network input layer, one or more neural network intermediate hidden layers, and a neural network output layer. Each neural network layer includes multiple nodes (or neurons). The nodes of a neural network layer are typically connected in series. The output of each node of a given neural network layer is connected to the input of one or more nodes of the subsequent neural network layer. Each node is a logic programming unit that implements an activation function (also known as a transfer function) to transform or manipulate data based on its inputs, weights (if any), and bias factor(s), if any, to generate an output. Each node's activation function produces a particular output in response to specific input(s), weights (if any), and bias factor(s). Each node's input may be a scalar, vector, matrix, object, data structure, and / or other item or reference thereto. Each node may store its respective activation function, weights (if any), and bias factor(s), independently of other nodes. In some example embodiments, the determination of one or more output nodes of the neural network output layer may be calculated or determined using a scoring function and / or a decision tree function using previously determined weights and bias factors, as understood in the art.

[0094] In some instances, the classification (the output of the second neural net) indicates whether the subject is origin Pathology or Neurology origin Presence of pathology or neurological origin The conclusion may be one or more conclusions about whether a neuropathology has been pre-screened and requires further investigation. origin The pathology conclusion can be based on one or more pathologies, where the one or more pathologies are classified by a neural network, determined or calculated using, for example, a combined weighted score, a scorecard, or a probabilistic determination. For example, the presence or probabilistic classification of both amyloid-beta and tau neurofibrillary tangles can be used to determine the pathology. origin This may lead to a highly probabilistic conclusion of pathology. In some instances, the conclusion may also be based on changes in the patient's pathology over time, for example, by comparison with the patient's past spectroscopy information. In some instances, the hyperspectral reflectance information is also used as input to a neural network, which may then origin Further aid in classifying pathology.

[0095] From the above description, it will be apparent that variations and modifications may be made to the disclosed embodiments in order to adapt them to various uses and conditions, and such embodiments are also within the scope of the appended claims.

[0096] The recitation of a list of elements in any definition of a variable herein includes the recitation of that variable as any single element or combination of the listed elements. The recitation of an embodiment herein includes that embodiment as any single embodiment or in combination with any other embodiment or portion thereof.

[0097] All patents and publications mentioned in this specification are herein incorporated by reference to the same extent as if each individual patent or publication was specifically and individually indicated to be incorporated by reference.

Claims

1. a light source configured to illuminate a retina of the eye with light; one or more imaging devices configured to receive light returned from the retina and generate one or more spatio-spectral images of the retina; a computing device configured to: acquire a retinal image mosaic comprising one or more spatio-spectral images from one or more regions of the retina; analyze the retinal image mosaic with one or more neural networks to extract segmented blood vessels of the retina from the retinal image mosaic; evaluate the segmented blood vessels; identify one or more biomarkers indicative of neurogenic pathology at or along walls of the segmented blood vessels; and generate a digital representation indicative of the presence or absence of biomarkers at or along walls of the segmented blood vessels; Including, The neurogenic pathology is selected from the group consisting of Alzheimer's disease, Parkinson's disease, amyotrophic lateral sclerosis, multiple sclerosis, prion diseases, motor neuron diseases (MND), Huntington's disease (HD), spinocerebellar ataxia (SCA), spinal muscular atrophy (SMA), and cerebral amyloid angiopathy (CAA); Imaging system.

2. The imaging system of claim 1 , wherein the one or more imaging devices include a spectral sensor.

3. The imaging system of claim 2 , wherein the spectral sensor comprises a hyperspectral sensor or a multispectral sensor.

4. The imaging system of claim 1 , wherein the biomarker comprises amyloid formation or tau protein formation.

5. The imaging system of claim 1 , further comprising a retinal observation device, wherein the one or more imaging devices and the light source are integrated into the retinal observation device.

6. The imaging system according to claim 5 , wherein the retinal observation device is a fundus camera.

7. The imaging system of claim 1 , wherein for each of the one or more regions of the retina, the one or more spatio-spectral images include spectral images at multiple wavelength ranges.

8. The imaging system of claim 7 , wherein the spectral images include spatial information about corresponding regions of the retina.

9. The imaging system of claim 8 , wherein the spatial information includes texture, structure, and pattern in the corresponding region of the retina.

10. The imaging system of claim 1 , wherein the analysis of the retinal image mosaic uses a pixel-by-pixel analysis of the one or more spatio-spectral images.

11. a light source configured to illuminate a retina of the eye with light; one or more imaging devices configured to receive light returned from the retina and generate at least one spectral image and at least one spatial image of the retina; a computing device configured to: acquire a retinal image mosaic comprising at least one spectral image and at least one spatial image of one or more regions of the retina; extract segmented blood vessels of the retina from the retinal image mosaic; evaluate the segmented blood vessels to identify one or more biomarkers indicative of a neurogenic disease along walls of the segmented blood vessels; and generate a digital representation indicative of the presence or absence of biomarkers along walls of the segmented blood vessels; Including, The neurogenic disease is selected from the group consisting of Alzheimer's disease, Parkinson's disease, amyotrophic lateral sclerosis, multiple sclerosis, prion disease, motor neuron disease (MND), Huntington's disease (HD), spinocerebellar ataxia (SCA), spinal muscular atrophy (SMA), and cerebral amyloid angiopathy (CAA); Imaging system.

12. 12. The imaging system of claim 11, wherein the one or more imaging devices include a spatial camera configured to generate the at least one spatial image and a hyperspectral camera configured to generate the at least one spectral image.

13. The imaging system of claim 11 , wherein the one or more imaging devices include a hyperspectral camera configured to generate an image comprising the spectral image and the spatial image.

14. The imaging system of claim 11 , wherein the biomarker comprises amyloid formation or tau protein formation.

15. a light source configured to illuminate a retina of the eye with light; one or more imaging devices configured to receive light returned from the retina and generate one or more spatio-spectral images of the retina; a computing device configured to acquire one or more spatio-spectral images from one or more regions of the retina, extract segmented blood vessels of the retina from the one or more spatio-spectral images of the retina, analyze the segmented blood vessels to identify one or more biomarkers indicative of neurogenic pathology at or along walls of the segmented blood vessels, and generate a digital representation indicative of the presence or absence of the biomarkers at or along walls of the segmented blood vessels; Including, The neurogenic pathology is selected from the group consisting of Alzheimer's disease, Parkinson's disease, amyotrophic lateral sclerosis, multiple sclerosis, prion diseases, motor neuron diseases (MND), Huntington's disease (HD), spinocerebellar ataxia (SCA), spinal muscular atrophy (SMA), and cerebral amyloid angiopathy (CAA); Imaging system.

16. 10. The imaging system of claim 1, wherein the computing device comprises a convolutional neural network configured to receive the one or more spatio-spectral images and automatically extract retinal blood vessels from the one or more spatio-spectral images.

17. The imaging system of claim 1 , wherein the computing device is configured to output two probability maps indicative of retinal vessel segmentation.

18. 18. The imaging system of claim 17, wherein the first probability map is for an artery and the second probability map is for a vein.

19. 20. The imaging system of claim 17, wherein the two probability maps are images, and the intensity of each pixel in the two probability maps indicates the presence of a blood vessel in the image.

20. The imaging system of claim 17 , wherein the computing device is configured to binarize the two probability maps by a threshold.

21. 21. The imaging system of claim 20, wherein the computing device is configured to process the binarized probability map to identify the structure of the blood vessel and spectral features along the blood vessel and detect the one or more biomarkers indicative of the neurogenic pathology.

22. The imaging system of claim 1 , wherein the digital representation is a heat map overlaid on the retinal image mosaic.

23. 23. The imaging system of claim 22, wherein the heat map is configured to show one or more hot spots in the form of a spatio-spectral pattern indicative of protein formation.

24. The system of claim 1, wherein the computing device is further configured to analyze the structure of the segmented blood vessels to detect one or more biomarkers indicative of the neurogenic pathology as spectral features along the segmented blood vessels.

Images