Systems and methods for detection of ocular disease

Abstract

To provide systems and methods for detection of ocular disease.SOLUTION: This invention generally relates to computer systems for ocular imaging and disease evaluation. Disclosed herein are computer systems in part for image processing. Also disclosed herein are systems for processing ocular images of multiple imaging modalities to detect ocular diseases. Also disclosed herein are methods comprising systems described herein.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates generally to computer systems for ocular imaging and disease assessment.

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. Provisional Patent Application No. 62 / 407,875, filed October 13, 2016, U.S. Provisional Patent Application No. 62 / 470,624, filed March 13, 2017, and U.S. Provisional Patent Application No. 62 / 515,475, filed June 5, 2017. The entire contents of the above-mentioned patent applications are incorporated herein by reference. [Background technology]

[0003] Because ocular biopsies and evaluation of systemic (e.g., non-ocular) samples such as blood or urine are not routine ophthalmological care, ophthalmologists and vision specialists rely on in vivo imaging for disease diagnosis, classification, and stratification, patient management, and clinical trial design.

[0004] However, for many potentially blinding eye diseases, it is currently not possible to adequately characterize disease features or complex phenotypes using image-based biomarkers. Thus, it is often not possible to adequately diagnose, stage, triage, prognose, monitor, or predict response to treatments or interventions, or predict the safety of those treatments or interventions. New image-based computational analysis tools offer such opportunities.

[0005] As a non-limiting example, it is currently not possible to easily identify patients at imminent risk of progressing to blinding late-onset age-related macular degeneration (AMD), and therefore studies aimed at reducing the development of geographic atrophy (GA) or choroidal neovascularization (CNV) (e.g., progression from early to late disease) are not feasible. Furthermore, because central vision is not an appropriate endpoint for studies of dry AMD, phase III clinical trial designs are typically limited to enrolling patients with pre-existing patches of geographic atrophy (GA) of a defined size range, and prospectively measuring and comparing the rate of expansion of geographic atrophy in study and control eyes. Such endpoints address the disease only after late effects have developed. There is a clear unmet need to develop better methods to classify patients with blinding eye diseases using image-based biomarkers both qualitatively (e.g., describing new phenotypes) and quantitatively (e.g., measuring the extent, amount, en face two-dimensional (2D) area or three-dimensional (3D) volume of disease features, and dynamic changes in these features over time). Summary of the Invention [Problem to be solved by the invention]

[0006] Disclosed herein is a computing system for storage and analysis of a plurality of ocular images. In some embodiments, the computing system may include a computer processor. In some embodiments, the computing system may include a communication interface configured to import a plurality of ocular images, the plurality of ocular images corresponding to a plurality of different imaging modalities. In some embodiments, the computing system may include a database configured to store the plurality of ocular images. In some embodiments, the computing system may include a computer memory including a computer program. In some embodiments, the computer program may be executable by the computer processor to perform one or more of: (i) controlling the communication interface to import the plurality of ocular images, (ii) processing the plurality of ocular images, and (iii) controlling the database to store the plurality of ocular images. In some embodiments, the plurality of different imaging modalities may include delayed near-infrared analysis (DNIRA). In some embodiments, the plurality of different imaging modalities may include, by way of non-limiting examples, one or more of infrared reflectance (IR), infrared autofluorescence (IR-AF), near-infrared autofluorescence (NIR-AF), confocal laser scanning ophthalmoscopy (cSLO), fundus autofluorescence (FAF), color fundus photography (CFP), optical coherence tomography (OCT), and optical coherence tomography angiography (OCT-A). In some embodiments, the plurality of different imaging modalities may include a functional modality. In some embodiments, the functional imaging modality may include, by way of non-limiting examples, microperimetry, visual field testing, electroretinography (ERG), scotopic imaging (DA), or low-light visual acuity testing. In some embodiments, the plurality of eye images may include a plurality of different file formats. In some embodiments, the database may include a local storage unit.In some embodiments, the database may include an external storage unit communicatively coupled to the computer processor via a computer network. In some embodiments, the multiple ocular images may be acquired from the subject during a single clinical visit. In some embodiments, the multiple ocular images may be acquired from the subject during multiple clinical visits. In some embodiments, the multiple ocular images may be acquired from multiple subjects. In some embodiments, the process may include generating image metadata for each of the multiple ocular images. In some embodiments, the image metadata may include a timestamp and a tag for future reference. In some embodiments, the process may include generating a composite image from the multiple ocular images. In some embodiments, the process may include generating a pseudo-composite image from the multiple ocular images. In some embodiments, the computer program may include a graphic user interface (GUI) configured to enable user selection of the multiple analysis modules. In some embodiments, the multiple analysis modules may perform image analysis on the multiple ocular images. In some embodiments, the multiple analysis modules may control a database to store the multiple ocular images. In some embodiments, the multiple analysis modules may include a pre-processing module configured to normalize and register the multiple ocular images, thereby generating a multiple pre-processed ocular images. In some embodiments, the plurality of analysis modules may include a segmentation module configured to perform a segmentation analysis of the plurality of preprocessed ocular images. In some embodiments, the segmentation analysis may be based on one or more features selected from the group consisting of, by way of non-limiting example, circularity, granularity, area, location relative to the fovea, location relative to the optic nerve head, location relative to the periphery, number, clustering, thickness, degree of hyperfluorescence or hypofluorescence, degree of hyporeflectance or hyporeflectance, edge softness or hardness, confluence between or association of individual features, and presence or absence of features known in other modalities.In some embodiments, the known features of other modalities may include, by way of non-limiting example, features selected from the group consisting of blood or blood vessels, fluid, small hard drusen, medium drusen, large drusen, confluent drusen, serous, drusenoid, or hemorrhagic RPE detachment, areas of geographic atrophy, active or static choroidal neovascularization, nascent geographic atrophy, geographic atrophy, disintegrating pigment epithelial detachment, basement membrane deposits, basal linear deposits, subretinal hyperreflective material, hyperreflective material, hypofluorescent FAF, hyperfluorescent FAF, hyporeflective IR or hyperreflective IR-AF, hypofluorescent or hyperfluorescent NIR-AF, and scar. In some embodiments, the segmentation analysis can be mapped and / or masked on images of the subject's eye obtained at a subsequent clinical visit. In some embodiments, the segmentation analysis can be used to quantify size changes on images of the subject's eye obtained at a subsequent clinical visit. In some embodiments, the segmentation analysis may include storing the segmented ocular images in a database. In some embodiments, the segmentation analysis may include displaying the segmented ocular images. In some embodiments, the plurality of analysis modules may include a delta module configured to perform a delta analysis of the plurality of preprocessed ocular images. In some embodiments, the delta analysis may include determining changes between preprocessed ocular images acquired from a single subject during different clinical visits. In some embodiments, the delta analysis may include determining changes between preprocessed ocular images acquired from a single subject corresponding to a plurality of different imaging modalities. In some embodiments, the delta analysis may include generating an ocular difference image representative of changes between the preprocessed ocular images. In some embodiments, the delta analysis may include storing the ocular difference image in a database. In some embodiments, the delta analysis may include displaying the ocular difference image. In some embodiments, the ocular difference image may be displayed as an overlay on a preprocessed ocular image acquired at an earlier clinical visit.In some embodiments, the plurality of analysis modules may include a wide-area image analytics module configured to perform wide-area image analysis of a plurality of pre-processed ocular images, the plurality of ocular images may be acquired from a plurality of subjects. In some embodiments, the wide-area image analysis may include integrating quantitative features extracted from the plurality of pre-processed ocular images. In some embodiments, the integrated quantitative features may be used to identify and quantify the burden of a degenerative ocular disease. In some embodiments, the integrated quantitative features may be used to identify, characterize, treat, or prevent the onset of reversible or irreversible ocular tissue loss caused by an ocular disease. In some embodiments, the prevention may include performing ocular or systemic (non-ocular) application of one or more of gene therapy, cell-based therapy, laser-based therapy, and delivery of a pharmacological or biological therapy. In some embodiments, the integrated quantitative features can be processed by cognitive technology configured to perform one or more of: (i) phenotyping, genotyping, and / or epigenetic cognitive analysis; (ii) biomarker identification; (iii) estimating probability of ocular disease; (iv) estimating probability of efficacy of treatment for ocular disease; (v) estimating probability for recommending treatment for ocular disease; and (vi) estimating probability for recommending clinical trial enrollment; (vii) estimating likelihood of responding to treatment for ocular disease; and (viii) estimating safety of treatment for ocular disease.

[0007] In some embodiments, the plurality of analysis modules may include an analysis of unstructured data configured to perform an analysis of the plurality of ocular images. In some embodiments, the analysis of unstructured data, i.e., data that is not segmentable, may be based on selected and unselected features. These may include, as non-limiting examples, circularity, granularity, area, location relative to the fovea, location relative to the optic nerve head, location relative to the periphery, number, clustering, thickness, degree of hyperfluorescence or hypofluorescence, degree of hyporeflectivity or hyporeflectivity, edge softness or hardness, confluence between or combination of individual features, complex en face patterns, arrays, distributions, and the presence or absence of known features in other modalities. Methods may include pattern recognition, signal processing, statistical analysis, deep learning, cognitive computing, convolutional neural networks (CNN), and artificial intelligence. In some embodiments, the computer image analysis is semi-automated, fully automated, supervised, or unsupervised. In some embodiments, the computer image analysis may include features selected from the group consisting of, by way of non-limiting example, blood or blood vessels, fluid, small hard drusen, medium drusen, large drusen, confluent drusen, serous, drusenoid, or hemorrhagic RPE detachment, areas of geographic atrophy, active or static choroidal neovascularization, nascent geographic atrophy, geographic atrophy, disintegrating pigment epithelial detachment, basement membrane deposits, basal linear deposits, subretinal hyperreflective material, hyperreflective material, hypofluorescent FAF, hyperfluorescent FAF, hyporeflective IR or hyperreflective IR-AF, hypofluorescent or hyperfluorescent NIR-AF, and scar, along with known features of other modalities. In some embodiments, the segmentation analysis may be mapped and / or masked onto images of the subject's eye obtained at subsequent clinical visits. In some embodiments, the segmentation analysis may be used to quantify changes in size of images of the subject's eye obtained at subsequent clinical visits.In some embodiments, the segmentation analysis may not match known features of other modalities and may include, by way of non-limiting example, features selected from the group consisting of blood or blood vessels, fluid, small hard drusen, medium drusen, large drusen, confluent drusen, serous, drusenoid, or hemorrhagic RPE detachment, areas of geographic atrophy, active or static choroidal neovascularization, nascent geographic atrophy, geographic atrophy, disintegrating pigment epithelial detachment, basement membrane deposits, basal linear deposits, subretinal hyperreflective material, hyperreflective material, hypofluorescent FAF, hyperfluorescent FAF, hyporeflective IR or hyperreflective IR-AF, hypofluorescent or hyperfluorescent NIR-AF, and scar. In some embodiments, the segmentation analysis may be mapped and / or masked onto images of the subject's eye obtained at subsequent clinical visits. In some embodiments, the segmentation analysis may be used to quantify changes in size of images of the subject's eye obtained at subsequent clinical visits.

[0008] Also disclosed herein is a computer-implemented method for storing and analyzing a plurality of ocular images. In some embodiments, the method may include using a computer processor to import a plurality of ocular images corresponding to a plurality of different imaging modalities. In some embodiments, the method may include using a computer processor to process the plurality of ocular images. In some embodiments, the method may include using a computer processor to store the plurality of ocular images in a database. In some embodiments, the plurality of different imaging modalities may include delayed near-infrared analysis (DNIRA). In some embodiments, the plurality of different imaging modalities may include, by way of non-limiting examples, one or more of infrared reflectance (IR), infrared autofluorescence (IR-AF), near-infrared autofluorescence (NIR-AF), confocal laser scanning ophthalmoscopy (cSLO), fundus autofluorescence (FAF), color fundus photography (CFP), optical coherence tomography (OCT), and optical coherence tomography angiography (OCT-A). In some embodiments, the plurality of different imaging modalities may include a functional modality. In some embodiments, the functional modalities may include, by way of non-limiting examples, microperimetry, visual field testing, electroretinography (ERG), scotopic measurement (DA), or low-light visual acuity testing. In some embodiments, the plurality of ocular images may include a plurality of different file formats. In some embodiments, the database may include a local storage unit. In some embodiments, the database may include an external storage unit communicatively coupled to the computer processor via a computer network. In some embodiments, the plurality of ocular images may be acquired from a subject during a single clinical visit. In some embodiments, the plurality of ocular images may be acquired from a subject during multiple clinical visits. In some embodiments, the plurality of ocular images may be acquired from multiple subjects. In some embodiments, the processing may include generating image metadata for each of the plurality of ocular images. In some embodiments, the image metadata may include a timestamp and a tag for future reference.In some embodiments, the processing may include generating a composite image from the plurality of ocular images. In some embodiments, the processing may include generating a pseudo-composite image from the plurality of ocular images. In some embodiments, the method may further include using a graphic user interface (GUI) to select from a plurality of analysis processes. In some embodiments, the analysis process may perform image analysis on the plurality of ocular images. In some embodiments, the analysis process may control a database to store the plurality of ocular images. In some embodiments, the plurality of analysis processes may include pre-processing, which may include normalizing and registering the plurality of ocular images, thereby generating a plurality of pre-processed ocular images. In some embodiments, the plurality of analysis processes may include segmentation analysis of the plurality of pre-processed ocular images. In some embodiments, the segmentation analysis may be based on one or more features selected from the group consisting of, by way of non-limiting example, circularity, granularity, area, location relative to the fovea, location relative to the optic nerve head, location relative to the periphery, number, clustering, thickness, degree of hyper- or hypo-fluorescence, degree of hyper- or hypo-reflectivity, edge softness or hardness, confluence between or bonding of individual features, and presence or absence of features known in other modalities. In some embodiments, the known features of the other modalities may include features selected from the group consisting of blood or blood vessels, fluid, small hard drusen, medium drusen, large drusen, confluent drusen, serous, drusenoid, or hemorrhagic RPE detachment, areas of geographic atrophy, active or static choroidal neovascularization, nascent geographic atrophy, geographic atrophy, disintegrating pigment epithelial detachment, basement membrane deposits, basal linear deposits, subretinal hyperreflective material, hyperreflective material, hypofluorescent or hyperfluorescent FAF, hyporeflective or hyperreflective IR, hyporeflective or hyperreflective NIR, and scar. In some embodiments, the segmentation analysis can be mapped and / or masked on images of the subject's eye obtained at a subsequent clinical visit. In some embodiments, the segmentation analysis can be used to quantify size changes in images of the subject's eye obtained at a subsequent clinical visit.In some embodiments, the segmentation analysis may include storing the segmented ocular images in a database. In some embodiments, the segmentation analysis may include displaying the segmented ocular images. In some embodiments, the multiple analysis processes may include a delta analysis of the multiple preprocessed ocular images. In some embodiments, the delta analysis may include determining changes between preprocessed ocular images acquired from a single subject during different clinical visits. In some embodiments, the delta analysis may include determining changes between preprocessed ocular images acquired from a single subject corresponding to a plurality of different imaging modalities. In some embodiments, the delta analysis may include generating an ocular difference image representative of changes between the preprocessed ocular images. In some embodiments, the delta analysis may include storing the ocular difference image in a database. In some embodiments, the delta analysis may include displaying the ocular difference image. In some embodiments, the method may further include displaying the ocular difference image as an overlay on a preprocessed ocular image acquired at an earlier clinical visit. In some embodiments, the multiple analysis processes may include wide-area image analysis of multiple pre-processed ocular images, where the multiple ocular images may be acquired from multiple subjects. In some embodiments, the wide-area image analysis may include integrating quantitative features extracted from the multiple pre-processed ocular images. In some embodiments, the method may further include using the integrated quantitative features to identify and quantify a burden of degenerative ocular disease. In some embodiments, the method may further include using the integrated quantitative features to prevent the development of reversible or irreversible ocular tissue loss caused by the ocular disease. In some embodiments, the prevention may include performing ocular or systemic (non-ocular) application of one or more of gene therapy, cell-based therapy, and delivery of a pharmacological or biological therapy.In some embodiments, the method may further include processing the integrated quantitative features by cognitive techniques to perform one or more of: (i) phenotyping, genotyping, and / or epigenetic cognitive analysis; (ii) biomarker identification; (iii) estimating probability of ocular disease; (iv) estimating probability of efficacy of treatment for ocular disease; (v) estimating probability for recommending treatment for ocular disease; and (vi) estimating probability for recommending clinical trial enrollment; (vii) estimating likelihood of responding to treatment for ocular disease; and (viii) estimating safety of treatment for ocular disease.

[0009] Also disclosed herein are methods of detecting ocular disease in a subject. In some embodiments, the method may include systemically administering a fluorescent agent to the subject. In some embodiments, the fluorescent agent may be internalized by or accumulated in a cell, cell layer, or tissue layer from the subject with the ocular disease. In some embodiments, the internalized agent may cause fluorescence in the cell, cell layer, or tissue layer, thereby producing image data. In some embodiments, the method may include analyzing the fluorescence using an algorithm. In some embodiments, the algorithm may include detecting a pattern of fluorescence from the image data from the subject. In some embodiments, the algorithm may include comparing the pattern of fluorescence from the image data from the subject to a pattern of fluorescence from a control subject or a subject with a different ocular disease. In some embodiments, the algorithm may include detecting the ocular disease if the image data from the subject has a greater or lesser degree of fluorescence than the image data from the control subject or the subject with a different ocular disease. In some embodiments, the cell or cell layer may be localized in an ocular tissue of the subject. In some embodiments, the cell may be a phagocyte. In some embodiments, the cell may be an immune cell or an antigen presenting cell, a retinal pigment epithelium cell, or a photoreceptor. In some embodiments, the cell may be a macrophage. In some embodiments, the cell may be a resident macrophage or a circulating macrophage. The macrophage may be an inflammatory macrophage, a perivascular macrophage, a parenchymal macrophage, a microglial cell, a dendritic cell, or another antigen presenting cell. In some embodiments, the internalized agent may produce fluorescence at least about 2 hours, 6 hours, 12 hours, or 24 hours after administration. In some embodiments, the internalized agent may produce fluorescence for less than about 7 days or 14 days after administration. In some embodiments, the pattern of fluorescence may include hyperfluorescence.In some embodiments, the pattern of fluorescence may include low fluorescence. In some embodiments, the fluorescent agent may be a near infrared dye. In some embodiments, the near infrared dye may be indocyanine green. In some embodiments, the administering may be ophthalmic, intraocular, vitreous, epithelial chorio, periocular, subtenon, intravenous, intraarterial, transdermal, oral, intranasal, intramuscular, subcutaneous, sublingual, buccal, or suppository. In some embodiments, the eye disease may be an eye disorder selected from the group consisting of dry age-related macular degeneration (AMD), wet AMD, reticular pseudodrusen, late-onset retinal degeneration, and any combination thereof. In some embodiments, the eye disease may be any that affects the retinal pigment epithelium (RPE), photoreceptors, macrophages, and cells of the immune system. In some embodiments, the ocular disease may be central serous retinopathy (CSR), adult vitelliform disease, uveitis, systemic diseases both primary and secondary (e.g., as non-limiting examples, sarcoid, rheumatoid disease, arthritis, etc.), white dot syndrome (including MEWDS (multiple vanishing white dot syndrome)), serpiginous choroidopathy, AMPPE (acute multifocal posterior plaque epitheliopathy), POHS (presumed ocular histoplasmosis), or serpiginous chorioretinitis. In some embodiments, the disease may be macular degeneration or cone dystrophy, such as, as non-limiting examples, Stargardt's disease. In some embodiments, the disease may be an inherited degenerative disease, such as, as, as non-limiting examples, Retinitis Pigmentosa (RP). In some embodiments, the ocular disease may be ocular melanoma, ocular tumor, or invasive tumor. In some embodiments, the algorithm may further comprise segmenting the image data based on one or more features to form segmented image data.In some embodiments, the one or more features may be selected from, by way of non-limiting examples, circularity, granularity, area, location relative to the fovea, location relative to the optic nerve head, location relative to the periphery, number, clustering, thickness, degree of hyper- or hypo-fluorescence, degree of hyper- or hypo-reflectivity, edge softness or hardness, confluence between or joining of individual features, complex two-dimensional (2D) patterns, and presence or absence of known features in other modalities. In some embodiments, the one or more features may include hyperfluorescent dots. In some embodiments, the method may further include comparing the segmented image data to second segmented image data generated at a subsequent clinical visit. In some embodiments, the method may further include comparing the segmented image data to second segmented image data generated at a previous clinical visit. In some embodiments, the method may further include comparing the segmented image data to patient risk factors, additional imaging modalities, functional modalities, or epigenetic factors. In some embodiments, the method may further include comparing the segmented image data to patient risk factors. The patient risk factors may be selected from, by way of non-limiting example, drusen, pseudodrusen, RPE detachment, pigment, hyperreflective material, basement membrane deposits, basal linear deposits, hypofluorescent FAF, hyperfluorescent FAF, geographic atrophy, choroidal neovascularization, genetic single nucleotide polymorphisms (SNPs), copy number variation (CNV), genome-wide association studies (GWAS), exome sequencing, complete genome sequencing, genomics, proteomics, transcriptomics, ocular biomarkers, systemic (blood, urine, tissue) biomarkers, environmental risk factors, or any combination thereof. In some embodiments, the method may further include comparing the segmented image data to an additional (vs. "second") imaging modality. In some embodiments, comparing the segmented image data may include delta analysis of the segmented image data. In some embodiments, the delta analysis may include determining changes between segmented image data obtained from the subject during different clinical visits.In some embodiments, the delta analysis may include determining changes between segmented image data acquired from the subject corresponding to a plurality of different imaging modalities. In some embodiments, the delta analysis may include generating a difference image of the eye representative of changes between the segmented image data. In some embodiments, the second imaging modality may be selected from, by way of non-limiting examples, infrared reflectance (IR), infrared autofluorescence (IR-AF), near-infrared autofluorescence (NIR-AF), confocal laser scanning ophthalmoscopy (cSLO), fundus autofluorescence (FAF), color fundus photography (CFP), optical coherence tomography (OCT), and optical coherence tomography angiography (OCT-A). In some embodiments, the method may further include comparing the segmented image data to a functional modality. The functional modality may be selected from, by way of non-limiting examples, microperimetry, electroretinography (ERG), visual field testing, scotopic imaging (DA), or low-light visual acuity testing. In some embodiments, the method may further include comparing the segmented image data to an epigenetic factor. The epigenetic factors may be selected from smoking, age, body mass index, obesity index, dietary intake, dietary vitamin consumption, lipid panel, cholesterol level, blood pressure, family history, concomitant medications, or pre-existing conditions. In some embodiments, the algorithm may further include a process for transforming the image data for display. In some embodiments, the method may further include pre-processing the image data by registration and normalization of the image data. In some embodiments, the method may further include post-processing the image data. In some embodiments, post-processing the data may include transforming the image data to generate output image data. In some embodiments, post-processing the data may include displaying changes in segments of the image data over time. In some embodiments, post-processing the data may include generating a graphic user interface of visual elements representing the image output data.In some embodiments, the algorithm further includes disease diagnosis, severity classification, monitoring, prognosis, or selection and prediction of treatment and its response. In some embodiments, the imaging method may be used as a surrogate biomarker for diagnosis, prognosis, disease progression, treatment selection, and prediction of treatment response or clinical trial design, or any combination thereof. In some embodiments, the imaging method can be used as a companion diagnostic for detection and treatment of ocular disease or clinical trial enrollment or exclusion. In some embodiments, the method may further include administering a treatment to the subject. In some embodiments, the treatment may include administering a drug or a pharma- ceutically acceptable salt thereof to the subject. In some embodiments, the administering may be ophthalmic, intraocular, periocular, vitreous, epithelial choroid, intravenous, intraarterial, transdermal, oral, intranasal, intramuscular, subcutaneous, sublingual, buccal, or suppository. In some embodiments, the method may further include communicating the results via a communication medium.

[0010] The present specification also provides, for example, the following items: (Item 1) 1. A computing system for storage and analysis of multiple eye images, comprising: A computer processor; a communications interface configured to import the plurality of ocular images, the plurality of ocular images corresponding to a plurality of different imaging modalities, the plurality of different imaging modalities comprising one or more of delayed near infrared analysis (DNIRA), infrared reflectance (IR), confocal laser scanning ophthalmoscopy (cSLO), fundus autofluorescence (FAF), color fundus photography (CFP), optical coherence tomography (OCT), or OCT angiography; a database configured to store the plurality of eye images; 1. A computer memory comprising: (i) controlling the communication interface to import the plurality of eye images; (ii) processing the images of the plurality of eyes; and (iii) controlling the database to store the plurality of eye images; a computer memory comprising a computer program executable by said computer processor to perform one or more of the following: A computing system comprising: (Item 2) Item 14. The computing system of item 1, wherein the plurality of different imaging modalities comprises functional modalities. (Item 3) 3. The computing system of claim 2, wherein the functional imaging modality comprises microperimetry, visual field testing, electroretinography (ERG), scotopic imaging (DA), or low-light visual acuity testing. (Item 4) 2. The computing system of claim 1, wherein the plurality of different imaging modalities comprises DNIRA and FAF. (Item 5) 2. The computing system of claim 1, wherein the plurality of eye images comprises a plurality of different file formats. (Item 6) 2. The computing system of claim 1, wherein the database comprises a local storage unit. (Item 7) 2. The computing system of claim 1, wherein the database comprises an external storage unit communicatively coupled to the computer processor via a computer network. (Item 8) Item 14. The computing system of item 1, wherein the multiple eye images are acquired from the subject during a single clinical visit. (Item 9) Item 14. The computing system of item 1, wherein the multiple eye images are acquired from a subject during multiple clinical visits. (Item 10) 2. The computing system of claim 1, wherein the multiple eye images are acquired from multiple subjects. (Item 11) 2. The computing system of claim 1, wherein the processing includes generating image metadata for each of the plurality of eye images. (Item 12) Item 12. The computing system of item 11, wherein the image metadata comprises a timestamp and a tag for future reference. (Item 13) 2. The computing system of claim 1, wherein the processing includes generating a composite image from the multiple eye images. (Item 14) 2. The computing system of claim 1, wherein the processing includes generating a pseudo-synthetic image from the multiple eye images. (Item 15) 2. The computing system of claim 1, wherein the computer program comprises a graphic user interface (GUI) configured to enable user selection of a plurality of analysis modules to (i) perform image analysis on the plurality of ocular images, or (ii) control the database to store the plurality of ocular images. (Item 16) Item 16. The computing system of item 15, wherein the plurality of analysis modules comprises a pre-processing module configured to normalize and register the plurality of ocular images, thereby generating a plurality of pre-processed ocular images. (Item 17) Item 17. The computing system of item 16, wherein the plurality of analysis modules comprises a segmentation module configured to perform a segmentation analysis of the plurality of preprocessed eye images. (Item 18) Item 18. The computing system of item 17, wherein the segmentation analysis is based on one or more features selected from the group consisting of circularity, granularity, area, location relative to the fovea, location relative to the optic nerve head, location relative to the periphery, number, clustering, thickness, degree of hyperfluorescence or hypofluorescence, hyporeflectance or degree of hyporeflectance, edge softness or hardness, confluence between or association of individual features, and presence or absence of features known in other modalities. (Item 19) 20. The computing system of claim 18, wherein the known features of other modalities include features selected from the group consisting of blood or blood vessels, fluid, small hard drusen, medium drusen, large drusen, confluent drusen, serous, drusenoid, or hemorrhagic RPE detachment, areas of geographic atrophy, active or static choroidal neovascularization, nascent geographic atrophy, geographic atrophy, disrupted pigment epithelial detachment, and scar. (Item 20) 20. The computing system of claim 17, wherein the segmentation analysis is mapped and / or masked onto an image of the subject's eye obtained at a subsequent clinical visit. (Item 21) Item 18. The computing system of item 17, wherein the segmentation analysis is used to quantify changes in size of images of the subject's eye obtained at subsequent clinical visits. (Item 22) 20. The computing system of claim 17, wherein the segmentation analysis includes storing the segmented eye image in the database. (Item 23) 20. The computing system of claim 17, wherein the segmentation analysis includes displaying the segmented eye image. (Item 24) Item 17. The computing system of item 16, wherein the plurality of analysis modules comprises a delta module configured to perform a delta analysis of the plurality of preprocessed ocular images. (Item 25) 25. The computing system of claim 24, wherein the delta analysis comprises determining changes between pre-processed ocular images obtained from a single subject during different clinical visits. (Item 26) 25. The computing system of claim 24, wherein the delta analysis comprises determining changes between pre-processed ocular images obtained from a single subject corresponding to a plurality of different imaging modalities. (Item 27) 25. The computing system of claim 24, wherein the delta analysis comprises generating an ocular difference image representative of changes between pre-processed ocular images. (Item 28) 28. The computing system of claim 27, wherein the delta analysis includes storing a difference image of the eye in the database. (Item 29) 28. The computing system of claim 27, wherein the delta analysis includes displaying a difference image of the eye. (Item 30) 30. The computing system of claim 29, wherein the difference image of the eye is displayed as an overlay on a pre-processed image of the eye acquired at an earlier clinical visit. (Item 31) Item 17. The computing system of item 16, wherein the plurality of analysis modules comprises a wide area image analytics module configured to perform wide area image analysis of the plurality of preprocessed ocular images, the plurality of ocular images being obtained from a plurality of subjects. (Item 32) Item 32. The computing system of item 31, wherein the global image analysis includes integrating quantitative features extracted from the multiple preprocessed eye images. (Item 33) 33. The computing system of claim 32, wherein the integrated quantitative features are used to identify and quantify the burden of a degenerative eye disease. (Item 34) 33. The computing system of claim 32, wherein the integrated quantitative features are used to prevent the onset of irreversible ocular tissue loss caused by a degenerative eye disease. (Item 35) 35. The computing system of claim 34, wherein the prevention comprises performing one or more of the following ocular medications: gene therapy, cell-based therapy, laser-based therapy, and local delivery of a pharmacological therapy. (Item 36) Item 33. The computing system of item 32, wherein the integrated quantitative features are processed by cognitive technology configured to perform one or more of: (i) phenotyping, genotyping, and / or epigenetic cognitive analysis; (ii) biomarker identification; (iii) estimating probability of ocular disease; (iv) estimating probability of efficacy of treatment for ocular disease; (v) estimating probability for recommending treatment for ocular disease; and (vi) estimating probability for recommending clinical trial enrollment. (Item 37) 1. A computer-implemented method for storing and analyzing a plurality of eye images, comprising: (i) using a computer processor, importing a plurality of ocular images corresponding to a plurality of different imaging modalities, the plurality of different imaging modalities comprising one or more of delayed near infrared analysis (DNIRA), infrared reflectance (IR), confocal laser scanning ophthalmoscopy (cSLO), fundus autofluorescence (FAF), color fundus photography (CFP), optical coherence tomography (OCT), or OCT angiography; (ii) processing the plurality of images using the computer processor; and (iii) using said computer processor to store said plurality of eye images in a database; A method comprising: (Item 38) 38. The method of claim 37, wherein the plurality of different imaging modalities comprises functional modalities. (Item 39) 38. The method of claim 37, wherein the plurality of different imaging modalities comprises DNIRA and FAF. (Item 40) 39. The method of claim 38, wherein the functional modality comprises microperimetry, visual field testing, electroretinography (ERG), scotopic imaging (DA), or low-light visual acuity testing. (Item 41) Item 38. The method of item 37, wherein the plurality of eye images comprises a plurality of different file formats. (Item 42) Item 38. The method of item 37, wherein the database comprises a local storage unit. (Item 43) Item 38. The method of item 37, wherein the database comprises an external storage unit communicatively coupled to the computer processor via a computer network. (Item 44) 38. The method of claim 37, wherein the multiple ocular images are obtained from the subject during a single clinical visit. (Item 45) 38. The method of claim 37, wherein the multiple ocular images are obtained from the subject during multiple clinical visits. (Item 46) 38. The method of claim 37, wherein the multiple eye images are obtained from multiple subjects. (Item 47) 40. The method of claim 37, wherein the processing includes generating image metadata for each of the plurality of eye images. (Item 48) Item 48. The method of item 47, wherein the image metadata comprises a timestamp and tags for future reference. (Item 49) 38. The method of claim 37, wherein the processing includes generating a composite image from the multiple eye images. (Item 50) 38. The method of claim 37, wherein the processing includes generating a pseudo-synthetic image from the multiple eye images. (Item 51) 40. The method of claim 37, further comprising using a graphic user interface (GUI) to select from a plurality of analysis processes: (i) performing image analysis on the plurality of ocular images; or (ii) controlling the database to store the plurality of ocular images. (Item 52) 52. The method of claim 51, wherein the multiple analysis processes include pre-processing, the pre-processing including normalizing and registering the multiple ocular images, thereby generating a multiple pre-processed ocular images. (Item 53) 53. The method of claim 52, wherein the multiple analysis processes include segmentation analysis of the multiple preprocessed eye images. (Item 54) 54. The method according to item 53, wherein the segmentation analysis is based on one or more features selected from the group consisting of circularity, granularity, area, location relative to the fovea, location relative to the optic nerve head, location relative to the periphery, number, clustering, thickness, degree of hyperfluorescence or hypofluorescence, hyporeflectance or degree of hyporeflectance, edge softness or hardness, confluence between or association of individual features, and presence or absence of features known in other modalities. (Item 55) 55. The method of claim 54, wherein the known features of other modalities include features selected from the group consisting of blood or blood vessels, fluid, small hard drusen, medium drusen, large drusen, confluent drusen, serous, drusenoid, or hemorrhagic RPE detachment, areas of geographic atrophy, active or static choroidal neovascularization, nascent geographic atrophy, geographic atrophy, disintegrating pigment epithelial detachment, and scar. (Item 56) 54. The method of claim 53, wherein the segmentation analysis is projected and / or masked onto an image of the subject's eye obtained at a subsequent clinical visit. (Item 57) 54. The method of claim 53, wherein the segmentation analysis is used to quantify changes in size of images of the subject's eye obtained at subsequent clinical visits. (Item 58) 54. The method of claim 53, wherein the segmentation analysis includes storing the segmented eye image in the database. (Item 59) 54. The method of claim 53, wherein the segmentation analysis includes displaying the segmented eye image. (Item 60) 53. The method of claim 52, wherein the multiple analysis processes include delta analysis of the multiple preprocessed ocular images. (Item 61) 61. The method of claim 60, wherein the delta analysis comprises determining the changes between pre-processed ocular images obtained from a single subject during different clinical visits. (Item 62) 61. The method of claim 60, wherein the delta analysis comprises determining changes between pre-processed ocular images obtained from a single subject corresponding to a plurality of different imaging modalities. (Item 63) Item 61. The method of item 60, wherein the delta analysis comprises generating an ocular difference image representative of changes between pre-processed ocular images. (Item 64) 64. The method of claim 63, wherein the delta analysis comprises storing a difference image of the eye in the database. (Item 65) 64. The method of claim 63, wherein the delta analysis includes displaying a difference image of the eye. (Item 66) 66. The method of claim 65, further comprising displaying the difference image of the eye as an overlay on a pre-processed eye image acquired at an earlier clinical visit. (Item 67) 53. The method of claim 52, wherein the multiple analysis processes include wide area image analysis of the multiple pre-processed ocular images, the multiple ocular images being obtained from multiple subjects. (Item 68) 68. The method of claim 67, wherein the global image analysis comprises integrating quantitative features extracted from the multiple preprocessed eye images. (Item 69) 70. The method of claim 68, further comprising using the integrated quantitative features to identify and quantify a burden of a degenerative eye disease. (Item 70) 70. The method of claim 68, further comprising using the integrated quantitative signature to prevent the onset of irreversible ocular tissue loss caused by a degenerative eye disease. (Item 71) 71. The method of claim 70, wherein said prevention comprises administering to the eye one or more of gene therapy, cell-based therapy, and local delivery of pharmacological therapy. (Item 72) 69. The method of claim 68, further comprising processing the integrated quantitative features by cognitive techniques to perform one or more of: (i) phenotyping, genotyping, and / or epigenetic cognitive analysis; (ii) biomarker identification; (iii) estimating probability of ocular disease; (iv) estimating probability of efficacy of treatment for ocular disease; (v) estimating probability for recommending treatment for ocular disease; and (vi) estimating probability for recommending clinical trial enrollment. (Item 73) 1. A method for detecting eye disease in a subject, comprising: a. systemically administering to the subject a fluorescent agent, wherein the fluorescent agent is internalized by cells from the subject having the ocular disease, and the internalized agent causes fluorescence in the cells, thereby producing image data; b. analyzing the fluorescent emissions using an algorithm, the algorithm comprising: i. detecting a pattern of fluorescence from the image data from the subject; ii. comparing the pattern of fluorescence from the image data from the subject to a pattern of fluorescence from a control subject; iii. detecting the eye disease when the image data from the subject has a greater or lesser degree of fluorescence than the image data from the control subject. The analyzing step includes: A method comprising: (Item 74) 74. The method of claim 73, wherein the cells are localized within an ocular tissue of the subject. (Item 75) 74. The method of item 73, wherein the cell is a phagocyte. (Item 76) 74. The method of claim 73, wherein the cell is a macrophage, a retinal pigment epithelial cell, or a photoreceptor. (Item 77) 77. The method of claim 76, wherein the cell is a macrophage, and the macrophage is an inflammatory macrophage, a perivascular macrophage, a parenchymal macrophage, a microglial cell, a dendritic cell, or an antigen-presenting cell. (Item 78) 74. The method of claim 73, wherein the internalized agent produces a fluorescent emission at least about 2 hours after the administering. (Item 79) 74. The method of claim 73, wherein the internalized agent produces fluorescence for less than about 7 days after the administering. (Item 80) 74. The method of claim 73, wherein the pattern of fluorescence comprises hyperfluorescence. (Item 81) 74. The method of claim 73, wherein the pattern of fluorescence comprises low fluorescence. (Item 82) 74. The method of claim 73, wherein the fluorescent agent is a near-infrared dye. (Item 83) 83. The method of claim 82, wherein the near infrared dye is indocyanine green. (Item 84) 74. The method of item 73, wherein the administering is ophthalmic, intraocular, vitreous, intravenous, intraarterial, transdermal, buccal, intranasal, intramuscular, subcutaneous, sublingual, buccal, or suppository. (Item 85) 74. The method of claim 73, wherein the eye disease is an eye disorder selected from the group consisting of dry age-related macular degeneration (AMD), wet AMD, reticular pseudodrusen, late-onset retinal degeneration, and any combination thereof. (Item 86) 74. The method of claim 73, wherein the eye disease is ocular melanoma or an eye tumor. (Item 87) 74. The method of claim 73, wherein the algorithm further comprises segmenting the image data based on one or more features to form segmented image data. (Item 88) 88. The method of claim 87, wherein the one or more features are selected from circularity, granularity, area, location relative to the fovea, location relative to the optic nerve head, location relative to the periphery, number, clustering, thickness, degree of hyper- or hypo-fluorescence, degree of hyper- or hypo-reflectivity, edge softness or hardness, confluence between or bonding of individual features, and presence or absence of features known in other modalities. (Item 89) 88. The method of claim 87, wherein the one or more features include hyperfluorescent dots. (Item 90) 88. The method of claim 87, further comprising comparing the segmented image data to second segmented image data generated at a subsequent clinical visit. (Item 91) 88. The method of claim 87, further comprising comparing the segmented image data to a second segmented image data generated at a previous clinical visit. (Item 92) 88. The method of claim 87, further comprising comparing the segmented image data to a patient risk factor, a second imaging modality, a functional modality, or an epigenetic factor. (Item 93) 88. The method of claim 87, further comprising comparing the segmented image data to patient risk factors, wherein the patient risk factors are selected from drusen, RPE detachment, geographic atrophy, choroidal neovascularization, genetic single nucleotide polymorphisms, copy number polymorphisms, complete genome sequencing, ocular biomarkers, or systemic biomarkers, or any combination thereof. (Item 94) 88. The method of claim 87, further comprising comparing the segmented image data to a second imaging modality. (Item 95) 95. The method of claim 94, wherein the comparing of the segmented image data comprises a delta analysis of the segmented image data. (Item 96) 96. The method of claim 95, wherein the delta analysis comprises determining changes between the segmented image data obtained from the subject during different clinical visits. (Item 97) 96. The method of claim 95, wherein the delta analysis comprises determining changes between the segmented image data acquired from the subject corresponding to a plurality of different imaging modalities. (Item 98) 96. The method of claim 95, wherein the delta analysis includes generating a difference image of the eye representative of changes between the segmented image data. (Item 99) 95. The method of claim 94, wherein the second imaging modality is selected from infrared reflectance (IR), confocal laser scanning ophthalmoscopy (cSLO), fundus autofluorescence (FAF), color fundus photography (CFP), optical coherence tomography (OCT), and OCT angiography. (Item 100) 88. The method of claim 87, further comprising comparing the segmented image data to a functional modality, wherein the functional modality is selected from microperimetry, electroretinography (ERG), visual field testing, scotopic imaging (DA), or low-light visual acuity testing. (Item 101) 90. The method of claim 87, further comprising comparing the segmented image data to epigenetic factors, wherein the epigenetic factors are selected from smoking, age, body mass index, obesity index, dietary intake, dietary vitamin consumption, lipid panel, cholesterol level, blood pressure, family history, or pre-existing conditions. (Item 102) Item 74. The method of item 73, wherein the algorithm further comprises a process for transforming image data for display. (Item 103) The method further comprising: a. preprocessing the image data by registering and normalizing the image data; b. post-processing the image data, i. transforming the image data to generate output image data; ii. displaying changes in said segments of image data over time; and iii. generating a graphical user interface of visual elements representing said image output data; said post-processing comprising: Item 103. The method according to item 102, comprising: (Item 104) 74. The method of claim 73, wherein the algorithm further comprises diagnosis, severity classification, prognosis, or treatment selection. (Item 105) 74. The method of claim 73, wherein the imaging method is used as a surrogate biomarker for diagnosis, prognosis, disease progression, clinical trial design, or any combination thereof. (Item 106) 74. The method of claim 73, wherein the imaging method is used as a companion diagnostic for the detection and treatment of the eye disease. (Item 107) 74. The method of claim 73, further comprising administering a treatment to the subject. (Item 108) 108. The method of claim 107, wherein the treatment comprises administering to the subject a drug or a pharma- ceutically acceptable salt thereof. (Item 109) 109. The method of item 108, wherein said administering is ophthalmic, intraocular, vitreous, intravenous, intraarterial, transdermal, buccal, intranasal, intramuscular, subcutaneous, sublingual, buccal, or suppository. (Item 110) 100. The method of any one of items 73 to 109, further comprising communicating the results via a communication medium. (Item 111) 1. A method for assessing a subject for a blinding eye disease, comprising: systemically administering to the subject a fluorescent compound that is internalized by retinal pigment epithelial (RPE) cells or macrophages; detecting fluorescence of the RPE layer or macrophages in the subject's eye(s) at least 24 hours after administering the fluorescent compound; evaluating the fluorescence pattern of the RPE layer or macrophages for one or more characteristics indicative of a compelling ocular disease or a stage thereof; Including, The fluorescence pattern is classified as to the presence or absence or stage or type of blinding eye disease using a classifier algorithm. method. (Item 112) 112. The method of claim 111, further comprising providing treatment to a patient for a blinding eye disease, wherein the patient is determined to have a blinding eye disease or the progression thereof. (Item 113) Item 113. The method of item 112, wherein the classifier algorithm is trained using fluorescence patterns determined for an experimental group comprising subjects with a blinding eye disease or a stage thereof, and a comparator group not having the blinding eye disease. (Item 114) Item 114. The method of item 113, wherein the classifier algorithm is trained using pattern recognition and / or clustering analysis. (Item 115) Item 114. The method of item 112 or 113, wherein the classifier utilizes one or more classification schemes selected from a decision tree, a logistic regression, a principal component analysis, a naive Bayes model, a support vector machine model, and a nearest neighbor model. (Item 116) 116. The method of any one of items 111 to 115, wherein the features are determined by machine learning, optionally using a single-part classifier or anomaly detector algorithm. (Item 117) 117. The method of any one of items 111 to 116, wherein the fluorescent pattern(s) comprises a loose pattern, a tight pattern, a leopard print pattern, a grey stain pattern, an oil stain pattern, a fingering potato pattern, a bullet hole pattern, and a large bright dot pattern. (Item 118) 118. The method of any one of items 111 to 117, wherein the pattern indicates the presence of drusen or pseudodrusen in the absence of geographic atrophy, macular atrophy (MA), or choroidal neovascularization, and indicates the presence of drusen or pseudodrusen in the presence of either geographic atrophy or choroidal neovascularization. (Item 119) 119. The method of any one of items 111 to 118, wherein the pattern indicates the presence, absence, regression, or extent of drusen, pseudodrusen, fish-shaped lesions, inflammatory infiltrates, subretinal fluid, variably formed hyperpigmented or hypopigmented areas in the absence of geographic atrophy, patchy cutis, or choroidal neovascularization, or the presence of drusen, pseudodrusen, fish-shaped lesions, inflammatory infiltrates, subretinal fluid, variably formed hyperpigmented or hypopigmented areas in the presence of either geographic atrophy, patchy cutis, or choroidal neovascularization. (Item 120) 119. The method of any one of items 111 to 118, wherein the pattern indicates total number of drusen, size of drusen, presence of soft confluent drusen, hyperpigmentation, RPE detachment, hypopigmentation, hyporeflective subretinal material, hyperreflective dots, subretinal basement membrane and / or basal linear deposits. The novel features of the exemplary embodiments are set forth with particularity in the appended claims. A better understanding of the features and advantages will be obtained by reference to the following detailed description which sets forth illustrative embodiments, in which the principles of the disclosed systems and methods are utilized, and the accompanying drawings. [Brief description of the drawings]

[0011] [Figure 1] Schematic of a system for processing, storage, and retrieval of ocular images. Image acquisition is followed by cloud-based image analysis to generate a complex subject phenotype. Together with genetic and epigenetic data, this subject phenotype may feed downstream steps of biomarker development, clinical trial design improvement, drug development, new treatments, and ultimately selection from a menu of specific treatment options for personalized medicine. (QA=Quality Assurance) [Diagram 2]FIG. 1 is a schematic diagram of a system for processing, storage, and retrieval of ocular images. The system 100 connects to an image acquisition device 106 to receive ocular images of one or more modalities. The system 100 also connects to an external image storage unit 102 to access ocular images of one or more modalities. The system 100 processes the images to generate output data. In particular, the system 100 generates a graphical representation (of visual elements) of the output data for display on a user device 104. The user device 104 also provides user configurations and user parameters to the system 100 to provide feedback and input data as well as dynamically control the display to update the graphical representation based on timing and subject data. [Diagram 3] 1 is a flow diagram of a method for processing, storage, and retrieval of eye images.The system 100 includes a server 200, a database 202, a client application 204, a user device 206, and an imaging device 208. [Figure 4] Schematic diagram of a system for processing steps, analytics, and end use of image data for drug development of new therapeutics. The diagram shows the flow of data from the clinical environment through computational steps, integration with other modalities including phenotype, genomic data, and epigenetics, and finally to the end user for therapeutic development. [Diagram 5]FIG. 1 is a schematic diagram of a system for processing, storage, and retrieval of ocular images. System 100 includes a retinal image capture unit 402, a display device 450, and a server 200. Server 200 includes various hardware and software units, including a processing unit 404, a post-processing unit 408, and a global analytics unit 408. Server 200 stores ocular images received from retinal image capture unit 402. Server 200 integrates the analysis from components 406 and 408. Server 200 generates a graphical representation of output data for display as part of a GUI on display device 450. Server 200 receives commands, configurations, and parameters from display device 450. Server 200 displays the processed images on display device 450. [Figure 6] 1 is a schematic diagram of a system for the key steps of processing ocular DNIRA images. A raw image input 602 is provided to an image pre-processing unit 404 to generate qualified image data 608 by steps of registration 410 and normalization 414. An image post-processing unit 406 further transforms and processes the qualified image data 608 using segmentation 424 and feature extraction 434. A delta unit 614 generates a delta image as output data. Outputs include an image output 618 and a quantitative output 620. [Figure 7] 1 is a schematic diagram of a computing device for implementing aspects of a system for processing, storage, and retrieval of ocular images. Shown is a system 100 (including a server 200 or a user device 104, 206) for processing, storage, and retrieval of ocular images. The system includes a processor 702, a memory 704, an I / O interface 706, and a network interface 708. [Figure 8]1 is a flow diagram of a method for processing, storage and retrieval of ocular images, including the different components described herein. The diagram shows the simultaneous analysis of two images 802, 804, including the components of registration 410, which is improved through cycles of correction until sufficient quality is reached, feature extraction 810 using an algorithm developed to uniquely identify disease-related features 812, comparison 814 of the two images, and classification 816. [Figure 9] FIG. 1 illustrates the importance of image registration in sequential imaging and analysis of the eye. Registration allows for addressing slight variations in alignment, tilt, skew, magnification, and any other parameters that may change during each patient visit. An example is shown using FAF images. Additionally, this figure clearly illustrates the magnification of central GA relative to FAF, showing only those sites where tissue has died and damage is irreversible. [Figure 10] FIG. 1 shows segmentation and feature extraction of DNIRA images. In the left panel, DNIRA identifies widespread areas of hypofluorescence (central black area). In the middle panel, identification of areas of DNIRA hypofluorescence is shown. In the right panel, segmentation is used to select areas of low fluorescent DNIRA signal, allowing further feature extraction and quantification. [Figure 11-1] (FIGS. 11A to 11D) Representative DNIRA of diseased retinas. Arrows indicate areas of hypofluorescence. Yellow lines indicate the borders of areas of hypofluorescence. [Figure 11-2] (FIGS. 11A to 11D) Representative DNIRA of diseased retinas. Arrows indicate areas of hypofluorescence. Yellow lines indicate the borders of areas of hypofluorescence. [Figure 12]

[0023] Figure 1 shows a representative DNIRA image compared to corresponding IR, FAF, and OCT images. The left panel is the IR, the middle is the FAF image, and the right panel is the DNIRA image. The bottom panel shows the corresponding OCT. [Figure 13]Representative DNIRA images compared to corresponding IR, FAF, and fundus photographs. Top left is a fundus photograph showing drusen, bottom left is the corresponding DNIRA image showing the black areas where drusen are located. Right panel shows IR (top) and FAF (bottom) where drusen are usually difficult to detect. [Figure 14] Representative DNIRA images of a patient with central GA observed on FAF (left). The DNIRA image (right) shows a greater number of hypofluorescent signals, a greater total area of ​​hypofluorescence, and a greater amount of periphery / border of hypofluorescence. The difference between these two images represents the "delta" portion of the signal. [Figure 15A] (FIGS. 15A and 15B) Representative output data as a visual representation comparing DNIRA to an underlying image showing RPE detachment on OCT. [Figure 15B] (FIGS. 15A and 15B) Representative output data as a visual representation comparing DNIRA to an underlying image showing RPE detachment on OCT. [Figure 16] FIG. 13 shows image output data as a visual representation comparing IR, FAF, and OCT to DNIRA. [Figure 17A] (Figures 17A and 17B) Representative images of quantification of total hypofluorescent DNIRA area as a marker of progressive or dynamic changes, and comparison of DNIRA signal across multiple time points / sessions to observe changes associated with increasing or decreasing DNIRA signal. [Figure 17B] (Figures 17A and 17B) Representative images of quantification of total hypofluorescent DNIRA area as a marker of progressive or dynamic changes, and comparison of DNIRA signal across multiple time points / sessions to observe changes associated with increasing or decreasing DNIRA signal. [Figure 18] FIG. 1 shows a comparison of DNIRA versus color photography and clinical grading of uveal melanoma cases. [Figure 19-1](FIGS. 19A-19J) Analysis of tumor-associated macrophages (TAMs) using DNIRA. DNIRA images were collected into a composite image and correlated with other modalities including fundus imaging, FAF, and OCT. Regions of interest were selected, analyzed, and compared in patients with melanoma, indeterminate lesions, or benign nevi using imaging and quantification techniques. ANOVA and Bonferroni multiple comparisons of site dot density were analyzed across groups. [Figure 19H] (FIGS. 19A-19J) Analysis of tumor-associated macrophages (TAMs) using DNIRA. DNIRA images were collected into a composite image and correlated with other modalities including fundus imaging, FAF, and OCT. Regions of interest were selected, analyzed, and compared in patients with melanoma, indeterminate lesions, or benign nevi using imaging and quantification techniques. ANOVA and Bonferroni multiple comparisons of site dot density were analyzed across groups. [Figure 19I] (FIGS. 19A-19J) Analysis of tumor-associated macrophages (TAMs) using DNIRA. DNIRA images were collected into a composite image and correlated with other modalities including fundus imaging, FAF, and OCT. Regions of interest were selected, analyzed, and compared in patients with melanoma, indeterminate lesions, or benign nevi using imaging and quantification techniques. ANOVA and Bonferroni multiple comparisons of site dot density were analyzed across groups. [Figure 19J] (FIGS. 19A-19J) Analysis of tumor-associated macrophages (TAMs) using DNIRA. DNIRA images were collected into a composite image and correlated with other modalities including fundus imaging, FAF, and OCT. Regions of interest were selected, analyzed, and compared in patients with melanoma, indeterminate lesions, or benign nevi using imaging and quantification techniques. ANOVA and Bonferroni multiple comparisons of site dot density were analyzed across groups. [Figure 20] FIG. 1 shows comparative retinal images obtained by IR imaging, FAF imaging, and DNIRA imaging to detect mid-gray signals in the DNIRA image. Unlike FAF, the DNIRA image reveals abnormal or "diseased" RPE as a darker gray color. [Figure 21-1] (FIGS. 21A-21E) Examples of patterns with interwoven, lacy, net-like, or speckled configurations. In some examples, patterns are observed that are coarser (wider) or finer (closer) in aspect. These patterns may indicate different subtypes of AMD and therefore different responses to treatment options. [Figure 21-2] (FIGS. 21A-21E) Examples of patterns with interwoven, lacy, net-like, or speckled configurations. In some examples, patterns are observed that are coarser (wider) or finer (closer) in aspect. These patterns may indicate different subtypes of AMD and therefore different responses to treatment options. [Figure 22] Structural and functional aspects of DNIRA signal in RPE. The top panel shows the relative intensity of DNIRA signal to the corresponding tissue changes in the bottom panel. In the bottom panel, normal RPE, Bruch's membrane (BrM), and choroidal vasculature are shown with commonly observed changes occurring in dry AMD. Hypofluorescence is observed in association with GA, RPE detachment, and drusen or other subretinal pigment epithelial deposits (basal linear, basement membrane deposits). Mid-gray signal is shown in areas where the RPE is in its correct anatomical position relative to the choroidal vessels, but where cellular metabolism is presumed to be abnormal. [Diagram 23]Structural and functional aspects of DNIRA signal in macrophages. The relative intensity of DNIRA signal is shown in the upper panel against the corresponding tissue changes in the lower panel. Hyperfluorescent dots (of a specific size, shape, and motility) associated with DNIRA signal are identified as phagocytic immune cells, i.e. macrophages. Green arrows indicate the transfer / uptake of ICG dye from the choroidal circulation to the retina / RPE complex. The lower panel shows the presence of phagocytic macrophages absorbing the dye. [Figure 24] FIG. 1 shows a representative example of hypofluorescent DNIRA total area as a marker of disease and disease burden (diagnostic or prognostic biomarker). [Diagram 25] FIG. 1 shows a representative example of percent change in hypofluorescent DNIRA total area as a marker of prognosis or response to intervention. [Figure 26A] (FIG. 26A and FIG. 26B) Representative examples of DNIRA signatures for identifying early stage disease and distinct phenotypes of early stage disease (diagnostic biomarkers). [Figure 26B] (FIG. 26A and FIG. 26B) Representative examples of DNIRA signatures for identifying early stage disease and distinct phenotypes of early stage disease (diagnostic biomarkers). [Figure 27] (FIG. 27A and FIG. 27B) Representative examples of DNIRA features for distinguishing diseases of similar clinical presentation to aid in diagnostic accuracy (diagnostic biomarkers). [Figure 28A] (FIG. 28A and FIG. 28B) Not Yet Diagnosed (NYD) DNIRA-characterized genotypes of monogenic and monogenic disorders. [Figure 28B] (FIG. 28A and FIG. 28B) Not Yet Diagnosed (NYD) DNIRA-characterized genotypes of monogenic and monogenic disorders. [Figure 29] FIG. 1 shows representative examples of DNIRA features for monitoring disease progression or the effect of treatment over time (monitoring biomarkers). [Diagram 30]FIG. 1 shows representative examples of DNIRA features for monitoring disease progression compared within and across modalities and other imaging biomarkers over time (monitoring and prognostic biomarkers). [Diagram 31] FIG. 1 shows representative examples of DNIRA signatures as biomarkers for identifying patients likely to progress to late stage disease (prognostic biomarkers). [Diagram 32] FIG. 1 shows representative examples of DNIRA features as biomarkers for quantifying known disease aspects to predict progression (prognostic biomarkers). [Diagram 33] FIG. 1 shows a representative example of using DNIRA to correlate with disease pathogenesis (predictive biomarker). [Figure 34A] (FIG. 34A to FIG. 34D) Representative example of DNIRA for detecting multiple types of hyperfluorescent dot signals (predictive biomarkers). [Figure 34B] (FIG. 34A to FIG. 34D) Representative example of DNIRA for detecting multiple types of hyperfluorescent dot signals (predictive biomarkers). [Figure 34C] (FIG. 34A to FIG. 34D) Representative example of DNIRA for detecting multiple types of hyperfluorescent dot signals (predictive biomarkers). [Fig. 34D] (FIG. 34A to FIG. 34D) Representative example of DNIRA for detecting multiple types of hyperfluorescent dot signals (predictive biomarkers). [Diagram 35] FIG. 1 shows a representative example of DNIRA for detecting static and dynamic hyperfluorescent dot signals (predictive biomarkers). [Diagram 36] FIG. 1 shows a representative example of DNIRA to detect macrophage activity at sites of tissue damage and central serous retinopathy (predictive biomarker, prognostic biomarker). [Figure 37]FIG. 1 shows a representative example of DNIRA for detecting two populations of bright hyperfluorescent dots in diffuse trickling AMD (diagnostic biomarker, prognostic biomarker). [Figure 38] FIG. 1 shows a representative example of DNIRA for detecting reticular pseudodrusen. [Figure 39] Representative example of DNIRA applied to the analysis of inflammatory eye disease. In this example of presumed acute multifocal posterior plaque-like epitheliopathy (AMPPE), DNIRA demonstrates abnormal RPE / outer retinal layer pigment uptake. [Diagram 40] FIG. 1 shows the use of DNIRA as a marker of high-risk AMD characteristics to identify progression from early to late disease. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] overview Ocular diseases or disorders that may reduce or eliminate a person's vision among other effects are of major medical concern. For example, age-related macular degeneration (AMD) is the leading cause of ocular dysfunction, including irreversible blindness, especially in subjects over the age of 50. All subjects begin with early, non-exudative, so-called "dry" AMD. In the early stages, dry AMD is characterized by deposits known as drusen, or in some cases pseudodrusen, that are clinically visible in the posterior pole of the eye known as the macula. Advanced AMD may take two forms - late "wet" or late "dry". The primary treatment option for advanced "wet" AMD is regular intraocular injections of antiangiogenic drugs. These injections are given after pathological new blood vessels grow beneath or into the central retina where they leak, bleed, or cause tissue damage and scarring. In contrast, there is no treatment for early or late dry AMD. Late dry AMD is characterized by the death or atrophy of photoreceptors and their associated "nurse" cell layer, the retinal pigment epithelium (RPE). Irreversible patches of RPE and loss of photoreceptor tissue are known as "geographic atrophy" (GA). GA may occur centrally in the fovea, depriving subjects of their central vision early in the disease (measured as a decrease in visual acuity), or may accumulate extra or juxta‐foveally, preserving central vision despite significant disease. These non‐foveal scotomas, or blind spots, cause significant disability, particularly with respect to reading speed, tracking, and detection or identification of text, as well as non‐central objects. Thus, visual acuity (e.g., reading an eye chart) is not an appropriate endpoint for clinical trials, nor is it an appropriate measure for clinical trial power‐up.

[0013] Because ocular biopsies and evaluation of systemic (e.g., non-ocular) samples, such as blood or urine, are not routine ophthalmologists and optometrists rely on in vivo imaging for disease diagnosis, classification, and stratification, patient management, and clinical trial design. Fundus autofluorescence (FAF) is the current standard method of imaging dry AMD and relies on the noninvasive detection of endogenous fluorophores, such as lipofuscin pigments, in the RPE monolayer, which provide a diffuse homogenous ground-glass incandescence in normal eyes when stimulated with fluorescent light, for example, in the blue spectrum. At sites of RPE loss (e.g., sites of GA), the autofluorescent FAF signal is also lost. Two-dimensional (2D) areas of GA can be easily quantified due to their widespread hypofluorescent and often sharply bordered black signals. These black signals contrast significantly with the rest of the image, which exhibit variable shades of gray (grayscale) and therefore cannot be reliably measured both during and between patient visits. Newer methods of optical coherence tomography (OCT), when reconstructed to form slabs, may be viewed en face to reveal areas of GA. However, currently, no imaging method other than FAF provides easily quantifiable areas of widespread hypofluorescent, or black, signal to characterize AMD or other ocular diseases or disorders.

[0014] In the absence of other suitable criteria for disease, the quantification of GA and its rate of expansion over time may be used to drive clinical trial design, for example to serve as inclusion criteria or endpoints, respectively. However, FAF detection of GA identifies areas of RPE and photoreceptor loss after the tissue is already missing and therefore cannot be salvaged or restored. It does not provide a measure of tissue at risk for future loss, does not provide an indication of tissue suitable for therapeutic protection, does not identify who may be at imminent risk of developing GA (progressing from early to late dry AMD), and it does not provide insight into disease etiology and disease subtype. Thus, there is a lack of methods to characterize and quantify the characteristics of early and late dry AMD.

[0015] Thus, there is a need for new, or at least alternative, ocular imaging methods for eye diseases or disorders, including AMD. Such methods may be independent of the patient's fixation (which may be lost in central GA), may not rely on prolonged or intensive activity (often weakened due to patient fatigue), and may not require significant cognitive abilities (which may be reduced in the elderly, especially among patients with neurodegenerative diseases). There is also a need for new and alternative imaging techniques and systems for processing, storing, and retrieving ocular images from multiple imaging modalities. Furthermore, there is a movement towards personalized medicine, image-guided medicine, and biomarker discovery to accelerate / automate drug discovery and optimize health care and medical solutions for patients. In this way, scarce resources can be better distributed in the drug development phase and variability during clinical trials can be minimized. To do this, there is an immense need for machine learning, deep learning, artificial intelligence, medical analytics, and cloud computing systems.

[0016] Disclosed herein are methods, systems, and compositions that may be useful for the diagnosis, treatment, or prevention of ocular diseases or disorders, including the use of in vivo imaging methods, image processing methods, and the generation of graphical representations using output from image processing as part of an integrated platform.

[0017] Disclosed herein is a novel functional imaging method, Delayed Near Infrared Analysis (DNIRA), that has produced unprecedented images of the retina (membrane of the eye) containing both known and unknown disease phenotypes. These complex images can be compared against an individual's genetic makeup (their genotype) and their comorbidities, medications, and lifestyle history (their epigenetics). This imaging can be used as a diagnostic tool and to help identify specific subjects who may benefit from specific clinical trials and drug development programs, and ultimately to guide clinical trial design and treatment outcomes. In some aspects, when DNIRA is used to specifically image subjects with AMD, DNIRA is referred to as AMD Imaging (AMDI). In some aspects, when DNIRA is used to image tumor-associated macrophages (TAM) in ocular melanoma patients, DNIRA is referred to as tumor-associated macrophage imaging (TAMI). However, because the methodology has utility for multiple ocular disorders, the methodology can generally be referred to by its broad, generic name, DNIRA.

[0018] Thus, described herein is a method for detecting an ocular disease or disorder or its subsequent blinding complication(s) comprising administering to a subject in need thereof an effective amount of an agent capable of enhancing optical detection of RPE, photoreceptors, or other components of the retina or associated tissues, including cells of the immune system, based on the presence of endogenous fluorophores, luminophores, or chemiluminescent compounds, or based on (but not necessarily limited to) physiological uptake, processing, activation, or accumulation of exogenous fluorophores, luminophores, or chemiluminescent compounds provided locally or systemically and which may be differentially processed or metabolized by those tissues.

[0019] In some aspects, described herein are methods for detecting and imaging the physiological uptake, processing, activation, or accumulation of a systemically administered dye, including administering to a subject a systemic dye, such as indocyanine green (ICG) or a derivative thereof, and using specialized imaging equipment to photograph its presence during the hours and days following administration using appropriate excitation and emission filters or optical systems without further administration of the dye.

[0020] In some aspects, described herein is a system for processing the resulting images using the method to generate a graphical representation of output data. In some aspects, described herein is a system that can provide logical hardware and software units for analyzing and qualitatively quantifying the resulting images using the method. In some aspects, described herein is a method for providing an integrated platform for image acquisition, processing, and output. The platform can store and retrieve imaging data and generate a graphical representation of the imaging data. In some aspects, described herein is a method for providing a dynamic graphical representation of imaging data from the resulting images in a graphical user interface (GUI) using the method described herein. In some aspects, described herein is a system for providing a centralized or networked image processing platform for processing, storage, and retrieval of ocular images from multiple imaging modalities and / or cloud servers.

[0021] In various aspects, the systems described herein may relate to AMDI (AMD Imaging) consisting of the clinical application of delayed near infrared analysis (DNIRA) along with logical processes, hardware and software units, and an integrated platform for its analysis, interpretation, dissemination, application, and integration with other imaging and non-imaging modalities used to evaluate subjects with AMD or RPD, other potentially blinding diseases, and other ocular abnormalities where macrophages are present.

[0022] In various aspects, the systems described herein may relate to TAMI (Tumor-Associated Macrophage Imaging), which consists of the clinical application of DNIRA along with logical processes, hardware and software units, and an integrated platform for its analysis, interpretation, dissemination, application, and integration with other imaging and non-imaging modalities used to evaluate subjects with ocular melanoma, ocular tumors, and other ocular abnormalities where macrophages are present.

[0023] The embodiments described herein further relate to the use of DNIRA / AMDI / TAMI as an endpoint analysis for entrance inclusion or exclusion criteria or clinical trial design. Some embodiments may include the development and application of complementary biomarkers for clinical trial design and drug development. Some embodiments may include the use of DNIRA / AMDI / TAMI as markers for the diagnosis and / or prognosis and / or progression of ocular diseases or disorders (including, without limitation, AMD and RPD, ocular melanoma, diabetic retinopathy, inherited retinal diseases, uveitis, and others).

[0024] Also described herein are methods and compositions that may be useful for the diagnosis, treatment, or prevention of ocular diseases or disorders, including the use of in vivo imaging methods, their outputs, their processes, and management as part of an integrated system for computer-aided analytical diagnosis (CAAD). The system can use as input images of the eye (macula / posterior retina) from various imaging modalities for comparative processing, and provide as output a graphical representation of multiple criteria, including prognostic and diagnostic scores on which to base personalized treatment for ocular diseases. The images may include standard clinical macular images, as well as novel imaging types described herein.

[0025] The system can use data images from multiple modalities and user configurations or parameters as input. The system performs image processing to obtain measurements and other report data for the images. The system has one or more processing components to register the images, segment objects of interest from the images, and define the pattern of the segmented objects and / or the intensity texture of the images. The information can be integrated by the system into feature vectors for each subject. The subject's data can be used as input for statistical models to generate diagnostic and prognostic scores, either categorical or continuous in nature. The system may rely on expert users to provide training or configuration information for image processing, and a feedback loop to review the output from image processing to ensure suitability of the results.

[0026] Also described herein is a system for image processing and display including an image capture device for capturing image data, a data storage for storing and retrieving the image data, a processor configured with a pre-processing unit, a post-processing unit, and a wide-area analytics unit for transforming the image data to generate output image data, the data storage storing the output image data, and a client application for connecting to a display device to generate and update a graphic user interface (GUI) of visual elements representing the image output data. In some aspects, the pre-processing unit can be configured to register and normalize the image data. In some aspects, the pre-processing unit can be configured to register the image data using dynamically adjusted registration parameters using a feedback loop of control commands to verify, approve, and adjust registration parameters. In some aspects, the post-processing unit can be configured to segment the image data and extract features from the segmented image data. In some aspects, the wide-area analytics unit is configured to comparatively process the image data to generate comparative image data as a portion of the output image data.

[0027] In some aspects, the systems described herein may have a delta image unit to cross-reference and compare segments of image data over a period of time and to automatically monitor image data for changes to characteristics of the segments of image data across modalities. In some aspects, a first segment of image data at a first time point may be mapped or masked onto a subsequent segment of image data at a subsequent time point. In some aspects, the system may have a delta image unit for comparing a first image segment to a second image segment to generate difference image data, and a client application for generating visual elements of the different image data to visualize the level of change between the image segments.

[0028] Also described herein is an apparatus for image processing and display, including a data storage for storing and retrieving image data, a pre-processing unit, a post-processing unit, and a wide area analytics unit configured to transform the image data to generate output image data, the data storage storing the output image data, a processor configured in the wide area analytics unit, and a client application for connecting to a display device to generate and update a graphical user interface of visual elements representing the image output data. In some aspects, the pre-processing unit is configured to register and normalize the image data. In some aspects, the pre-processing unit can be configured to use a feedback loop of control commands to verify, approve, and adjust the registration parameters, and to use and register dynamically adjusted registration parameters. In some aspects, the post-processing unit can be configured to segment the image data and extract features from the segmented image data. In some aspects, the wide area analytics unit can be configured to comparatively process the image data to generate comparative image data as a portion of the output image data. In some aspects, the device may have a delta image unit to cross-reference and compare segments of image data over a period of time and to automatically monitor the image data for changes to the characteristics of the segments of image data across modalities. In some aspects, a first segment of image data at a first time point may be mapped or masked onto a subsequent segment of image data at a subsequent time point. In some aspects, the device may have a delta image unit for comparing a first image segment to a second image segment to generate difference image data, and a client application for generating visual elements of the different image data to visualize the level of change between the image segments.

[0029] Also disclosed herein is a process for transforming image data for display, including pre-processing the image data by registering and normalizing the image data, post-processing the image data by transforming the image data to generate output image data showing changes in segments of the image data over time, and generating and updating a graphical user interface of visual elements representing the image output data. In some aspects, the process may include using computer algorithms / analytics to compare the detected features to the genetic risk factors and epigenetics of the subject.

[0030] Details of various examples are described in the accompanying description below. Methods and materials similar or equivalent to those described herein can be used in the practice or testing of various aspects. Exemplary methods and materials are described by way of example. Other features, objects, and advantages will be apparent from the description and from the claims. In the specification and the appended claims, the singular forms include the plural unless the context clearly dictates otherwise. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art.

[0031] In some cases, the methods described herein may include administration of a fluorescent dye, such as ICG, a drug that has long been used to evaluate the ocular (retinal and choroidal) vasculature, which can be detected in phagocytic cells, such as RPE and macrophages, within hours, days, and weeks after systemic vascular administration using a suitable optical imaging system without further injection of the dye. This method differs from autofluorescence, which relies on endogenous fluorophores without the provision of dye, and from angiography, which is used to evaluate vascular anatomy and patency, performed during and within minutes or hours immediately following the passage and recirculation of the dye. Such a method, i.e., DNIRA, can enhance visualization of the eroded RPE, thereby making visible the loss of GA-like RPE and photoreceptor patches, similar to the method of FAF described for subjects with late dry AMD. Thus, some embodiments may include a method for enhancing visualization of the RPE / photoreceptor layers of a subject.

[0032] In some cases, the method may include visualizing, detecting, and quantifying (via a generated graphical representation) the ability of RPE / retinal cells to internalize / incorporate exogenous pigments as a measure of cell viability, thus providing an image-based measure of cell activity, metabolism, and / or dysfunction for the first time. In some embodiments, the method may include visualizing, detecting, and / or quantifying disease burden. In some embodiments, the method may include detecting tissue at imminent risk of loss. In some embodiments, the method may include identifying subjects suitable for clinical investigation or treatment for secondary prevention of GA among subjects with early AMD.

[0033] In some cases, immune cells, particularly cells that can internalize / take up exogenous dyes, fluorophores, chromophores, or chemiluminescent compounds (by way of non-limiting examples, ingestion, passive influx, membrane uptake), can be detected in vivo in the subject's eye. Data can range from ImmunoD and pulsed DNIRA observations. Thus, in some embodiments, the method can include first visualizing phagocytic immune cells in the human eye, including, by way of non-limiting examples, microglia, macrophages, monocytes, and dendritic cells. In some embodiments, the method can include co-localization of immune cells with progressing disease, thus suggesting their role in disease progression. Such diseases can include diseases of innate immune dysfunction.

[0034] Also disclosed herein are logical algorithms that can distinguish DNIRA from other known imaging methods, such as FAF, infrared, or other en face methods at any wavelength of light, or other cross sectional methods, such as optical coherence tomography (OCT).

[0035] definition The terminology used herein is for the purpose of describing specific cases only and is not intended to be limiting. As used herein, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly dictates otherwise. Furthermore, the terms "including," "includes," "having," "has," "with," or variations thereof, wherever used in either the detailed description and / or the claims, such terms are intended to be inclusive in a manner similar to the term "comprising."

[0036] The term "about" or "approximately" may mean within an acceptable error range of a particular value as determined by one of ordinary skill in the art, which may depend in part on how the value is measured or determined, such as the limits of the measurement system. For example, "about" may mean plus or minus 10% per calculation in the art. Alternatively, "about" may mean plus or minus 20%, plus or minus 10%, plus or minus 5%, or plus or minus 1% of a given value. Alternatively, particularly with respect to biological systems or processes, the term may mean within an order of magnitude, within a 5-fold range, or within a 2-fold range of a value. When particular values ​​are recited in the application and claims, the term "about" should be assumed to mean within an acceptable error range of the particular value, unless otherwise indicated. Also, when ranges and / or subranges of values ​​are provided, the ranges and / or subranges may include the endpoints of the ranges and / or subranges.

[0037] As used herein, the terms "subject," "patient," or "individual" may encompass mammals and non-mammals. Mammals may be any member of the mammalian class, including, but not limited to, humans, non-human primates, such as chimpanzees, monkeys, or other monkey species, livestock, such as cattle, horses, sheep, goats, pigs, domestic animals, such as rabbits, dogs (or canines), and cats (or felines), and laboratory animals, including rodents, such as rats, mice, and guinea pigs. Non-mammals may include birds, fish, and the like. In some embodiments, the subject may be a mammal. In some embodiments, the subject may be a human. In some examples, the human may be an adult. In some examples, the human may be a child. In some examples, the human may be between 0 and 17 years of age. In some examples, the human may be between 18 and 130 years of age. In some examples, the subject may be male. In some examples, the subject may be female. In some examples, the subject may be diagnosed with a disease or disorder or may be suspected of having a disease or disorder. In some instances, the disease or condition may be cancer. The subject may be a patient. The subject may be an individual. In some instances, subject, patient, or individual may be used interchangeably.

[0038] The terms "treat," "treating," "treatment," "ameliorate," or "ameliorating," and other grammatical equivalents as used herein, may include alleviating or attenuating the symptoms of a disease or condition, inhibiting a disease or condition, such as preventing the onset of a disease or condition, relieving a disease or condition, causing regression of a disease or condition, alleviating a condition caused by a disease or condition, or arresting the symptoms of a disease or condition. The term "preventing" may include preventing additional symptoms, ameliorating or preventing the underlying metabolic cause of a condition, and may include prophylactic methods.

[0039] The terms "age-related macular degeneration imaging" (AMDI) and "delayed near-infrared analysis" (DNIRA) can be used interchangeably to describe the methods of ocular imaging described herein.

[0040] The terms "connected" or "coupled" may include both direct couplings (wherein the two elements that are coupled to each other are in contact with each other) and indirect couplings (wherein at least one additional element is located between the two elements).

[0041] Figure 1 is a flow diagram of the process involved in obtaining, identifying, processing DNIRA images, and using them in downstream applications. Following image acquisition, cloud-based image analysis is used to generate a complex subject phenotype. This subject phenotype, along with genetic and epigenetic data, may feed downstream steps of biomarker development, clinical trial design improvement, drug development, new treatments, and ultimately selection from a menu of specific treatment options for personalized medicine. (QA=Quality Assurance)

[0042] Imaging System The apparatus, systems, and methods described herein may be implemented in a combination of both hardware and software. Various aspects may be implemented in programmable computers, each computer including at least one processor, a data storage system (including volatile or non-volatile memory or other data storage elements or a combination thereof), and at least one communication interface.

[0043] Although the exemplary embodiments may represent a single combination of inventive elements, all possible combinations of the disclosed elements are contemplated and are within the scope of the disclosure provided herein. For example, if one embodiment includes elements A, B, and C and a second embodiment includes elements B and D, then the scope of the disclosure provided herein may include other remaining combinations of A, B, C, or D, even if not explicitly disclosed.

[0044] Program code may be applied to perform functions described herein and to generate output information. The output information may be applied to one or more output devices. In some examples, the communication interface may be a network communication interface. In some examples, elements may be combined. For example, the communication interface may be a software communication interface, such as a communication interface for inter-process communication. In some examples, there may be a combination of communication interfaces implemented as hardware, software, and combinations thereof.

[0045] Throughout the description, numerous references may be made to servers, services, interfaces, portals, platforms, or other systems formed from computing devices. It should be understood that use of such terms is to be considered to represent one or more computing devices having at least one processor configured to execute software instructions stored on a computer-readable tangible non-transitory medium. For example, a server may include one or more computers operating as a web server, database server, or other type of computer server to perform the roles, responsibilities, or functions described.

[0046] The technical solutions described herein may take the form of a software product. The software product may be stored in a non-volatile or non-transitory storage medium, which may be a compact disc read-only memory (CD-ROM), a USB flash disk, or a removable hard disk. The software product includes some instructions that enable a computer device (personal computer, server, or network device) to execute the methods described herein.

[0047] The methods described herein can be implemented by physical computer hardware, including computing devices, servers, receivers, transmitters, processors, memory, displays, and networks. The methods described herein may include physical machines and computer hardware devices that are specifically configured for the execution of the methods. In some aspects, methods implemented by electronic machines and electronic machines adapted to process and convert electromagnetic signals representing various types of information are contemplated.

[0048] 2 is a schematic diagram of an apparatus for processing, storing, and retrieving ocular images according to some embodiments. The system 100 can be connected to an image acquisition device 106 to receive ocular images of one or more modalities. The system 100 can also be connected to an external image storage unit 102 to access ocular images of one or more modalities. The system 100 can process the images as described herein to generate output data. In particular, the system 100 can generate a graphical representation (of visual elements) of the output data for display on a user device 104. The user device 104 can also provide user configurations and user parameters to the system 100 to dynamically control the display and update the graphical representation based on timing and subject data, as well as to provide feedback and input data.

[0049] The system 100 can be connected to other components in various ways, including direct coupling and indirect coupling through the network 108. The network 108 (or multiple networks) can transmit data. The network 108 may involve wired connections, wireless connections, or a combination thereof. The network 108 may involve different network communication technologies, standards, and protocols, such as, for example, Global System for Mobile Communications (GSM), Code Division Multiple Access (CDMA), Wireless Local Loop, WiMAX, WiFi, Bluetooth, Long Term Evolution (LTE), etc. The network 108 may involve different physical media, such as coaxial cable, optical fiber, transceiver station, etc. Example network types include the Internet, Ethernet, plain old telephone service (POTS) lines, Public Switched Telephone Network (PSTN), Integrated Services Digital Network (ISDN), Digital Subscriber Line (DSL), and others, including any combination thereof. The network 108 may be a local area network or a wide area network.

[0050] FIG. 3 is a schematic diagram of a system 100 for processing, storing, and retrieving ocular images according to some embodiments. The system 100 includes a server 200, a database 202, a client application 204, a user device 206, and an imaging device 208. The system 100 may be useful for diagnosing, treating, or preventing ocular diseases or disorders. The system 100, and in particular the imaging device 208, can use in vivo imaging methods to generate output data. The system 100 provides an integrated platform for computer-aided analysis diagnosis (CAAD). The system 100 can receive images of the eye (macula) from various imaging modalities (captured by the imaging device 208) for comparative processing, and provide as output a graphical representation of multiple criteria, including prognostic and diagnostic scores on which to base personalized treatment for ocular diseases. The images may include standard clinical macular images as well as novel imaging types described herein.

[0051] The system 100 may include a client application 204 for exchanging data between the server 200, the user device 206, and the imaging device 108. The server 200 uses as input data images from multiple modalities captured by the imaging device 108, and user configurations or user parameters received by the user device 206. The server 200 performs image processing to obtain measurements and other reporting data for the images. The server 200 has one or more processing components to register the images, segment objects of interest from the images, and define patterns of segmented objects and / or intensity textures of the images. The information may be integrated by the server 200 into feature vectors for each subject. The subject's data may be used as input for statistical models to generate diagnostic and prognostic scores, either categorical or continuous in nature. The system 100 may interact with expert users to provide training or configuration information for image processing, and to provide a feedback loop to review the output from the image processing to verify suitability of the results.

[0052] The client application 204 may be either web-based or downloaded and can run locally on a designated computing machine (user device 206). The user device 206 displays a graphic user interface (GUI) that is controlled and generated by the client application 204. The GUI may display various graphical representations of the output data and receive control commands to update the image processing. The client application 204 includes a set of processing and analysis tools that provide a platform for image analysis and storage of the ocular images.

[0053] Once imported, the images go through a pre-processing module / stage. Pre-processing may be performed using a suitable computer system via a local client application or a web-based application. The images may be automatically normalized and registered. This is a process performed on an image-by-image basis to create a new set (replica) of successfully normalized and registered images. Given the inherent variability across images acquired over time, if an image is not processed satisfactorily (failure to normalize and / or register), the clinician or diagnostician may be prompted or the image may be tagged for evaluation by a clinical expert at a later date. FIG. 4 is a flow diagram of a method 300 for processing, storage, and retrieval of ocular images according to some embodiments. At 302, the system 100 acquires ocular images with various modalities. The system 100 may use an imaging device 208 or acquire images from a database 202.

[0054] In some embodiments, the system 100 can acquire ocular images in various modalities from an external image storage unit 102 or an image acquisition device 106. During the image acquisition stage, the system 100 can use various ocular imaging equipment to acquire images. Once acquired, the images are stored locally in a database or on a dedicated local server 200 or externally in the image storage unit 102. From there, clinicians and diagnosticians can use the user devices 104, 206 to import the acquired images into the client application 204 for further processing and analysis. Images can be imported in various ways and in different formats. Images can be sequenced in time for a particular subject viewed by a diagnostician (e.g., an ophthalmologist) over a period of time. The clinician or diagnostician can use the user devices 104, 206 to import and process images as they are acquired per patient visit, or import a section of images spanning a period of time for multiple patients. Image metadata can establish the appropriate sequential time frame of images and create tags for future reference. At 304, the system 100 can perform pre-processing of the eye image. Pre-processing may generally involve registration and normalization of the eye image. At 306, the system 100 can perform post-processing of the eye image. Post-processing may generally involve segmentation and feature extraction of the eye image. At 308, the system 100 can perform wide area analytics of the eye image to generate output data. Wide area analytics may generally involve comparison processing and / or report generation. The server 200 generates a graphical representation of the output data for display on the user device 206. The server 200 stores and retrieves image data and output data along with data related to configurations and parameters. The database 202 can store historical and comparative data sets for machine learning and pattern matching by the server 200 in generating the output data. For example, aspects of the workflow may be performed by the server 200, the client application 204, or a combination thereof.

[0055] FIG. 5 is a schematic diagram of a system 100 for processing, storage, and retrieval of eye images according to some embodiments. The system 100 may include a retinal image capture unit 402, a display device 450, and a server 200. The server 200 may include various hardware and software units, including a pre-processing unit 404, a post-processing unit 408, and a global analytics unit 408. The server 200 can store the eye images received from the retinal image capture unit 402. The server 200 can generate a graphical representation of the output data for display as part of a GUI on the display device 450. The server 200 may receive commands, configurations, and parameters from the display device 450. The server 200 can display the processed images on the display device 450. The pre-processing unit 404 generally involves registration 410 and normalization 414 of the stored eye images 412.

[0056] The registration 410 can transform different sets of image data, for example, to a target image space, domain, or coordinate system. The image data can be from multiple modalities, different imaging devices, times, depths, or viewpoints. The registration 410 can provide the ability to compare or integrate image data obtained from different modalities, formats, and measurements. The registration 410 may receive registration parameters 422 to use a transformation model to transform different sets of image data from the stored eye images 412 from a reference image space to a target image space. The transformation model may receive the registration parameters 422 as input. Feedback may be received to indicate whether the registration parameters 422 require improvement 420. The pre-processing unit 404 adjusts the parameters 416 if control commands or data processing indicate that they require improvement 420. The registration 410 receives a validation or approval 418 before updating the stored eye images 412 with the registered image data.

[0057] Normalization 414 may be a process that can change the range of pixel intensity values ​​of the stored eye images 412. Once registered and normalized, a post-processing unit 460 further processes the stored retinal images 412 to generate unique information regarding the stages and progression of various eye diseases. Additional processing features can be performed standalone or sequentially.

[0058] The segmentation unit 404 detects and segments regions of the retina image 412. Eye images inherently have several characteristic features, e.g., vasculature, that are unlikely to change over time. Eye images can be used to generate visualizations of physiological changes resulting from biological significance. By segmenting regions and features of interest on eye images, the segmented regions may be automatically monitored over time for changes, e.g., shape or size. For example, for a single patient where four images are acquired 1-2 months apart, the segmentation unit 404 can segment image regions and features of interest in the image acquired at the first visit to serve as a baseline for feature size and shape. When a series of images are comparatively analyzed, the original segmentation can be mapped and / or masked on subsequent images. Additionally, the segmented regions in the first image can be used to quantify size changes in subsequent images.

[0059] Feature selection 430 can be used to select and identify features within and segments of the image 412. Feature extraction 434 extracts features that are intended to be informative and non-redundant, and may facilitate subsequent learning and generalization steps. Feature extraction 430 may involve dimensionality reduction. Feature selection 430 may facilitate converting a large amount of image data into a reduced set of features (e.g., feature vectors). The selected features contain relevant information from the input data, so that the comparison process can be performed by using this reduced representation instead of the complete initial data.

[0060] Wide area image analytics 408 can be used to derive output values ​​based on comparative processing of images. As images 412 are processed for subjects at multiple visits, quantitative information 428 can be generated and stored throughout the process. If a clinician or diagnostician is interested in compiling and analyzing key characteristic changes for multiple subjects over time, the data can be selected, pooled, and reprocessed to produce aggregated quantitative results 428. For example, if a clinician is interested in analyzing changes in geographic atrophy for a subject pool undergoing a common therapeutic treatment for a particular disease, they can aggregate subject results across subject visits to produce quantitative measurements and information for the subject pool.

[0061] The quantitative metric 428 may generate different output data values. An example quantitative metric 428 is delta. A feature of the system 100 may be the ability to analyze images and cross-reference and compare characteristic features across different modalities over time. Feature selection 430 and feature extraction 430 are tools for isolating features from segments of the image 412.

[0062] The feature difference comparison analysis 446 can be performed between two images taken consecutively in time within the same modality, but it can also be used to observe and measure differences on images separated at multiple intervening imaging time points (e.g., time point 1 vs. time point 4), and it can be extended to compare images taken from different modalities, such as fundus autofluorescence (FAF), infrared (IR), or DNIRA. The reporting unit 448 can generate a graphical representation of the output data for display on the display device 450 as part of the GUI. The data selection 436 selects and isolates data sets for comparison or use by the report generation 448.

[0063] FIG. 6 illustrates various units and applications of the apparatus and methods described herein. An exemplary apparatus configuration consistent with FIG. 5 is illustrated in FIG. 6 and FIG. 7. FIG. 6 is a schematic diagram of a system for processing, storage, and retrieval of ocular images according to some embodiments. As illustrated, in some embodiments, the server 200 may or may not include aspects of wide area analytics 408. FIG. 6 is a schematic diagram of a system for processing, storage, and retrieval of ocular images according to some embodiments. A raw image input 602 may be provided to an image pre-processing unit 404 to generate qualified image data 608. An image post-processing unit 406 further transforms and processes the qualified image data 608 using segmentation 424 and feature extraction 434.

[0064] The delta unit 614 may generate a delta image as output data. A feature of the system 100 may be the ability to process images to cross-reference and compare characteristic features of the images across different modalities over time. Feature selection 430 and feature extraction 430 may be tools to isolate features from segments of the image 412.

[0065] Comparative analysis of feature differences can be performed between two images taken consecutively in time within the same modality. Comparative analysis can also be used to observe and measure differences on images separated by multiple intervening imaging time points (e.g., time point 1 vs. time point 4), and can be extended to compare images taken from different modalities, such as fundus autofluorescence (FAF), OCT, fundus photography, red-free, infrared (IR), angiography, or DNIRA.

[0066] The image discrimination, or delta process 614 utilizes statistical and algorithmic processes to compare two or more pre-processed images to each other and determine or visualize the level of change between them, and the differences can be quantified on a pixel-by-pixel or pixel site level. Typically, two images 412 are compared to each other by subtracting one from the other to generate a difference image. This difference image can be an example output data that can be used to generate a visual representation for the GUI and display device 450.

[0067] The delta image generation unit 614 is configured to generate a series of delta image outputs linked to a comparative time reference, either image by image or with reference to a common baseline image. The selection of images for this delta generation process 614 may be automated or may involve manual input. A clinician or diagnostician can configure the analysis to compare a single image or a series of images to a baseline reference and assemble the difference data to highlight key changes over time.

[0068] Output delta images (quantitative output 620 and image output 618) may be created and stored (locally or externally). The delta image output can be viewed through an integrated viewer as part of the GUI, processed, and displayed on the display device 450. When visualized in an integrated format, the difference images may be overlaid with different colors to allow the user to observe how changes have appeared over time. Delta image analysis can also present an overlay format where each image is superimposed on top of the previous image in a visual representation. By allowing the display device to display the image differences relative to the ocular structures, the user is given the opportunity to see how the differences compare to the ocular structures, making disease diagnosis easier.

[0069] FIG. 7 is a schematic diagram of a computing device for implementing aspects of a system 100 (including a server 200 or user devices 104, 206) for processing, storage, and retrieval of eye images according to some embodiments. For simplicity, only one computing device is shown, but the system may include more computing devices operable by a user to access remote network resources and exchange data. The computing devices may be the same or different types of devices. The computing device may include at least one processor, a data storage device (including volatile or non-volatile memory or other data storage elements or a combination thereof), and at least one communication interface. The components of the computing device may be connected to other components in various ways, including directly coupled and indirectly coupled via a network, distributed over a wide geographical area, and connected via a network (sometimes referred to as "cloud computing").

[0070] In some aspects, the computing device may be a server, a network appliance, a set-top box, an embedded device, a computer expansion module, a personal computer, a laptop, a personal data assistant, a mobile phone, a smart phone device, an imaging device, a display terminal, and a wireless hypermedia device, or any other computing device capable of being configured to perform the methods described herein. A computing device (e.g., system 100, server 200, or user device 104, 206) may include at least one processor 702, a memory 704, at least one I / O interface 706, and at least one network interface 708.

[0071] Each processor 702 may be, for example, any type of general-purpose microprocessor or microcontroller, digital signal processing (DSP) processor, integrated circuit, field programmable gate array (FPGA), reconfigurable processor, or any combination thereof. Memory 704 may include any suitable combination of any type of computer memory located either internally or externally, such as, for example, random access memory (RAM), read only memory (ROM), compact disc read only memory (CD-ROM), electro-optical memory, magneto-optical memory, erasable programmable ROM (EPROM), electrically erasable programmable read only memory (EEPROM), ferroelectric RAM (FRAM®), etc.

[0072] Each I / O interface 706 allows a computing device to interconnect with one or more input devices, such as a keyboard, mouse, camera, touch screen, and microphone, or one or more output devices, such as a display screen and speakers. Each network interface 708 allows a computing device to communicate with other components, exchange data with other components, access and connect to network resources, serve applications, and run other computing applications by connecting to a network (or networks) over which data can be transmitted. A computing device may operate to register and authenticate users (e.g., using a login, unique identifier, and password) before providing access to applications, local networks, network resources, other networks, and network security devices. A computing device may serve one user or multiple users.

[0073] 8 is an example flow diagram of a method for processing, storage, and retrieval of eye images. The process uses as input a first image 802 and a second image 804 (and additional images) to compare features of the images over time. As output, the process generates visual elements or representations for a GUI to show changes to the features over time.

[0074] In some aspects, the process can provide comparative analysis of images within and between various imaging modalities. The method can provide two or more images 802, 804 with each other by establishing point correspondences between the images or target image space for registration 410. Registration 410 is performed to visualize and address characteristic differences between images taken at a particular time or over a period of time within an imaging modality, such as individual DNIRA images, or between images of different modalities, such as DNIRA with FAF, OCT, fundus photography, red-free, infrared (IR), angiography, and OCTA images. An example of registration is shown in FIG. 8. The registration 410 process maps each point in one image onto the respective corresponding point in the other image. In certain aspects, registration 410 can be achieved by using correlation and / or feature location matching and determining registration parameters 422. The images can then be registered with each other via the registration function 410. Alternatively, a mechanical deformation model can be used for registration 410. Different registration methods may be utilized to register images with each other before comparing the images for differences or changes. This includes fully automatic registration as well as computer-assisted manual registration, or different registration approaches using various degrees of manual intervention. To register two or more images, the images may be pairwise registered. For example, if a first image and a second image are registered, and separately a second image and a third image are registered, then by configuration of the registration function, the first image and the third image are effectively registered. This concept can be extended to any number of images. Thus, using the pairwise registration 410 process, one can register any number of images. Additionally, the registration 410 process of simultaneously registering two or more images may be utilized by current technology.

[0075] In some examples, registration 410 may be based on landmark extraction or may be feature based. Registration 410 of the images may be achieved by modeling the large-scale motions and local distortions of anatomical structures. In some examples, registration parameters 422 of the model that define these motions are estimated and selected. A search is then performed to find the parameter values ​​that result in the best registration of images within and between modalities over time. The quality of this registration may be based on a comparison of corresponding points in the original images.

[0076] In some aspects described herein, pixel intensity can vary significantly between images, both within and between imaging modalities, potentially obscuring valuable features of the image. Image normalization 414 can correct for the variable intensity between images. Registration 410 can transform an image into a target image space for comparison against other images in a common target image space.

[0077] Dealing with low fluorescence signals Disclosed herein is an en face functional imaging system and method for the assessment of ocular diseases or disorders, such as non-exudative (dry) age-related macular degeneration (AMD), that can identify both tissue loss (structural damage) and "disease" or abnormal tissue (functional damage). The method may be referred to herein as AMD imaging (AMDI) or delayed near infrared analysis (DNIRA). By way of illustration, the DNIRA method can be compared to fundus autofluorescence (FAF). Like FAF, DNIRA uses structured computer learning based on the detection of widespread areas of hypofluorescence (black) in an otherwise grayscale image, and can generate images that are easily quantifiable. FAF may only be able to quantify advanced late dry AMD with geographic atrophy (GA) only after irreversible loss of retinal pigment epithelium (RPE) and photoreceptors.

[0078] The imaging system can be used to register, segment, and provide inter-modal comparisons between DNIRA and any other imaging modality capable of producing easily quantifiable images. For example, the imaging system can register, segment, and provide intra-modal comparisons of DNIRA over time against any other imaging modality over time. In some aspects, an imaging system for registering and providing inter-modal and intra-modal comparisons of DNIRA and FAF against other en face or cross-sectional imaging modalities capable of producing images of the eye. In some aspects, an imaging system for registering and providing inter-modal and intra-modal comparisons of DNIRA and FAF against any functional modality capable of producing information about the state / health of the fundus. By way of non-limiting example, such methods could include microperimetry, low light visual acuity, and dark adaptation measurements.

[0079] The imaging systems described herein can register and provide inter- and intra-modal comparisons of DNIRA, FAF, and other imaging modalities to other currently used non-imaging modalities, such as, by way of example and not by way of limitation, genetic analysis of single nucleotide polymorphisms in genes such as CFH, C2, C3, C5, C6, C7, C8, C9, CFI, CFD, MCP-1, MCP-2, MCP-3, ARMS2 / HTRA1, TIMP3, SLC16A8, COL4A3, COL8A1, CETP, LIPC, APOE, VEGFA, VEGFC, ABCA1, ABCA4, MMP9, VTN, NFkappaB, etc. Alternatively, images may be compared to genetic analysis of copy number variations of any of the above genes, such as, without limitation.

[0080] Additionally, the imaging systems described herein can register and provide inter- and intra-modal comparisons of DNIRA, FAF, and other modalities for other factors, such as co-morbidity analysis, including, but not limited to, cardiovascular diseases such as coronary artery disease, atherosclerosis, and stroke, as well as chronic kidney disease; neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, multiple sclerosis, Huntington's disease, amyotrophic lateral sclerosis, and neurological trauma; inflammatory diseases such as rheumatoid arthritis, osteoarthritis, ankylosing spondylitis, systemic lupus erythematosus, celiac disease, Crean's disease, Behcet's disease; and pulmonary diseases such as chronic obstructive pulmonary disease (COPD), including asthma, emphysema, and allergies. In some embodiments, the imaging system can register and provide inter-modal and intra-modal comparisons of DNIRA and FAF against ocular and systemic biomarkers, including but not limited to monocyte chemotactic factors (MCPs), e.g., MCP-1, MCP-2, MCP-3; interleukins (ILs), e.g., IL-6, IL-8; vascular endothelial growth factors (VEGFs), e.g., VEGF-A; tumor necrosis factors (TNFs), e.g., TNFa; nitric oxide synthases (NOSs), e.g., NOS-2; complement components, e.g., C1-C9; complement regulators, e.g., CFD, CFH, CFI; apolipoproteins (Apo), e.g., ApoE, ApoB, C-reactive protein, and eotaxins, e.g., eotaxin-1. In some exemplary aspects, the imaging system can register and provide inter- and intra-modal comparisons of DNIRA, FAF, and other imaging modalities against environmental / epigenetic risk factors such as smoking, obesity, blood pressure, body mass index, waist-to-hip ratio, cholesterol levels, lipid panels, family history, age, and dietary vitamin intake, and medications.

[0081] The embodiments described herein can provide two iteratively refined data analysis processes for the study. Beta version 1 may be configured for non-cognitive image analysis with novel biology-based iterative refinement including clinician input, and also to include multi-modal image comparison. Beta version 2 may be configured for cognitive and non-cognitive image analysis with iterative refinement for large-scale data analytics, cloud-based analysis, for multi-modal image-based comparison, and for intra-modal comparison with other subject-centric databases (e.g., genetics, comorbidities, medications, diet).

[0082] The aspects described herein can provide an imaging system to perform the following: image output from an acquisition device, native image registration, image region segmentation, feature extraction specific disease related observations (compared to other modalities and over time), inter-modal image analysis, and generation of delta image data (e.g., DNIRA subtracting another imaging modality such as FAF), and segmentation over time.

[0083] When the system 100 generates image output from an acquisition device, there may be challenges with "best" subject images, slow conversion to non-confidential image output formats, long directory names, and the need to be suitable for downstream batch processing. The purpose for generating image output from an acquisition device is for selection of appropriate images for downstream analysis that is device independent. The system 100 may be suitable for use by a physician, for example, with an image-based interactive GUI interface that is rendered and updated, for example, on a user device 104.

[0084] When the system 100 performs native image registration, there may be challenges due to the spherical shape of the human eye that causes tilt, skew, subtle magnification changes between images, and potentially human features that result in significant differences over time (e.g., cataracts). The goal of native image registration is cross-modal and time image comparison. This allows for detection of image-based features that occur in one or more modalities and that increase, decrease, or remain unchanged in size (area) over time. Computational requirements are identification of branching points on retinal vessels, visualization of the optic nerve head (ONH). Optimization is that images are processed within concentric rings / zones that include a) an outer ring that is discarded due to distortion, b) a central ring that is not used due to lack of medium for the region of interest and larger vessels. Optimization may be that images are iteratively reviewed and a physician or expert provides feedback on acceptable and unacceptable differences as training input for the machine learning process. Optimization may be for batch processing with a threshold such as >85% of all images (e.g., to reject <15% for manual registration).

[0085] The optimization may be a three image registration using vessel markings (images taken at three separate time points). In various embodiments, more or less images may be used. FIGS. 17A and 17B show the image output data as a visual representation of three images at three separate time points. The output of native image registration may be a stack of registered images for downstream manipulation and processing. The system 100 may be suitable for physician input for iterative refinement on parameters. The system 100 may be suitable for export-ready cloud-based processing for large scale data analysis platforms.

[0086] When the system 100 performs image segmentation, there may be a challenge after registration to distinguish the black regions from the grayscale background. After registration, the black regions need to exclude certain features, such as blood vessels and ONH. The purpose of segmentation is to identify and quantify high contrast features identified as widespread hypofluorescence (black). For example, using FAF techniques, the black regions correspond to sites of irreversible tissue loss. FIG. 17A shows progressive irreversible loss of tissue visualized by FAF imaging. Over time, the black regions increase in size or remain the same (e.g., regression or shrinkage of the black is observed).

[0087] DNIRA provides a quantifiable functional readout of tissue function. Using DNIRA, system 100 can detect dark areas in the image that correspond to sites of irreversible tissue loss and to abnormal or diseased tissue. Over time, the dark areas may remain the same, increase in size, or decrease in size. System 100 may perform the determination of a "baseline DNIRA" image. System 100 can obtain a baseline DNIRA image by removing the signal from the pre-dye NIR image from the post-dye NIR image.

[0088] When the system 100 performs feature extraction, there may be challenges related to identifying and quantifying disease features other than tissue loss. This process currently relies on manual interpretation of multimodal images, including four en face and one cross sectional method. The system 100 can use the techniques described herein to automate aspects of this process. The purpose of feature extraction is to identify disease-related features that aid in subject diagnosis, staging, severity classification, and ultimately aid in clinical trial design and monitoring response to treatment. The purpose of feature extraction is to facilitate the quantification and classification of drusen, RPE detachment (RPED), and other disease features using image processing. The purpose of feature extraction is to enable the observation and quantification of dynamic changes in disease using graphical representations and visual elements of a GUI displayed on the user device 104. For example, identification of the following en face features: a) small solid drusen, b) medium drusen, c) large drusen, d) confluent drusen, e) RPE detachment (e.g., drusenoid, serous), f) areas of geographic atrophy (GA), g) blood informs the physician of disease severity (based on clinical examination and imaging methods) and risk to the subject. System 100 for performing DNIRA uses structured computer processes, providing the possibility to identify and quantify clinically relevant features (beyond simply GA). It may detect the following in the image data: a) small black dots with a circularity of 0.8 or greater, <65 μm diameter; b) small black dots with a circularity of 0.8 or greater, 65-125 μm diameter; c) large black dots with a circularity of 0.8 or greater, >125 μm diameter; d) rectangular, oval, or round structures deemed to result from a collection of drusen [fingering potatoes] with a circularity >0.2, >125 μm in one dimension; e) large black areas >200 μm diameter, circularity >0.5; f) black areas seen in both DNIRA and FAF; g) blood.

[0089] The system 100 performs inter-modal image analysis and generates delta image data (DNIRA minus FAF). The delta image defines areas of black DNIRA that coincide with black GA (and thus represent GA) for black DNIRA that exceeds black FAF (and thus represent potential damaged or abnormal tissue present outside of GA) in any imaging session. Suitable methods for en face segmentation include DNIRA and FAF. The system 100 performs inter-modal image analysis and generates delta image data for identification of disease-related features in registered and segmented images, both between modalities (FAF and DNIRA only) and over time. The system 100 performs inter-modal image analysis and generates delta image data for identification of en face features to inform physicians of disease severity and subject risk based on FAF, including sites of geographic atrophy (GA). DNIRA uses a structured computational process and offers the possibility to identify and quantify GA. For example, using DNIRA, areas that appear black and are equal or larger in area than black areas also observed in FAF at the same spatial location, >200 μm, or alternatively measured relative to the diameter of the ONH. The output can be delta image data that is the "DNIRA minus FAF image" to identify areas of corresponding black signal.

[0090] Identification of en face features may inform the physician of abnormal RPE / PhR physiology in areas outside the GA (e.g., areas of diseased tissue rather than dead tissue), including areas of abnormal cellular physiology not currently known to the physician and that do not involve the GA. Using DNIRA, discontinuous areas >200 μm (or measured relative to the ONH) that appear black may not be evident with FAF, providing improvement. Delta image data includes "DNIRA minus FAF" image data to identify areas of black signal that extend beyond the boundaries of the GA or reside in different spatial regions.

[0091] En face grayscale methods include infrared (IR) reflectance, red free (RF) color, color (RGB), color (CNYK), [useful for wet AMD, but may be less useful for dry AMD] angiography, en face optical coherence tomography (OCT), and OCT angiography (OCTA). The goal of cross-modal image analysis is to identify disease-related features of registered images in grayscale images, both across modalities and over time. Cross-modal image analysis may perform identification of the following en face features: a) small hard drusen, b) medium drusen, c) large drusen, d) confluent drusen, e) RPE detachment (e.g., drusenoid, serous, hemorrhagic), f) blood, g) geographic atrophy (GA), informing physicians of disease severity and patient risk at the time of exam and for currently available en face imaging methods (except FAF). Inter-modal image analysis can use unstructured computational processes (eg, cognitive computing) to generate output data and transform image data.

[0092] Cross-sectional methods include optical coherence tomography (OCT) and OCT angiography (OCTA). Cross-modal image analysis may be performed to identify the following cross-sectional features detected by OCT: a) presence and number of drusen in the RPE and / or sub-RPE space, b) drusenoid RPE detachment, c) pseudodrusen in the subretinal space, d) geographic atrophy (loss of RPE / PhR with greater OCT signal in deeper tissues), e) drusen volume, f) areas of drusen regression, g) outer nuclear layer (ONL) thickness and volume, h) hyperreflective OCT foci, i) presence of nascent GA, j) presence of stationary CNV, k) subfoveal choroidal thickness, l) disruptive pigment epithelial detachment, informing the physician of disease severity and subject risk. Enface DNIRA or other imaging modalities may be compared against OCT using single slices or 3-dimensional reconstructions that generate graphical representations of cross-sectional imaging modalities.

[0093] The system 100 may perform segmentation over time, where the different features selected and extracted are followed over time (e.g., to determine whether they are increasing, staying the same, or decreasing in area or number). The system 100 may perform segmentation and, when followed over time, the FAF may allow the physician to quantify the sites of GA and measure their rate of expansion over time. This step may allow identification of subjects with the highest rate of GA expansion, which may be useful for clinical trial enrollment and endpoint analysis. However, it may not follow the changes prior to irreversible tissue loss. Furthermore, quantification of GA suggests that the disease is unidirectional (e.g., features accumulate but do not regress), which may be false. Three main parameters may be assessed: A. change in overall black DNIRA (over GA), B. change in overall drusen, C. change in RPE detachment. The purpose of segmentation over time may be to quantify clinically relevant features of the disease (e.g., drusen, RPED, etc.) and determine their changes over time, including area of ​​GA, total area above GA (as a potential measure of disease burden), and areas of sub-segmented features of the disease (e.g., small, medium, large drusen, RPED, etc.). The data confirms that areas of black DNIRA not only increase, but may decrease over time, confirming a dynamic feature of the disease most typically observed in color photographs (e.g., drusen regression). This may be drusen regression. System 100 performs segmentation over time to generate delta time (DNIRA), which can be measured as a decrease in black DNIRA signal over time at small, medium, or large. System 100 performs segmentation over time to generate the following outputs: a) change in overall black DNIRA over time; b) change in "delta DNIRA minus FAF" over time; and c) change in specifically segmented features, such as change in small drusen, change in medium drusen, and change in large drusen.

[0094] Changes in disease burden (outside of GA) cannot be detected using known techniques. This parameter will provide a measure of disease burden and aid in clinical trial enrollment to identify globally diseased or dysfunctional tissue. The output of "Delta Image = DNIRA minus FAF" may show areas of increasing size (area) over time.

[0095] System 100 performs segmentation over time to quantify the extent of diseased or dysfunctional tissue. The output "Delta Image = DNIRA minus FAF" shows areas that remain unchanged in size (area) over time. The output "Delta Image = DNIRA minus FAF" shows areas that decrease in size (area) over time.

[0096] Changes in drusen accumulation over time can be difficult to identify and even more difficult to quantify. Drusen are not uniformly detected by FAF or other en face methods. DNIRA may be used to detect and quantify dynamic changes in drusen accumulation and regression. It is widely supported that drusen accumulation (number, area, or 3D volume) is proportional to disease. The output "Delta Image = DNIRA minus FAF" shows areas that increase in size (area) over time. The output "Delta Image = DNIRA minus FAF" shows areas that remain unchanged in size (area) over time.

[0097] Drusen regression (often seen on color photographs or by number, area, or 3-dimensional volume) may be a sign of progressive disease. Drusen loss may herald late stage MAD with GA or CNVM. The output "Delta image = DNIRA minus FAF" shows areas of decrease in size (area) over time.

[0098] Drusen may be detected using OCT with 3D slab analysis. It is believed that accumulation of drusen (number, area, or 3D volume) may be proportional to disease. The output "Delta Image = DNIRA minus FAF" shows areas of increasing size (area) over time, which can be used to compare to OCT.

[0099] Changes in RPE detachment (RPED) size (area, height, volume) over time can be difficult to identify and difficult to quantify. System 100 can detect and quantify dynamic changes in RPED size (area) and number. It is believed that RPED resolution [especially drusenoid diversity] is indicative of risk of progression to late blinding GA. Output "Delta Image = DNIRA minus FAF" shows areas of RPEDs that increase in size (area) over time. Output "Delta Image = DNIRA minus FAF" shows areas that remain unchanged in size (area) over time. Output "Delta Image = DNIRA minus FAF" shows areas that decrease in size (area) over time.

[0100] While accumulation of drusen reflects ongoing disease, areas of drusen regression (often seen in color photographs or by number, area, or three-dimensional volume) are indicative of progressive disease. Loss of drusen may herald late-stage MAD with GA or CNVM. In some embodiments, AMDI can detect regression of drusen, allowing for the determination of AMD progression.

[0101] In some embodiments, AMDI images can be compared to other imaging modalities, such as OCT-angiography (OCT-A), allowing for the identification of sites of pathological blood flow (reduced flow through either the choriocapillaris or other layers of the choroid) or potentially increased flow into areas of leaky or non-leaky neovessels (choroidal neovasculature) or vascular lesions (choroidal plexus, polyps, aneurysms, anastomoses, etc.).

[0102] Example output data 11A-11D are exemplary illustrations of visual representations that can be visualized using the systems described herein. FIG. 11A shows image output data as a visual representation of small and medium black dots (<65 μm and 65-125 μm). FIG. 11B shows image output data as a visual representation of large drusen (>125 μm). FIG. 11C shows image output data as a visual representation of confluent drusen. FIG. 11D shows image output data as a visual representation of GA.

[0103] In some cases, the output data may be shown as multiple black dots as depicted in Figures 12 and 13. Unlike FAF, the DNIRA signal is black and the RPE is separated from the BM and choroid by structures such as drusen (arrows overlaid on the DNIRA image indicate representative drusen). As shown in Figure 13, soft confluent drusen demonstrate a rectangular foot shape with widespread hypofluorescence (black) using DNIRA. Soft drusen are usually defined by color fundus photography. Soft drusen are grayscale and therefore cannot be quantified at other wavelengths.

[0104] In some cases, the output image data obtained by DNIRA may be compared output image data obtained by other methods to reveal features of image pairs that may not necessarily be evident when analyzing the images in isolation. For example, FIG. 12 shows comparative retinal images obtained by IR imaging, FAF imaging, DNIRA imaging, and OCT imaging. The arrows indicate the location of drusen visible in the DNIRA image. These drusen are not readily visible in the corresponding IR or FAF images. FIG. 14 also shows the difference between images obtained using DNIRA along with images obtained by FAF. FIG. 14 shows a representative DNIRA image of a patient with central GA observed with FAF (left). As shown in FIG. 14, the image output data from FAF imaging (left) shows less black (GA), whereas the DNIRA image (right) shows more black. The DNIRA image (right) shows a greater number of hypofluorescent signals, a greater total area of ​​hypofluorescence, and a greater amount of peripheral / border of hypofluorescence. The difference between these two images represents the "delta" portion of the signal. The delta image data may be "DNIRA black area - (minus) FAF black area", which is the area of ​​diseased or dysfunctional tissue in the absence of GA.

[0105] FIG. 15A shows the image output data as a visual representation, the top image shows a large RPE detachment that appears black using DNIRA, and the bottom image shows the elevation of the RPE layer (high dark white line). The blue horizontal line indicates the height at which the OCT cross section was taken. FIG. 15B shows the image output data as a visual representation, with the top of the DNIRA showing the black area and the bottom of the OCT showing a small drusenoid RPE detachment (resolution >200 μm). The right image has the top of the DNIRA showing some remaining black signal, and the bottom of the OCT showing the resolution of the RPE detachment. Note also that the right image shows bright dots associated with the RPED suggesting the presence of macrophages.

[0106] In some embodiments, analysis of image data from multiple imaging modalities can be used to visualize RPE features. Figure 16 shows image output data as a visual representation with a left image for IR, a center image for FAF, and a right image for DNIRA. Areas of hypofluorescence (black) indicate loss of RPE.

[0107] In some examples, images can be collected as a function of time to elucidate changes in disease state or condition. For example, FIG. 17A shows image output data as a visual representation of a left image at time t=1, a center image at t=2, and a right image at time t=3 (all 4 months apart). The black areas are larger in the first image than in the second image, indicating reduced features. FIG. 18 and FIGS. 19A-19J show that a collection of TAMi composite images can enable correlation of the disclosed imaging method with current clinical modalities. Fundus autofluorescence (FAF, FIG. 19B) and TAMi composite images (FIG. 19C) were collected as overlays on a color photograph (FIG. 19A). Multimodal analysis allowed identification of regions of interest (FIG. 19F) that included the tumor site (i), areas of current or former fluid (ii), and the periphery (extra-lesional site (iii). mm 2 Site areas at 100 nm were determined using ImageJ measurements and used in the hyperfluorescent dot density calculation (Figure 19H). These were compared across patients with melanoma, indeterminate lesions, or benign nevi as determined by clinician ground truth. Mean dot density is shown as number of dots / mm2 ± SEM. One-way ANOVA and post hoc multiple comparisons (Figure 19I) show significantly higher dot density in the melanoma arm when considering lesion and fluid regions, but not within the periphery (Figure 19J). Multiple comparisons of dot density at each site within the risk groups showed that melanomas had significantly higher dot density in lesion and fluid regions when compared to the periphery. This was not observed in the other risk groups.

[0108] Low Fluorescence NIRA Signal In some aspects, the method may include identifying discontinuous regions of the eye, such as, for example, in the macula, where pigments, fluorophores, chromophores, or chemiluminescent molecules that would otherwise be internalized, absorbed, processed, or accumulated cannot (or may be, albeit to a relatively small extent), thus preventing or reducing their detection. This may occur, for example, within a particular layer or tissue, such as, by way of non-limiting example, discontinuous regions of the RPE monolayer. Such regions, made visible by their low fluorescence / black signal, may be identified and quantified by DNIRA / AMDI, both within a single image and within their rate of change calculated across multiple images over time.

[0109] Similar observations have been made using FAF, for example, where areas of absent signal spatially coincide with sites of RPE / photoreceptor loss such as occurs in geographic atrophy. FAF detects endogenous fluorophores that are produced by photoreceptors and accumulate in the RPE, so when the RPE is lost, the FAF signal is also lost. Because areas of GA cannot regress, e.g., become smaller in size, the widespread hypofluorescent / black areas of FAF will either remain the same size or grow larger over time.

[0110] The aspects described herein relate, in part, to temporal changes in DNIRA images that manifest as an expansion over time. Without wishing to be bound by any theory, such an increase in areas of hypofluorescent / black DNIRA could represent an increase in disease burden. As such, these sites could represent areas where abnormal or damaged RPE / photoreceptors are unable to internalize pigment. Again, without wishing to be bound by any theory, such an increase could represent (i) sites where changes in tissue anatomy prevent or reduce pigment uptake in the presence of otherwise normal RPE / photoreceptor cells, or (ii) sites where anatomy is normal but cellular physiology is abnormal.

[0111] In some embodiments, the temporal changes in DNIRA can be utilized to demonstrate a reduction in widespread hypofluorescent / black areas over time. Without wishing to be bound by any theory, a reduction in areas of hypofluorescent / black DNIRA may represent a reduction, amelioration, or alleviation of disease burden. In this manner, these sites may represent areas where abnormal or damaged RPE / photoreceptors are better able to internalize dye. Again, without wishing to be bound by any theory, such reduction in hypofluorescent / black signal may also represent areas where changes in tissue anatomy have been ameliorated, thereby facilitating dye uptake in the presence of otherwise normal RPE / photoreceptor cells. By way of example, such reduction may occur with resolution of RPE detachment (RPED), which allows the RPE layer to reappoint itself with the underlying Bruch's membrane, thus facilitating uptake of dye, such as ICG, from the underlying choroidal vasculature. Such a reduction in the size of hyperfluorescent / black sites has not been reported using FAF or other imaging modalities. Thus, these features demonstrate the dynamic qualities of DNIRA, further distinguishing it from FAF and other static imaging modalities.

[0112] The embodiments described herein may relate to temporal changes in DNIRA / AMDI that manifest residual widespread hypofluorescent / black areas that remain unchanged over time. When coincident with widespread hypofluorescent / black areas in FAF, they are hypothesized to reflect areas of RPE / photoreceptor loss after RPE / photoreceptors die and are gone. In contrast, other areas of widespread hypofluorescent / black areas in DNIRA may reflect anatomical regions that, without wishing to be bound by any theory, prevent dye uptake from the underlying choroidal vasculature, such as ICG, and that remain largely unchanged. For example, in addition to RPE detachment that separates RPE / photoreceptors from the underlying choroidal vasculature, the interposition of small deposits known as drusen, basement membrane deposits, or basal linear deposits may also prevent dye uptake (FIG. 10). Thus, the presence of widespread hypofluorescent / black dots that are measured to be consistent in size and location with drusen observed with other imaging modalities (e.g., IR, red-free, color fundus photography, OCT, angiography, OCTA) would provide a novel method for measuring total drusen area. Without wishing to be bound by any theory, drusen volume is considered a global measure of disease burden, but its quantification is difficult using imaging modalities such as IR or FAF due to its grayscale nature and the need for three-dimensional (3D) volumetric reconstruction with other methods such as OCT. In this embodiment, the ability to quantify small widespread hypofluorescent / black dots in a given size range, e.g., larger than droplets (small non-pathological age deposits) and smaller than drusenoid RPEDs (e.g., 200-250 μm), would provide an entirely novel method for quantifying disease burden.

[0113] Mid-gray low fluorescence signal In some instances, AMDI identifies areas of grey that are neutral in tone - neither widespread hyperfluorescence / black, nor bright hyperfluorescence. These 2-dimensional areas occur in some, but not all, subjects and may occur in the presence or absence of pre-existing GA (e.g., as determined by FAF). Unlike widespread hypofluorescence / black areas, no structural abnormalities (e.g., no drusen, no RPE detachment, no GA) were observed. Instead, the RPE cells and retina appear anatomically normal in cross-section (e.g., OCT) as well as using other en face modalities (e.g., IR, FAF, red-free, color, angiography, OCTA, and others). These cells may be physiologically disturbed, damaged, or genetically different from normal cells, and thus unable to internalize dyes such as ICG. Thus, in some embodiments, AMDI can be used to identify and quantify dysfunctional RPE / retina.

[0114] Figure 20 shows comparative retinal images obtained by IR imaging, FAF imaging, and DNIRA imaging. Unlike FAF, areas of apparently "normal" RPE appear mid-gray in the DNIRA image. Unlike FAF, the DNIRA image reveals abnormal or "diseased" RPE as darker gray (highlighted by a yellow line) against a mid-gray background.

[0115] Processing of superfluorescent signals In some aspects described herein, in vivo functional imaging methods can be utilized for the detection of phagocytic immune cells, such as macrophages or microglia (collectively macrophages), in the eyes of subjects with ocular diseases or disorders. System 100 performing DNIRA identifies putative macrophages in the eyes of subjects with intraocular tumors and AMD, and places these potentially aggressive cells in close proximity to sites of tissue damage and inflammation. Furthermore, complex two-dimensional patterns of hyperfluorescence, normal fluorescence, and hypofluorescence are initially identified using DNIRA - these grayscale images can be used by system 100 to perform cognitive computing for their analysis and interpretation.

[0116] Currently, there may be no method for detecting immune cells in the eye of a living human, and because tissue biopsy is not a viable option, there are no methods to provide clinicopathological correlation between in vivo study results and postmortem tissue pathology.

[0117] In some cases, the system 100 can perform the following: 1) use registered DNIRA images to identify spatiotemporal distribution of putative macrophages; 2) use registered DNIRA images to identify co-distribution of putative macrophages with high-risk and late features of disease; 3) use registered DNIRA images to identify changes in spatiotemporal distribution and co-distribution of macrophages and high-risk and late features of disease over time; 4) identify novel complex patterns of disease phenotypes; and 5) compare DNIRA images to other en face and cross-sectional modalities over time. The system 100 can generally perform image output processing from the image acquisition unit, native image registration, and image segmentation. The system 100 can generally perform image output processing from the image acquisition unit. There may be challenges related to a) selection of the "best" subject images, b) slow conversion to non-sensitive image output formats, c) long directory names, and d) the need to be amenable to downstream batch processing. The system 100 can perform image output processing for selection of appropriate images for downstream analysis that is generally device independent. The system 100 can perform native image registration.

[0118] System 100 can generally perform image segmentation. As with post-registration, there may be challenges. Individual bright dots may be distinguished from surrounding black and grayscale background regions. Additional bright putative macrophage dots may be quantified over time. System 100 can generally perform image segmentation to identify the spatiotemporal distribution of bright macrophages. Macrophages are believed to be associated with disease activity. System 100 can generally perform image segmentation to select a region of interest "ROI". System 100 can generally perform image segmentation to threshold bright signals to separate them from an otherwise grayscale background. System 100 can generally perform image segmentation to identify an appropriate dot size. System 100 can generally perform image segmentation to quantify the number of bright DNIRA macrophages within the ROI. System 100 can generally perform image segmentation to indicate the margin of error for thresholding to capture bright signals. The system 100 can perform a segmentation of the image to indicate the range of tolerance for capturing the dot size generally.

[0119] System 100 can perform image segmentation to correlate the distribution of generally bright macrophages with sites of high-risk disease features. High-risk features of disease can predict disease progression. Correlating macrophages to these high-risk features can therefore provide another biological predictor of disease and increase predictive power. System 100 can perform image segmentation to select an ROI generally within an area of ​​disease activity. System 100 can perform image segmentation to quantify the number of bright DNIRA macrophages within an ROI, generally using the processes described herein.

[0120] System 100 can generally perform image segmentation to identify changes in spatiotemporal distribution and co-distribution of macrophages, as well as high-risk and late features of disease over time. Macrophages are highly plastic and dynamic cells that change spatially and temporally based on their tissue environment, and thus the changes may further help predict future disease activity. System 100 can generally perform image segmentation to threshold bright signals to separate them from an otherwise grayscale background. System 100 can generally perform image segmentation to identify appropriate dot sizes. System 100 can generally perform image segmentation to capture bright dots within an ROI for quantification. System 100 can generally perform image segmentation to compare bright dots within an ROI from one time point to the next time point showing changes in quantity (increase / decrease over time) or changes in localization within the same ROI (no increase / decrease, but spatial change).

[0121] The system 100 can perform image segmentation to generally identify complex patterns of disease phenotypes. Imaging macrophages in vivo in a subject's eye has not been previously reported, and disclosed herein is a method to identify new patterns that have not been previously identified and need to be characterized. The system 100 can perform image segmentation to generally threshold bright signals to separate them from an otherwise grayscale background. The system 100 can perform image segmentation to generally identify appropriate dot sizes. The system 100 can perform image segmentation to generally identify clusters of grouped dots. The system 100 can perform image segmentation to generally identify commonalities of clusters of dots to indicate different patterns that share characteristics across subjects. For example, similar numbers of grouped dots, similar distances between clusters, and similar localization patterns may provide indications.

[0122] System 100 can perform image segmentation to compare DNIRA images to other en face and cross-sectional modalities, generally over time. DNIRA provides features not previously seen by other ocular modalities, and comparisons to other modalities can enhance the capabilities associated with DNIRA, but also help identify potential features in those modalities whose biology was previously unknown. Example comparisons include DNIRA vs. FAF, DNIRA vs. OCT, DNIRA vs. IR, DNIRA vs. NIF-AF, DNIRA vs. RF, DNIRA vs. angiography, and DNIRA vs. OCTA.

[0123] The following provides an example of registration and analysis with respect to FIG. 20, which provides a graphical representation of example image data. The system 100 may register each time point using a linear stack registration with the SIFT plugin with the following (mostly default) settings (determined through trial and error): Initial Gaussian blur: 1.60 pixels, steps per scale octave: 3 pixels, feature descriptor size: 4 pixels, feature descriptor orientation bins: 8, nearest / next nearest ratio: 0.92, maximum registration error: 25 pixels, inclusion ratio = 0.05. The predictive transformation may be affine. The system 100 may remove a blurred version of each image from the original image. This should remove most of the image background, but the result may be low contrast. The blurred images provide a Gaussian blur of sigma = 50 μM. The system 100 may remove any remaining background using the "remove background" plugin with rolling ball radius = 4 μM. This helps remove any last traces of background. The system 100 can enhance the contrast, saturate 0.2% of the pixels, and normalize the histogram. The system 100 can use the "Bleach Correction" plug-in to stretch the histogram of time point 2 to mimic the pixel intensity of time point 1. This is done because averaging, shading, and more differences can alter the relative intensity of the dots between the two time points. A threshold can be applied so that the dots are similarly bright in both images. This plug-in was originally intended to compensate for the weakening of signal intensity when imaging sensitive fluorophores over long periods of time. The system can apply an "inter-mode" thresholding algorithm and convert it into a mask. In this step, anything above a threshold intensity value determined by the "inter-mode" algorithm can be captured and everything else can be removed. In theory, this should select only the brightest signals. In some embodiments, this selects all of the dots.

[0124] Additionally, system 100 can despeckle, watershed, and dilate (despeckle to remove noise, watershed algorithms to separate dots close to each other, and prevent the program from falsely identifying a collection of dots as a single large dot). System 100 uses the "Image Calculator" feature to step pixel by pixel and select for the "Max" and "Min" intensity values ​​between two time points, a description of which is outlined in the next slide. System 100 can filter out the "Max" image from the "Min" image. System 100 can filter out the "Max" image from the "Min" image with the following settings: size < 400 μM 2 , patches with circularity = 0.50 to 1.00 can be analyzed. 2 leads to a diameter of 22 μM. This is reaching the macrophage region in terms of size, but it also reaches the current resolution limit of DNIRA. The blurring associated with the point difference function means that the DNIRA dots are probably larger than their source. The system 100 can use appropriate means to perform a Mander colocalization test. This is a colocalization test commonly used for IHC, but is now being applied to test the system's ability to identify static vs. dynamic dots.

[0125] Superfluorescent DNIRA signal In some embodiments, DNIRA can identify small hyperfluorescent dots in the subject's eye. Surprisingly, some of these dots are stationary during the subject's visit. Meanwhile, others are dynamic, present at some times and absent at others, residing in different locations. Without wishing to be bound by any theory, these small hyperfluorescent dots identified using DNIRA may represent phagocytic immune cells, e.g., macrophages, that move, divide, appear or disappear over time. It has previously been demonstrated preclinically using DNIRA and two variants of DNIRA - ImmunoDNIRA and Pulsed-DNIRA - that cells, e.g., macrophages, can internalize ICG and then visualize it in the living eye using excitation / emission filters that match the dye's spectral properties. Disclosed herein is the unprecedented finding that stationary or dynamic populations of dots identified in the eyes of subjects with AMD are the appropriate size and shape to become immune cells, e.g., macrophages.

[0126] With reference to Figures 18 and 19, as a proof of concept, it is also observed that hyperfluorescent dots of similar size are present in eyes with uveal melanoma or high-risk indeterminate melanocytic lesions. They are absent from nevi and low-risk indeterminate lesions. Inflammation is a hallmark of cancer, and histopathological specimens of uveal melanoma, especially those of high grade (monosomy 3), are associated with large numbers of macrophages, whereas nevi are not. This correlation is confirmed in vivo using DNIRA to visualize tumor-associated macrophages (TAM) by a method named "TAMI" (TAM Imaging). Using TAMI, it is confirmed that regions of interest can be segmented based on their presence or absence of hyperfluorescent DNIRA-labeled dots. These sites correspond to differences in the concentration of bright dots that can be quantified. Hyperfluorescent dots may be identified by number, density, concentration, size or radius, fluorescence brightness, or location. Dot quantification correlates significantly with disease burden, as areas with a higher amount of dots correlate with the incidence of melanoma compared with indeterminate lesions or benign nevi.

[0127] Two-dimensional pattern of DNIRA signal In some embodiments, early studies described herein applying DNIRA / AMDI to subjects have identified novel two-dimensional (2D) grayscale patterns, as observed in Figures 21A-21E. These may occur centrally in the macula, or more peripherally in the macula or central retina. These patterns often have an interwoven, lacy, reticular, or speckled configuration. In some instances, patterns that are coarser (wider) in aspect or finer (closer) in aspect are observed. Without any precedent, in some cases, these are referred to as "loose patterns" (Figure 21A) or "solid patterns" (Figure 21B), respectively. These patterns may indicate different subtypes of AMD and therefore different responses to treatment options, as shown in Example 7.

[0128] Thus, in some embodiments, logical and software algorithms may be used to identify and classify or stratify subjects by their observed two-dimensional patterns using DNIRA / AMDI. Because these patterns are not easily described by traditional software programming (e.g., registration and segmentation), unstructured computational algorithms are applied to refine the description.

[0129] Delta - Comparison of FAF and DNIRA In some embodiments, methods are provided to distinguish areas of widespread hypofluorescence / black signal in FAF (where RPE / photoreceptors are dead and missing) from areas that are widespread hyperfluorescence / black in DNIRA. Compared to FAF, DNIRA images always have the same amount or a larger amount of widespread hypofluorescence / black areas than FAF does, allowing us to calculate the difference, or delta, shown in FIG. 14. This delta may represent RPE / photoreceptors with abnormal physiology rather than their loss. Thus, in some embodiments, DNIRA can detect areas of disease.

[0130] Some embodiments may relate to changes in DNIRA images compared to other imaging modalities (e.g., FAF, IR, color, red-free, segmented or non-segmented, cross-sectional or en face OCT, OCTA, which provide the difference, or delta, between the two modalities.

[0131] Moreover, it relates to the temporal changes of DNIRA over time compared to the temporal changes of FAF over time. For example, and without wishing to be bound by any theory, the widespread hypofluorescent / black DNIRA sites seen at time point 1 may precede and predict the changes seen with other modalities, such as FAF, at a later time. Thus, DNIRA can be seen as a method for detecting early changes. Thus, in some embodiments, DNIRA results in the identification of sites of RPE / photoreceptor damage prior to their loss and prior to the onset of widespread hypofluorescent / black sites of RPE / photoreceptor loss detected using FAF. Thus, DNIRA / AMDI can be used to identify subjects at risk of developing GA, which may then progress from, for example, early AMD to late dry AMD. In this way, DNIRA provides a novel method of identifying subjects for enrollment in clinical trials to prevent the onset of GA, or for treatment should GA occur.

[0132] Logic-biology-driven hardware and software for image processing Aspects described herein relate to the logical algorithms used to drive the design of the software and hardware. These include, but are not limited to, those described above, such as temporal changes in DNIRA and cross-modality differences between DNIRA and other methods, and cross-modality changes related to time. AMDI has not been reported before. The exemplary data disclosed herein is drawn from the first application of DNIRA to a clinical environment. Thus, each of these observations is novel. In some cases, the systems described herein may include software modules that implement the logical processes described herein. Additionally, the systems described herein may include an integrated image processing, image analysis, and image output platform.

[0133] In some aspects, such logic and software algorithms can identify, quantify, and localize sites of widespread hypofluorescent / black DNIRA over time and across modalities. In some aspects, such logic and software algorithms can identify, quantify, and localize hyperfluorescent dots and assess their number, size, and location. Such logic and software algorithms can identify, quantify, and localize bright hyperfluorescent dots to sites of widespread hypofluorescent / black DNIRA over time and across modalities. Without wishing to be bound by any theory, spatiotemporal localization of putative immune cells, such as macrophages, in close proximity to areas of RPE / photoreceptor loss is indicative of phagocytic activity with RPE / photoreceptor loss. Colocalization is described herein. Thus, DNIRA could be used to identify subjects suitable for immunomodulatory therapy.

[0134] In some aspects, the eye disease or disorder may be one or more of dry AMD, RPD, white dot syndrome (e.g., serpiginous choroidopathy, serpiginous retinopathy, acute posterior multifocal plaque-like pigment epitheliopathy (APMPPE), multiple vanishing white dot syndrome (MEWDS), acute zonular occult outer retinopathy (AZOOR), punctate inner choroidopathy (PIC), diffuse subretinal fibrosis (DSF)), late-onset retinal degeneration (LORD, e.g., Q1qTNF5 deficiency), and central serous retinopathy (CSR). In some aspects, the eye disease or disorder may be Lecithin Retinol Acryltransferase (LRAT) deficiency, optionally associated with lrat-related Leber congenital amaurosis, and retinal dystrophy, early onset, severe. In some aspects, the ocular disease or disorder is fundus albesic punctata, which may be associated with one or more of the following genetic locations: 3q22.1 (retinitis albesic punctata, RHO), 6p21.1 (retinitis albesic punctata, PRPH2), 12q13.2 (fundus albesic punctata, RDH5), 15q26.1 (retinitis albesic punctata, RLBP1), and 15q26.1 (fundus albesic punctata, RLBP1). In some aspects, the presence of phagocytic immune cells is measured by DNIRA.

[0135] In some embodiments, the ocular disease or disorder may be one or more of diabetic eye disease (e.g., diabetic retinopathy and DME), Vogt-Koyanagi-Harada disease (VKH), sarcoid uveitis, ocular histoplasmosis and / or presumed ocular histoplasmosis syndrome, idiopathic uveitis, autoimmune uveitis; uveitis associated with systemic diseases such as lupus, Crohn's disease, rheumatoid arthritis, and other known diseases of immune origin, posterior uveitis (including those that may not yet be diagnosed), anterior uveitis (e.g., iritis), Behcet's disease, polyarteritis nodosa, and Wegener's granulomatosis. This is shown in Example 21 and FIG. 39.

[0136] In some embodiments, RPD disease may be identifiable by the presence of one or more areas of distinct patterns of retinal imaging in the subject's eye. The retinal imaging may be one or more of white light, red-free light, blue light, FAF, near infrared (NIR), infrared (IR), angiography, and DNIRA, and / or the presence of one or more areas of abnormally fluorescent FAF in the subject's eye, and / or an increase (including a transient increase) in permeability through the subject's epithelial barrier between the choroid and retina compared to a non-disease state, and / or the presence of phagocytic immune cells across the subject's RPE compared to a non-disease state.

[0137] In some aspects, a method is provided that can determine whether an ocular disease or disorder in a subject will respond to treatment with an agent that inhibits or modifies the function of immune cells of the subject, comprising detecting the presence, detecting the absence, or measuring the amount of immune cells in the subject's eye, wherein the subject's eye fluoresces in response to light having a wavelength of about 600 nm to about 900 nm. In some embodiments, the light may have a wavelength of about 400 nm to about 900 nm.

[0138] The methods described herein may further include administering to the subject an effective amount of a fluorescent compound, and the detecting or measuring occurs at least one day after administration of the fluorescent compound. In some embodiments, the detecting or measuring may occur at least one day after administration of the effective amount of the fluorescent compound to the subject. In some embodiments, the methods described herein may further include detecting or measuring FAF in the subject's eye. In some embodiments, the methods described herein may further include correlating the FAF pattern to the presence, absence, or amount of immune cells in the subject's eye. In some embodiments, the detecting or measuring may include performing cSLO, FAF, DNIRA, OCT, or OCTA, and correlating these modalities to the presence, absence, or amount of immune cells in the subject's eye. In some embodiments, the immune cells may be cells of the subject's innate immune system and / or macrophages and / or microglial cells, dendritic cells, monocytes, mononuclear phagocytes, phagocytic immune cells.

[0139] DNIRA In various embodiments, system 100 can use optical imaging using various techniques. For example, such techniques include, but are not limited to, fundus photography, cSLO, FAF, angiography, OCT, OCTA, including three-dimensional reconstruction of such. In various embodiments, exposing the eye to light includes performing cSLO, FAF, DNIRA, angiography, or OCT. In various embodiments, the imaging is DNIRA. In various embodiments, any combination of the above techniques may be used.

[0140] The inventors have previously demonstrated that following systemic delivery of sodium iodate, patches of hypofluorescent FAF are not observed in vivo (in non-clinical settings, i.e., in non-human eyes) in areas of RPE loss, as would be expected from clinical trials (data not shown). However, by pre-labeling the RPE with a fluorescent dye, such as the near-infrared (NIR) dye indocyanine green (ICN), a technique called delayed near-infrared analysis (DNIRA), the RPE can be made visible using cSLO imaging. Once labeled, areas of RPE loss are revealed as quantifiable patches of hypofluorescence similar to those observed clinically. In various embodiments, NaIO3, FAF, and DNIRA may be used together to show, by way of example, the relationship between RPE loss, macrophages, macrophage polarization, and modulation of the M1 response.

[0141] In the case of DNIRA, for example, when administered at a given non-toxic dose, a compound suitable for fluorescence detection, including a near-infrared (NIR) dye such as ICG, can label the RPE and thus make it visible when viewed using an ICG excitation / emission filter in the following days or weeks. Importantly, this visualization in the following days and weeks can be without re-administration of the dye. Thus, in some aspects, a central component of the DNIRA technique is its timing. This differs from the current use of ICG or other blood contrast dyes, which are viewed immediately after injection, during the transit phase, or minutes to hours immediately after injection to determine the intravascular localization of the dye and its immediate extravasation.

[0142] In some embodiments, DNIRA can be used in experimental animals. In one embodiment, DNIRA causes patchy geographical areas of RPE loss (e.g., NAIO3), and is accompanied by administration (and optionally angiography) of a compound suitable for fluorescence detection, such as, but not limited to, ICG, for about one day or more prior to administration with a toxin or other agent, optionally followed by additional amounts of NaIO3 or another agent that causes the area of ​​patchy RPE loss to expand for about one day or more (or about one week, or about one month, or about three months). For example, other issues that cause the expansion of geographic atrophy (e.g., as the first administration, or the second, or the third, or the fourth administration) may be modulators of cell survival, cell death, survival, autophagy, proliferation, regeneration, etc.

[0143] In various embodiments, DNIRA techniques can be used in human subjects. For example, DNIRA in human subjects may not involve the use of toxins. DNIRA in human subjects may involve the evaluation of normal or disease-related changes in the eye using fluorescent dyes, with excitation / emission filters in place but no angiography, following a delay of hours or days after dye administration.

[0144] The expansion of geographic atrophy is a key outcome of clinical trial design that is acceptable to the US Food and Drug Administration (FDA). Thus, the methods and systems described herein can allow for the observation of geographic atrophy, particularly the expansion of geographic atrophy, in animal models, thus allowing for correlation between preclinical disease models and clinical trial design. The inability to clearly identify geographic atrophy, or the expansion of geographic atrophy, in animal eyes has prevented direct correlation between preclinical studies and clinical observations. Furthermore, in some embodiments, the systems described herein can allow for clinical assessment of the size and speed of expansion of geographic atrophy, including the expansion of geographic atrophy, in human subjects.

[0145] In some embodiments, the compound suitable for fluorescence detection may be suitable for imaging using various wavelengths of fluorescence. In some embodiments, these wavelengths range from visible light to infrared, such as from 390 nm to 1 mm, including blue light, white light, and near infrared. In some embodiments, the dye is a near infrared dye. In some embodiments, the dye is ICG.

[0146] In some embodiments, the DNIRA can be performed (and / or delayed near infrared fluorescence (DNIRF) is observed) about 6 hours, 12 hours, 24 hours, or about 2 days, or about 3 days, or about 4 days, or about 5 days, or about 6 days, or about 7 days, or about 10 days, or about 14 days, or about 21 days after administration. In some embodiments, the DNIRA can be performed at least 1 day after administration, or at least 2 days, or at least 3 days, or at least 4 days, or at least 5 days, or at least 6 days, or at least 7 days, or at least 10 days, or at least 14 days, or at least 21 days after administration. In some embodiments, the DNIRA can be performed at least about 2 hours, 4 hours, 8 hours, 24 hours, or at least about 7 days, or at least about 30 days, or at least 60 days, or at least 90 days after administration. In some embodiments, DNIRA may not be performed during the passage phase (e.g., the first passage of the dye as it flows through the blood vessels of the eye and into the eye tissue), or for several minutes thereafter. In some embodiments, angiographic imaging may not be required, thus distinguishing current dye-based imaging systems from DNIRA.

[0147] In some embodiments, visualization can be achieved using cSLO. In some embodiments, visualization can be achieved using white light and appropriate filters. In some embodiments, the ICG excitation / emission filters can be 795 nm (excitation) / 810 nm (emission) filters. In some cases, visualization can be achieved using a fundus camera or other ocular imaging device with the appropriate spectrum.

[0148] The RPE is a clinical epithelial monolayer that performs a "nurse cell" function for the specialized photoreceptors, rods, and cones of the eye. Eye diseases or disorders, such as AMD and RPD, are causally linked in part to abnormalities in the RPE, without wishing to be bound by any theory. DNIRA allows the RPE layer to be clearly identified in vivo in animal eyes. Furthermore, the primary technique used to detect the RPE in human eyes, FAF, is likely ineffective or insufficiently effective in rodent eyes, for example (as a non-limiting example), due to a relative paucity of fluorophores, such as lipofuscin. FAF imaging in human eyes is performed using the blue spectrum of non-coherent light in the presence of excitation / emission filters, or coherent blue light, to identify areas of absent RPE (e.g., hypofluorescent signal) or abnormal RPE (e.g., hyperfluorescent signal). In the absence of DNIRA, the inability to clearly identify the RPE in animal eyes has prevented direct correlation between preclinical studies and clinical observations.

[0149] Thus, in various aspects, methods are provided for visualizing the RPE layer in the eye of an animal for preclinical investigation of an eye disease or eye disorder, such as, for example, DNIRA. Further, as described herein, DNIRA or variations thereof can allow visualization of fluorescent immune cells in the eye of an animal. Further, as described herein, DNIRA or variations thereof can allow visualization of fluorescent immune cells in the eye of a human subject. In some aspects, the practice of the methods described herein with a human subject may not include administration of a toxin.

[0150] In some embodiments, DNIRA can be used in identifying drugs that are effective for treating eye diseases or eye disorders. In some embodiments, DNIRA can be used as a method for evaluating subjects who have or may have eye diseases or eye disorders (including, without limitation, AMD and RPD). In some embodiments, DNIRA can be a surrogate biomarker for the diagnosis and / or prognosis and / or progression of eye diseases or eye disorders (including, without limitation, AMD and RPD). For example, DNIRA can be used to identify patterns including lace-like, mesh-like, or leopard-print patterns of alternating hyperfluorescent DNIRA and hypofluorescent DNIRA that are not seen in other imaging modalities that indicate eye disease states (including, without limitation, AMD and RPD). DNIRA can also be used to identify and quantify areas of hyperfluorescent DNIRA and hypofluorescent DNIRA.

[0151] In various embodiments, DNIRA can be used to identify hypofluorescent features of the eye. For example, these areas appear black when imaged, thus allowing for easy quantification (in contrast to ICG imaging, or to hyperfluorescent signals, which are grayscale rather than black and white). In some embodiments, detection of hypofluorescent DNIRA can predict damaged or dead RPE. For example, hypofluorescent DNIRA may indicate one or more of the following: absence of RPE, abnormal / unhealthy RPE (which cannot take up ICG dye), RPE that is not in contact with Bruch's membrane (and thus is no longer in position to take up ICG dye from the choroidal vasculature), and the presence of lipids that may be located between the RPE and BM (thus preventing ICG uptake) or internal to the RPE (thus preventing RPE signal).

[0152] In various embodiments, DNIRA can be used to identify hyperfluorescent features of the eye. For example, these areas appear bright when imaged, thus allowing for easy quantification. Detection of hyperfluorescent DNIRA is, in some embodiments, predictive of macrophages, including M1 and / or M2 macrophages.

[0153] In various aspects, DNIRA can be used as a biomarker for diagnosis of ocular disease states (including, without limitation, AMD and RPD) and prompt further evaluation and / or treatment with one or more agents, including, without limitation, those described herein. In various aspects, DNIRA can be used as a biomarker for diagnosis of ocular disease states (including, without limitation, AMD and RPD) and prompt further evaluation and / or treatment with one or more agents, including, without limitation, those described herein. In various aspects, DNIRA can be used to improve identification of appropriate subjects for study recruitment and to assess disease progression. In various aspects, DNIRA can be used to monitor disease progression.

[0154] In various aspects, DNIRA can be used to identify sites of disease that may be amendable to gene therapy, stem / progenitor cell or other cell therapy, or combined gene / cell therapy. As a non-limiting example, sites suitable for rescue by cell replacement or nutrient support can be identified and targeted for therapy.

[0155] In various embodiments, DNIRA can be used as a comparative diagnostic for any of the agents described herein. In various embodiments, DNIRA can be used to evaluate a subject's response to any of the agents described herein (including evaluating the efficacy of the agents described herein and / or the likelihood of a response to any of the agents described herein). In various embodiments, the use of DNIRA can entail inclusion or exclusion criteria, or endpoint analysis for clinical trial design.

[0156] In various embodiments, the present invention relates to a method for assessing a binding eye disease using the systems and methods described herein. For example, in various embodiments, a method is provided for assessing a subject's eye using DNIRA (e.g., using the systems and methods described herein). In various embodiments, a method is provided for assessing a subject's eye by administering a fluorescent compound, such as indocyanine green (ICG), that is taken up by one or more cells (including, without limitation, retinal pigment epithelial (RPE) cells and cells of the immune system, such as macrophages), and exposing the eye to light having a wavelength of about 600 nm to about 900 nm, where the exposing occurs at least 24 hours (e.g., about 24 hours, 36 hours, or about 2 days, or about 3 days, or about 4 days, or about 5 days, or about 6 days, or about 7 days, or about 2 weeks, or about 1 month, etc.) after administering the fluorescent compound and evaluating the fluorescence pattern in the eye for one or more characteristics indicative of the binding eye disease or disease stage. Features indicative of a blinding eye disease may be selected and / or identified based on an analysis of a cohort of patients (e.g., including an experimental group with a blinding eye disease or stage thereof, and a comparator group without the blinding eye disease). Features may be selected and / or identified using pattern recognition and / or clustering analysis. The presence of the feature in the subject's image may be determined using a classifier algorithm that may be trained using images and analysis from the cohort(s).

[0157] Such methodologies allow for binning or classification of subject fluorescence pattern information to guide clinical trial design, such as, for example, directing enrollment or exclusion of patients from studies of binding eye disease treatments. For example, such methodologies, in various embodiments, allow for differentiation of subjects with binding eye disease from those without binding eye disease. Furthermore, in various embodiments, such methodologies allow for differentiation within or between subjects with blinding eye disease, for example, disease stage or disease progression, disease subtype, associated molecular genetics, pathogenetics, drug response, etc. Such differentiation and resulting clinical trial decision-making directions are further described elsewhere herein. Such differentiation and clinical treatment directions are consistent with customized or personalized clinical trial design or personalized medicine.

[0158] Moreover, in the presence of potential or various treatments, such methodologies allow for binning or classification of subject fluorescence pattern information to guide clinical decision making, such as, for example, directing treatment of a blinding eye disease as necessary. For example, such methodologies, in various embodiments, allow for differentiation between subjects with a blinding eye disease and those without a blinding eye disease. Furthermore, in various embodiments, such methodologies allow for differentiation within or between subjects with a blinding eye disease, for example, disease stage or disease progression, disease subtype, associated molecular genetics, pathobiology, drug response, etc. Such differentiation and resulting clinical decision direction are further described elsewhere herein. Such differentiation and clinical treatment direction are consistent with personalized medicine.

[0159] In various embodiments, the method allows for the comparison of eyes of a test group to eyes of a control group (e.g., without a blinding eye disease) to train an algorithm to classify images of subjects for the presence of a binding eye disease or the status or stage of a binding eye disease (e.g., progressing or regressing in response to treatment (if the method is being used for monitoring)). For example, in various embodiments, the method allows for the collection of a database of various patterns or features of eyes in a cohort, such cohorts including groups selected from diseased, non-diseased, early disease, late disease, disease subtypes, etc. Thus, in various embodiments, images are compared to allow for the extraction of features or patterns that may form the basis of classification. In various embodiments, pattern recognition using machine learning methodologies is used to continually enhance the ability of the classifier as further images are analyzed, which can provide a continually improved ability to distinguish conditions for clinical decision making (e.g., by providing an objective decision support tool to assist medical professionals in the diagnosis and prognosis of blinding eye conditions and to assist researchers in clinical trial design).

[0160] In various embodiments, upon analysis, various ocular image profiles allow extraction of various features that provide information, for example, about the presence, absence, extent, subtype, etc. of a blinding eye disease. Image profile analysis and / or feature extraction can utilize any suitable algorithm. Such algorithms may classify samples between groups that suffer from and do not suffer from a blinding eye disease. For example, samples may be classified based on the imaging features described herein in subjects with or suspected of having a blinding eye disease, for example, a non-blinding eye disease population (e.g., a cohort from the general population or a patient cohort with a disease other than a blinding eye disease) or for other blinding or non-blinding eye disease populations (e.g., a cohort with a particular blinding eye disease compared to a different binding or non-blinding eye disease). Various classification schemes are known for classifying samples between two or more groups, including decision trees, logistic regression, principal component analysis, naive Bayes models, support vector machine models, and nearest neighbor models. Furthermore, several different algorithms may be utilized for the analysis of the ocular imaging data. The algorithms may include, for example, machine learning single-part classifiers or anomaly detector algorithms. As indicated above, the detection algorithm may be based on the Support Vector Data Description (SVDD). Additional algorithms may include Support Vector Machines (SVMs), Relevance Vector Machines (RVMs), neural networks, neural analysis, large margin classifiers, kernel-based classifiers, classifiers based on Probability Density Function (PDF) estimators, classifiers based on Parzen PDF estimators, Bayesian classifiers, Constant False Alarm Rate (CFAR) detectors, classifiers based on fuzzy logic, and / or similar detection algorithms. Furthermore, predictions from multiple models or algorithms may be combined to generate an overall prediction.

[0161] In various embodiments, the evaluation methods utilize machine learning and computational intelligence techniques, such as deep neural networks, and combinations of supervised, semi-supervised, and unsupervised learning techniques, which in various embodiments provide image classification, image analysis, computer-aided diagnosis and prognosis, and / or pattern recognition information.

[0162] In various embodiments, the evaluation method utilizes a supervised algorithm (such as, by way of non-limiting example, linear regions, random forest classification, decision tree learning, ensemble learning, bootstrap aggregating, etc.) In various embodiments, the evaluation method utilizes an unsupervised algorithm (such as, by way of non-limiting example, clustering or association).

[0163] In various embodiments, the systems and methods enable pattern recognition of various features of relevance to blinding eye diseases, for example, as described herein.

[0164] In various embodiments, discriminant and / or classification analysis allows for binning of patients to inform regarding ocular status (e.g., presence or absence of disease, disease severity, disease identity, etc.).

[0165] In various embodiments, the systems and methods are useful in interpreting various ocular features, as described elsewhere herein.

[0166] In various embodiments, the present systems and methods are useful in detecting hypofluorescence patterns (such levels of hypofluorescence may be evaluated against an image standard, such as, for example, a decrease in fluorescence to about 90% or greater of hypofluorescence measured at the optic nerve head (ONH) (e.g., 60-90% of ONH, 20-60% of ONH)), including, by way of non-limiting example, the presence, absence, or degree of widespread hypofluorescence (PHoF), intermediate hypofluorescence (IHoF), and low hypofluorescence (LHoF) regions, which may have sharp (easily distinguished) or soft (blurred or fuzzy) boundaries, and complex fluorescence patterns (e.g., a "loose pattern" pattern, a "tight pattern" pattern, a "leopard print" pattern, a "gray stain" pattern, a "fingering potato" pattern, etc.).

[0167] In various embodiments, the present systems and methods are useful in detecting patterns of hyperfluorescence, including the presence, absence, or degree of widespread hyperfluorescence (PHrF), intermediate hyperfluorescence (IHrF), low hyperfluorescence (LHrF) regions, which may have sharp (easily distinguished) or soft (blurred or fuzzy) boundaries, or a variety of complex patterns including, by way of non-limiting examples, "bright dots," "bright points," "large bright points," and "halos."

[0168] In various embodiments, the systems and methods are useful in detecting complex two-dimensional patterns, such as, but not limited to, labeled "tight patterns," "loose patterns," "gray stains," "oil stains," "bullet holes," and the like.

[0169] In various embodiments, the systems and methods allow for detection of a transition from the presence of drusen or pseudodrusen in the absence of geographic atrophy, macular atrophy (MA), or choroidal neovascularization to the presence of drusen or pseudodrusen in the presence of geographic atrophy, macular atrophy, or choroidal neovascularization. In various embodiments, the systems and methods allow for detection of drusen regression (e.g., about 25%, or about 50%, or about 75%, or about 100% reduction in drusen).

[0170] In various embodiments, the systems and methods allow for detection of ocular tissue loss (e.g., about a 25%, or about a 50%, or about a 75%, or about a 100% reduction in ocular tissue loss).

[0171] In various embodiments, the systems and methods allow for detection of the presence, absence, or extent of drusen, pseudodrusen, fish-shaped lesions, inflammatory infiltrates, subretinal fluid, variably formed hyperpigmented or hypopigmented areas (target lesions) in the absence of geographic atrophy, patchy cutis, or choroidal neovascularization, or the presence of drusen, pseudodrusen, fish-shaped lesions, inflammatory infiltrates, subretinal fluid, variably formed hyperpigmented or hypopigmented areas (target lesions) in the presence of any of: geographic atrophy, patchy cutis, or choroidal neovascularization.

[0172] In various embodiments, the systems and methods allow detection of total number of drusen, size of drusen, presence of soft confluent drusen, hyperpigmentation, RPE detachment, hypopigmentation, hyporeflective subretinal material, hyperreflective dots, subretinal basement membrane and / or basal linear deposits, as shown in Example 22 and FIG. 40.

[0173] In various embodiments, the present systems and methods allow for detection of the extent of regions of low fluorescent DNIRA signal (utilizing, as non-limiting examples, area, square root of area, circumference, diameter of best-fit circle) that are consistent with high-risk features, such as, for example, total number of drusen, size of drusen, area of ​​drusen, volume of drusen, drusen border, presence of soft fuzzy drusen, presence of soft confluent drusen, hyperpigmentation, RPE detachment, hypopigmentation, hyporeflective subretinal material, hyperreflective dots, subretinal basement membrane or basal linear deposits, etc. An example of this is illustrated in Figure 40 and shown in Example 22.

[0174] In some embodiments, the methods provide for pattern recognition in the fluorescence of the RPE and outer retina of a subject, including patterns based on loss of RPE and outer retina fluorescence. The RPE and outer retina fluorescence patterns can be used to train classifier algorithms for the assessment of blinding eye diseases.

[0175] Thus, in various embodiments, the present invention provides ocular imaging-based biomarkers to enable ocular evaluation. In some embodiments, such ocular imaging-based biomarkers are to be used alone or in conjunction with other ocular imaging-based biomarkers (e.g., but not limited to, color, FAF, IR, NIR, OCT, OCT-A, angiography, etc.) or non-imaging-based biomarkers (e.g., including but not limited to, genomics, tissue biochemistry, disease functional measures, and other outcome measures such as, for example, QALYs (quality-adjusted life years)) to enhance disease diagnosis, to identify previously known and unknown subtypes of disease, to identify patients likely to develop disease, to identify patients likely to suffer from disease progression, to identify patients likely to respond or not respond to a particular intervention (e.g., therapy), and / or to identify patients whose future outcome or safety may be compromised by disease progression, disease intervention, or environmental influences.

[0176] In various embodiments, the present invention provides a method for identifying patients with blinding eye diseases early in the disease process / progression. If detected early, early intervention can be performed. Alternatively, the present invention provides a method for identifying patients with blinding eye diseases late in the disease process / progression. If detected late, less burdensome intervention can be performed.

[0177] In various embodiments, the invention provides methods for detecting patients likely to progress from early stage disease to late stage disease (eg, prognostic biomarkers).

[0178] In various embodiments, the invention provides methods for measuring or monitoring the rate of disease progression over time or the effect of an intervention (e.g., disease response) to one or more therapeutic agents (e.g., prognostic, predictive, or monitoring biomarkers).

[0179] In various embodiments, the present invention provides methods for determining patient prognosis, for example to identify the likelihood of clinical signs, disease recurrence, or disease progression in patients with blinding eye diseases such as AMD, RPD, maculopathy, central serous retinopathy, uveitis, inherited retinal degeneration, and other diseases mentioned elsewhere herein.

[0180] In various embodiments, the present invention provides methods for predicting the response of a patient or group of patients to an intervention (e.g., a therapy) to identify individuals who are more likely to experience favorable or unfavorable effects from exposure to a pharmaceutical or environmental agent than similar individuals without the use of biomarkers. For example, such methods predict that a patient may be a responder to a therapy. If a responder, in various embodiments, such a patient may be instructed to undergo treatment with the therapy. Alternatively, such methods predict that a patient may not respond to the therapy. If not, in various embodiments, such a patient may be instructed not to undergo treatment with the therapy and thus may be directed to an alternative therapy (e.g., a prognostic, predictive, monitoring, or safety biomarker).

[0181] In various embodiments, the present invention provides methods for indicating the occurrence of a biological response in an individual who has been exposed to a pharmaceutical or environmental agent (e.g., a predictive or safety biomarker).

[0182] In various embodiments, the present invention provides complementary biomarkers to drive clinical trial design. In various embodiments, the present invention provides companion biomarkers to drive clinical trial design. In various embodiments, the present systems and methods allow for the selection of patients likely to respond to a clinical trial drug or other clinical trial intervention, thus allowing for the recruitment of appropriate populations for clinical trials.

[0183] In various embodiments, the present invention provides methods for personalized drug development or custom medicine, for example, the methods allow for interventions and / or products to be tailored to individual patients based on the individual patient's predicted response or risk of disease, in various embodiments.

[0184] Illustratively, in various embodiments, the method provides a diagnostic biomarker or method. For example, the method and system identify a complex three-dimensional pattern of reduced hypofluorescence signal in patients with a family history of blinding eye disease (e.g., AMD) but no personal clinical diagnosis. In this way, DNIRA may allow early diagnosis of patients or subsets of patients likely to develop the disease. In contrast, patients without a known family or personal history do not have the same pattern of hypofluorescence. Such information indicates a ready intervention. In various embodiments, the diagnostic biomarker or method may allow early diagnosis of patients with a family or personal history of blinding eye disease (e.g., AMD) but lacking knowledge of such. Thus, such methods may indicate early treatment of patients who would otherwise go untreated.

[0185] In various embodiments, the present invention informs treatment with one or more therapeutic agents, such as, for example, without limitation, semapimod, MIF inhibitors, CCR2 inhibitors, CKR-2B inhibitors, 2-thioimidazole, CAS445479-97-0, CCX140, clodronic acid, clodronic acid-liposomal preparations, or gadolinium chloride, but also anti-VEGF agents, ACE inhibitors, PPAR gamma agonists or partial agonists, renin inhibitors, steroids, and agents that modulate autophagy.

[0186] Non-limiting examples of anti-VEGF agents useful in the present methods include ranibizumab, bevacizumab, aflibercept, KH902 VEGF receptor-Fc, fusion proteins, 2C3 antibody, ORA102, pegaptanib, bevasiranib, SIRNA-027, decursin, decursinol, picropodophyllin, guggulsterone, PLG101, eicosanoid LXA4, PTK787, pazopanib, axitinib, CDDO-Me, CD DO-Imm, shikonin, beta-, hydroxyisovaleryl shikonin, ganglioside GM3, DC101 antibody, Mab25 antibody, Mab73 antibody, 4A5 antibody, 4E10 antibody, 5F12 antibody, VA01 antibody, BL2 antibody, VEGF-related protein, sFLT01, sFLT02, peptide B3, TG100801, sorafenib, G6-31 antibody, fusion antibody, and antibody that binds to an epitope of VEGF. Additional non-limiting examples of anti-VEGF agents useful in the present method include substances that specifically bind to one or more of human vascular endothelial growth factor-A (VEGF-A), human vascular endothelial growth factor-B (VEGF-B), human vascular endothelial growth factor-C (VEGF-C), human vascular endothelial growth factor-D (VEGF-D) and human vascular endothelial growth factor-E (VEGF-E), and antibodies that bind to an epitope of VEGF.

[0187] In some embodiments, the anti-VEGF agent is the antibody ranibizumab or a pharma- ceutically acceptable salt thereof. Ranibizumab is marketed under the trademark LUCENTIS. In another embodiment, the anti-VEGF agent is the antibody bevacizumab or a pharma- ceutically acceptable salt thereof. Bevacizumab is marketed under the trademark AVASTIN. In another embodiment, the anti-VEGF agent is aflibercept or a pharma- ceutically acceptable salt thereof. Aflibercept is marketed under the trademark EYLEA. In one embodiment, the anti-VEGF agent is pegaptanib or a pharma- ceutically acceptable salt thereof. Pegaptanib is marketed under the trademark MACUGEN. In another embodiment, the anti-VEGF agent is an antibody or antibody fragment that binds to an epitope of VEGF, such as an epitope of VEGF-A, VEGF-B, VEGF-C, VEGF-D, or VEGF-E. In some embodiments, the VEGF antagonist binds to an epitope of VEGF such that the binding of VEGF and VEGFR is inhibited. In one embodiment, the epitope comprises a component of the three-dimensional structure of VEGF, where the epitope is exposed on the surface of the folded VEGF molecule, hi one embodiment, the epitope is a linear amino acid sequence derived from VEGF.

[0188] In various embodiments, the methods inform treatment with one or more therapeutic agents, including, but not limited to, vitamin supplements, complement inhibitors or activators, visual cycle modulators, amyloid inhibitors or inhibitors, neuroprotective agents, autophagy inhibitors, anti-inflammatory agents, modulators of macrophage behavior or activity, modulators of CD36, tyrosine kinase inhibitors, RNA modulators, gene therapy, or cell therapy. Examples of these include APL-2, eculizumab, LFG316, sirolimus, fuocinolone acetonide (Iluvien), fenretinide, emixustat, trimetazidine, alprostadil, moxaverine, sildenafil, MC-1101 (MacuClear), OT5551, tandospirone, GSK933766, lampalizumab, ARC-1905, RNG6, olacea, brimonidine, tartrate, FMX-103, CLG-561, BIO-201, BIO-203, tesidormab, These include isopropyl unoprostone, GS-030, ICR-14967, methotrexate, ONL-1204, RST-001, TT-231, KSI-401, OCU-100, RC-1 alpha, ACU-02, ACU-3223, alprostadil, AVT-2, HMR-59, INO-6001, INO-6002, MCT-355, NT-501, SIR-1046, SIR-1047, SIR-1076, zinc monocysteine, EBI-028, EG-30, RNA-144101.

[0189] In various embodiments, the methods inform treatment with one or more therapeutic agents, such as, for example, complement inhibitors or activators, visual cycle modulators, amyloid inhibitors or inhibitors, neuroprotective agents, autophagy inhibitors, anti-inflammatory agents, and modulators of macrophage behavior or activity.

[0190] In various embodiments, the methods provide for identification of disease subtypes, for example, among patients with a known diagnosis of a blinding eye disease (eg, AMD).

[0191] Currently, classification of AMD is limited to early (drusen), late (which may be wet (neovascular) or dry (with GA)). The coexistence of variably sized drusen (small, medium, and large) and pigment seen on clinical examination and fundus color photography may further stratify early disease into low-risk and medium-risk (or high-risk early) subcategories. However, given the multigenicity of AMD, with over 30 genes and 200 SNPs, this represents a significant simplification and precludes the development of broadly targeted (through over-inclusion clinical trial designs) or subtype-specific, e.g., customized, and ultimately personalized, therapies. Furthermore, the novel vivid images obtained using DNIRA, which acts as an image-based biomarker with well-defined and ill-defined features, are particularly well suited to the use of cognitive computing for novel AI-derived pattern recognition. Combined with genetic, epigenetic, and lifestyle data, DNIRA provides previously unknown correlations that point to efficient, customized clinical trial design and personalized therapy.

[0192] By way of example, subtypes in the context of wet AMD include separation by location of CNV (e.g., subfoveal, parafoveal, extrafoveal, and peripapillary) and / or traditional classification (e.g., completely classic, predominantly classic, minimally classic, and no occult classic).

[0193] In various embodiments, the method is utilized in distinguishing (e.g., differentiating, discriminating, perceiving) diseases of similar clinical presentation to aid in diagnostic accuracy and new diagnostic classification. For example, the method reveals complex image-based patterns that are absent in patients with ocular conditions other than AMD or that are frequently misdiagnosed as AMD. Reciprocally, the method may lead to the correct diagnosis of certain diseases that would otherwise be diagnosed as AMD, such as, but not limited to, Stargardt's disease, adult vitelliform disease, Bietti crystalline corneal retinal dystrophy, pattern dystrophy, and others. These other conditions may be included in a comparator cohort to enhance the ability of discrimination. Thus, in various embodiments, the method allows for improved diagnostic accuracy, enhanced clinical trial enrollment, and directed and more appropriate personalized therapy plans for individual patients.

[0194] In various embodiments, the method is utilized in monitoring the progression and / or efficacy of treatment of blinding eye disease over time. Thus, in various embodiments, the method is utilized as a monitoring biomarker. For example, illustratively, the ability to measure dark (e.g., low fluorescence) areas using the method makes it a useful monitoring biomarker. When continuously evaluated over time, the method detects and measures the presence, status, or extent of blinding eye disease, such as AMD, or provides evidence of intervention effects or exposure, including exposure to pharmaceuticals or environmental factors.

[0195] In various embodiments, the method is utilized to quantify or better qualitatively describe known disease aspects to predict progression (prognostic biomarkers). In various embodiments, the method identifies and quantifies known high-risk features of the disease, such as large soft and confluent drusen. The AREDS 1 and 2 clinical studies identified risk factors that correlate with an increased risk of progression from early to late AMD, and among these strongest predictors of progression is the presence of large soft confluent drusen with or without pigmentary changes. Soft drusen can be identified during clinical examination and documented using color fundus photography, but they have inherently unclear or "fuzzy" boundaries that prevent their precise quantification even with alternative imaging methods, such as optical coherence tomography (both transverse and en face). The methods allow for the detection and quantification of risk features, such as soft drusen, and therefore can identify patients at risk, stratify the risk of progression, and direct which patients to enroll in clinical trials, particularly trials aimed at preventing either late stage disease, choroidal neovascularization, geographic atrophy, or patchy cutis.

[0196] In various embodiments, the methods are used to quantify aspects of the disease that are currently hypothesized to predict progression (predictive biomarkers). In various embodiments, the methods identify, quantify, and follow over time spatiotemporal changes in disease features that are predicted to correlate with disease, but that are currently not quantifiable and therefore not evident. For example, in various embodiments, the methods can quantify total drusen load, changes in same over time, etc.

[0197] Moreover, over time, for example, large and medium drusen refer to disease risk and progression risk, but their dynamic changes are difficult to observe. The method allows the establishment of a Dynamic Drusen Index (DDI) that can calculate the total drusen burden at a single time point or multiple time points, or calculate the changes of a specific drusen subtype. Depending on the current classification of drusen, the DDI is defined as total DDI (for all drusen types), confluent dynamic large drusen index (LC-DDI), dynamic large drusen index (L-DDI), dynamic medium drusen index (M-DDI), and dynamic small drusen index (S-DDI). In various embodiments, the DDI indicates which patients are to be prioritized for participation in a specific clinical trial.

[0198] In various embodiments, the method identifies putative macrophages in the eye of a patient with a disease, such as, for example, without limitation, a disease described herein (e.g., AMD, RPD, inflammatory disease, inherited macular degeneration, retinal degeneration, and ocular tumors), thus identifying individuals who are more likely to experience favorable or unfavorable effects from exposure to pharmaceuticals or environmental factors, such as macrophage modulation therapy, than similar individuals without the use of biomarkers. For example, despite the current understanding that dry AMD is a disease of the innate immune system, there is currently no therapy available for dry AMD. Phase III studies targeting the complement cascade have not yet been successful. In various embodiments, the method allows for the investigation of the celluar arm of the innate immune system. For example, it allows for the investigation of macrophage populations (collectively, all phagocytic immune cells, including, but not limited to, perivascular macrophages, microglia, resident parenchymal macrophages, dendritic cells, and circulating monocytes) and / or the quantification of some macrophages over time or with treatment versus without treatment.

[0199] In some embodiments, the methods provide a method for demonstrating the occurrence of a biological response in an individual who has been exposed to pharmaceutical agents, such as, but not limited to, bindarit, its derivatives and related compounds, macrophage modulating agents, methotrexate, IkBa inhibitors and NFkB inhibitors, complement modulators, other anti-inflammatory compounds, or environmental factors, either alone or in combination.

[0200] In various embodiments, the methods are used to qualitatively assess and quantify pharmacodynamic responses to such treatments, thereby being useful for establishing proof of concept that a drug or other therapy produces a desired pharmacological response in humans that is believed to be associated with clinical benefit, and for guiding dose-response studies, in various embodiments, the methods can further provide evidence of target engagement.

[0201] In various embodiments, the methods are used to qualitatively assess and quantify the pharmacodynamic response to stem cell therapy - for example, the possible beneficial effects of macrophage modulation on their survival and / or use to restore damaged or absent RPE.

[0202] In various embodiments, exemplary blinding eye diseases relevant to the present evaluation methods include, for example, dry AMD, wet AMD, reticular pseudodrusen (RPD), LOAD (late onset retinal degeneration), retinal degenerations associated with clqTNF5 deficiency or its corresponding genetic mutations, or other macular degenerations including, but not limited to, Stargardt's disease, pattern dystofia, as well as retinitis pigmentosa (RP) and related diseases. In one embodiment, the macular degeneration is hereditary. In another embodiment, the blinding eye disease associated with the present evaluation method is an idiopathic disease that may be characterized by retinal inflammation with or without macular degeneration, including, but not limited to, white dot syndrome (e.g., serpiginous choroidopathy, serpiginous retinopathy, acute posterior multifocal platelet pigment epitheliopathy (APMPPE), multiple vanishing white dot syndrome (MEWDS), acute zonular occult outer retinopathy (AZOOR), punctate inner choroidopathy (PIC), diffuse subretinal fibrosis (DSF)), without wishing to be bound by any theory. In another embodiment, the blinding eye disease associated with the present evaluation method is central serous retinopathy (CSR). In another embodiment, the blinding eye disease associated with the present evaluation method is a retinopathy, including diabetic retinopathy.

[0203] In various embodiments, the evaluation methods are utilized in the evaluation of ocular tumors. For example, in some embodiments, the methods allow for the detection of macrophages, e.g., tumor-associated macrophages. For example, the methods allow for the detection of hyperfluorescent bright dots and / or the quantification of hyperfluorescent bright dot density, which may be correlated with clinical classification (e.g., as determined by clinician observation / sorting), thus allowing for standardization of diagnosis and less subjective classification schemes at all times. In various embodiments, the methods allow for the detection or prediction of ocular tumor associated fluids, ocular tumor height, etc.

[0204] In various embodiments, the present methods of evaluation include, for example, basal cell carcinoma, biliary tract cancer, bladder cancer, osteosarcoma, cancer of the brain and central nervous system, breast cancer, cancer of the peritoneum, cervical cancer, choriomas, colon cancer, connective tissue cancer, gastrointestinal cancer, endometrial cancer, esophageal cancer, eye cancer, head and neck cancer, gastric cancer (including gastrointestinal cancer), glioblastoma, liver cancer, hepatoma, intraepithelial neoplasia, kidney or renal cancer, laryngeal cancer, leukemia, liver cancer, lung cancer (including GI cancer), and the like. For example, small cell lung cancer, non-small cell lung cancer, lung adenocarcinoma, and squamous cell carcinoma of the lung), melanoma, myeloma, neuroblastoma, oral cancer (lips, lower, mouth, and pharynx), ovarian cancer, pancreatic cancer, prostate cancer, retinoblastoma, rhabdomyosarcoma, rectal cancer, respiratory system cancer, salivary gland cancer, sarcoma, skin cancer, squamous cell carcinoma, stomach cancer, testicular tumors, thyroid cancer, uterine cancer, urinary system cancer, vulvar cancer, (low-grade / follicular non-Hodgkin's lymphoma) Lymphomas, including Hodgkin's lymphoma and non-Hodgkin's lymphoma as well as B-cell lymphomas (including non-small cell lung cancer (NHL)), small lymphocytic (SL) NHL, intermediate-grade / follicular NHL, intermediate-grade diffuse NHL, high-grade immunoblastic NHL, high-grade lymphoblastic NHL, high-grade small noncleaved cell NHL, bulky mass disease NHL, mantle cell lymphoma, AIDS-related lymphoma, and Waldenstrom's lymphoma It is used in the evaluation of tumors of the eye, such as macroglobulinemia, chronic lymphocytic leukemia (CLL), acute lymphocytic leukemia (ALL), hairy cell leukemia, chronic myeloid leukemia, and other carcinomas and sarcomas, as well as tumors of the eye that are one or more of the following: post-transplant lymphoproliferative disorder (PTLD), abnormal blood vessel proliferation associated with neurocutaneous syndromes, edema (e.g., edema associated with brain tumors), and Meigs syndrome (e.g., metastases to the eye).

[0205] In various embodiments, the subject evaluation methods are used in the evaluation of ocular tumors, such as, for example, primary intraocular cancer (e.g., without limitation, intraocular melanoma or primary intraocular lymphoma), hi various embodiments, the subject evaluation methods are used in the evaluation of retinoblastoma or medulloepithelioma.

[0206] Image analysis platform and output The systems described herein can provide output in the form of either raw images for user analysis or processed images for end-user (consumer) utilization. The output may be represented as visual elements on a GUI display. In one embodiment, the output may consist of raw images resulting from application of DNIRA to subjects with AMD and / or other blinding eye diseases prior to their registration, alignment, enhancement, alignment, segmentation, or any other relevant image processing, provided as single images or composite images arranged randomly or sequentially.

[0207] In some embodiments, the output may consist of processed images resulting from application of the DNIRA to subjects with AMD and / or other blinding vision diseases after initial processing, which may include any or all of registration, alignment, enhancement, rectification, segmentation, or any other relevant image processing, provided as single images or composite images arranged randomly or sequentially.

[0208] The systems and devices described herein can provide image exploitation by and / or for independent users through a centralized image analysis center or platform. Independent users include, but are not limited to, physicians, vision care professionals, ophthalmic technical staff, vision scientists, contract research organizations, drug development departments, and those involved in clinical trial design and execution. The centralized image platform could include, by way of non-limiting examples, cloud-based systems, web-based systems, network-based systems, and locally active network (LAN)-based systems.

[0209] The systems and devices described herein can provide independent and / or centralized users with a graphic user interface (GUI) compatible with multiple systems for desktop, laptop, or handheld devices. The systems and devices described herein can provide centralized users with DNIRA image software for additional analysis. The DNIRA image software may be implemented for such analysis. The DNIRA image software enables processes such as, but not limited to, registration, alignment, contrast adjustment, sharpening, segmentation, etc., for structured analysis of predefined features, such as regions of hypofluorescent signals, hyperfluorescent signals, hyperfluorescent dots, complex two-dimensional patterns, etc.

[0210] The systems and devices described herein can provide a focused user with DNIRA image software that enables processes such as, but not limited to, registration, alignment, contrast adjustment, sharpening, segmentation, and the like, for unstructured analysis of undefined or complex features other than those addressed by structured analysis.

[0211] Graphic User Interface The system described herein can provide an iteratively improved graphic user interface (GUI) designed to optimize image manipulation and analysis guided by everyday and expert users. Such GUIs could be made available on handheld devices, iPads, laptops, and other personal computing systems. Such GUIs could be embedded into commercially available imaging equipment or purpose-specific devices optimized for DNIRA.

[0212] In aspects, the GUI may include a series of visual elements that dynamically update based on control commands and image output data as described herein. The GUI may include, for example, an image carousel that is a magnified image, a grid tiled array, or a non-magnified tiled array.

[0213] Software embedded chips and / or software programs for widespread use In some aspects, DNIRA images acquired from any suitable imaging system (e.g., cSLO or non-cSLO based) can be processed. Device independent software programs can be developed. In some aspects, the software programs may be made available for incorporation into current or future devices. In other aspects, the software programs are made available to current device and software manufacturers for post-acquisition image analysis.

[0214] A converged software system for data analysis and iterative program development. In some aspects, the systems described herein may include centralized image processing, which may be, for example, without limitation, cloud-based, internet-based, locally accessible network (LAN)-based, or a dedicated reading center using pre-existing or new platforms.

[0215] In some aspects, the software will rely on structured computations to provide, for example, registration, segmentation, and other functions where the centralized processing output is ready for downstream analysis. In this case, the output will be a single or a series of adjusted images (e.g., alignment, contrast, sharpness, etc.) provided to the user in a random or non-random order for subsequent use.

[0216] In a variation of this embodiment, the software relies on structured computation and the output is pre-analyzed for the user and compared against relevant modifiable or non-modifiable risk factors for disease. In this case, the output would be a single or series of adjusted images (e.g., alignment, contrast, sharpness, etc.) provided to the user along with a qualitative or quantitative description of the image content(s).

[0217] In some embodiments, the software relies on unstructured computation, artificial intelligence, or deep learning, whereby regions of variable grayscale, in addition to easily defined features (such as widespread hypofluorescent / bright regions or hyperfluorescent dots), will be analyzed iteratively using layered or multi-layered processing.Unstructured DNIRA analysis will be particularly suitable for cloud-based, networked, or reading center-based platforms, since unstructured computation benefits from the growing number of images.

[0218] In a variation of this embodiment, the software relies on unstructured computation, whereby high complexity sites will be analyzed iteratively using layered or multi-layered processes. As a non-limiting example, these may include complex two-dimensional patterns of varying grayscale density. The output will be a single or a series of adjusted images (e.g., alignment, contrast, sharpness, etc.) provided to the user, potentially over time, along with a qualitative or quantitative description of the image content(s), and for example, with and without processing.

[0219] In a further variation of this embodiment, the software would resort to unstructured computation, i.e. so-called "artificial intelligence" or "deep learning", whereby sites of variable greyscale, in addition to easily defined features (e.g. widespread hypofluorescent / bright areas or hyperfluorescent dots), would be analysed iteratively using a layered or multi-layered process, in which case output would be provided to the user along with a qualitative or quantitative description of the image content(s), potentially over time, and for example with and without processing.

[0220] AMDI phenotyping and genotyping The present system for classifying or describing subjects with AMD is limited. For example, in commonly used systems based on clinical examination and color fundus photography, subjects are classified as either early or late. All subjects with early disease are considered "dry" and are commonly diagnosed by the presence of 5 or more drusen larger than a "droplet" (<10 μm). Late AMD can be exudative (neovascular or "wet") or non-exudative (atrophic or "dry"), both of which are associated with blindness, and both forms of late disease can coexist. Late "wet" AMD can coexist with early dry or late dry.

[0221] Commonly used systems use so-called "simplified scores" to identify subjects at high risk of progressing from early to late disease (either exudative or nonexudative). Thus, the presence of two or more large drusen (>125 μm) in one eye, or pigmented masses in one eye, respectively, results in a score of 1; thus, the presence of both in both eyes results in a score of 4. Subjects with a score of 3 or higher are more likely to progress from early to late AMD. In some systems, these subjects are classified as "high risk early" or in some systems as "intermediate" AMD.

[0222] In contrast to these simple classifications, AMD is known to be a complex multigenic disease with over 20 associated mutations / SNPs that may be inherited singly or in combination. Disease manifestations are further modified by epigenetic risk factors such as smoking cessation, diet, and nutrient supply.

[0223] FAF is the standard method for the analysis of dry AMD, where sites of widespread hypofluorescence / dark signal indicate the presence of GA and therefore late stage disease. Complex two-dimensional patterns of hyperfluorescent FAF may also be observed, some associated with increased patch expansion when found in conjunction with patches of GA. These include, for example, streaky and diffuse (small speckled) patterns. To date, there is no clear correlation between the genetics of AMD (genotype) and the clinical pattern (phenotype) as determined by clinical examination or special imaging, e.g., FAF or OCT.

[0224] AMDI can identify previously unknown image-based patterns of disease, providing, for example, a novel phenotypic description of individual subjects.Therefore, there is a need for an image-based system that can stratify subjects into different subtypes of disease, which may differ according to their rate of progression, susceptibility to epigenetic risks, response to different therapies, and ultimately blindness rates.AMDI can dramatically improve the ability to define these groups.Because AMDI provides an unprecedented description of different patterns or subtypes of disease, the results of the centralized imaging platform, which uses structured and unstructured learning and is iteratively improved, can then be compared with GWAS analysis, epigenetic status, and other medical or demographic information.

[0225] Ecology of DNIRA Figures 22 and 23 are graphic illustrations of the biology of DNIRA signals in RPE and macrophages, respectively. Figure 22 shows the structural and functional aspects of DNIRA signals in RPE, with widespread hypofluorescent / black signals occurring in association with several pathological changes. Green arrows indicate the transfer / uptake of ICG dye from the choroidal circulation into the RPE monolayer. "Normal" levels of DNIRA signals are shown as bright green cells that can internalize the dye. In the bottom panel, normal RPE, Bruch's membrane (BrM), and choroidal vasculature are shown along with commonly observed changes that occur in dry AMD. Hypofluorescence is observed in association with GA, RPE detachment, and drusen or other sub-RPE deposits (basal linear, basement membrane deposits). Mid-gray signals are shown in areas where the RPE is in its correct anatomical position relative to the choroidal vessels, but where cellular metabolism is presumed to be abnormal. Figure 23 shows structural and functional aspects of the DNIRA signal of macrophages. Hyperfluorescent dots (of a particular size, shape, and motility) may appear in conjunction with the signal obtained in Figure 22. The green arrows indicate the movement / uptake of dye from the choroidal circulation to the retina / RPE layer where tissue macrophages can internalize the dye. In addition to the fact that macrophages can internalize the dye directly in the circulation before the cells themselves flow into the retina / RPE tissue. During inflammation, increased phagocytic activity may correlate with increased dye uptake and the generation of brightly fluorescent dots. Also, normally located deep within the choroid (macrophages) or inner retina (microglia), their recruitment to sites of RPE and photoreceptor damage would suggest the presence of bright DNIRA dots.

[0226] DNIRA can identify previously unknown image-based patterns of disease, providing a novel phenotypic description of an individual subject, for example. Thus, there is a need for an image-based system that can stratify subjects into different subtypes of disease, which may differ according to their rate of progression, susceptibility to epigenetic risks, response to different therapies, and ultimately blindness rates. DNIRA can dramatically increase the ability to define these groups. DNIRA can dramatically increase the ability to define these groups. Because DNIRA provides an unprecedented description of different patterns or subtypes of disease, the results of the centralized imaging platform, which uses structured and unstructured learning and is iteratively improved, can then be compared with GWAS analysis, epigenetic status, and other medical or demographic information.

[0227] pigment Although ICG dyes have been referenced, in some embodiments, other fluorescent compounds may be used in place of ICG and included in any of the embodiments described above. Such fluorescent compounds may be suitable for imaging using fluorescence of various wavelengths. In some embodiments, these wavelengths range from visible light to infrared, including, for example, blue light, white light, and near infrared, such as 390 nm to 1 mm. In some embodiments, the dye may be a near infrared dye.

[0228] In some embodiments, the fluorescent compound absorbs light at a wavelength between about 600 nm and about 900 nm and / or emits light at a wavelength between about 750 nm and about 950 nm. In some embodiments, the fluorescent compound may have the following characteristics: about 795 nm (excitation) and about 810 nm (emission).

[0229] The systems and methods described herein are further illustrated by the following non-limiting examples.

[0230] Working Example Example 1: DNIRA to identify, quantify, and follow sites of widespread low fluorescence / black signal Since DNIRA analysis has not been reported before, the present invention believes that all observations are novel. DNIRA can be used to identify areas of widespread hypofluorescence / black signal. These may or may not coincide with areas or distinguishing features identified by other imaging modalities. Unlike the widespread hypofluorescence / black signal of FAF, which represents the absence of RPE / photoreceptors and therefore remains the same in size or expands over time, the dark areas of DNIRA are dynamic and may expand, decrease, or remain the same. Therefore, software algorithms have been developed to identify and quantify the extent of the absent DNIRA signal, subject to software for registration, alignment, contrast adjustment, sharpening, and segmentation, without limitation. Figure 10 is a graphical illustration of DNIRA image analysis.

[0231] Subject consent is used for all clinical research protocols that are approved by a Research Ethics Board (REB), and subjects who are willing and able to consent are enrolled on a rolling basis.

[0232] Baseline in vivo imaging: Baseline AMDI images are acquired using a commercially available confocal laser scanning ophthalmoscope (cSLO) (e.g., Heidelberg Retinal Angiography, HRA-2, Heidelberg Engineering, Germany). Images are acquired using cSLO before systemic injection, and images with significant optical opacity are excluded from further imaging (unless clinically indicated). Among subjects with adequate optical clarity, additional images are acquired in red-free, FAF (488 / 500 nm excitation / emission), IR reflectance channel (830 nm), and ICG fluorescence channel (795 / 810 nm excitation / emission), along with optical coherence tomography (OCT) and fundus color photography. Within an appropriate short period, such as between 1 day and 28 days, subjects receive an ICG injection that may or may not include an angiogram.

[0233] ICG injection: ICG dye (Cardiogreen, Sigma) is freshly prepared to a final stock concentration of 5.0 mg / ml in sterile water prior to the experiment. A fine gauge catheter is inserted intravenously and ICG dye is injected. If angiography is performed, images are acquired prior to injection (baseline), during dye circulation, and at various intervals thereafter, usually up to 40 min or 2 h.

[0234] AMD Imaging: Within an appropriate time frame, for example 1 to 5 days (but potentially from 24 hours to 120 days post-ICG), subjects return for DNIRA images obtained using ICG excitation / emission filters or a prescribed (795 / 810 nm) laser system, but without injection of additional dye. Multiple images of each eye are acquired, including images centered on the fovea and optic nerve head. Images are also taken in the four quadrants of the macula, thereby allowing visualization of larger areas and post-mortem analysis of "pseudo-composite" images.

[0235] A logical process can be used to detect sites of widespread low fluorescence / dim DNIRA signal.

[0236] Software and hardware units are available for unprocessed high-resolution DNIRA images to undergo registration, alignment, and segmentation. If the baseline (pre-ICG injection) image has a clear signal before ICG injection, this signal is removed from all other images. When evaluated at a single time point, sites of widespread hypofluorescent / black DNIRA signal are identified and quantified. When evaluated over several points, the images are registered, aligned, and rates of change are identified. DNIRA has the ability to detect black / hypofluorescent signals that can then be analyzed and followed over time.

[0237] Example 2: Low Fluorescent DNIRA Total Area as a Marker of Disease and Disease Burden (Diagnostic and Prognostic Biomarker) FIG. 24 is a graphical representation of this embodiment.

[0238] "Total low-fluorescence DNIRA area(s)" reflects the overall burden of disease at a given time point. Low-fluorescence DNIRA signals are easily segmentable and quantifiable, providing a first and easily tractable measure of disease in addition to GA (determined by FAF, not shown here).

[0239] Subject consent will be used for all clinical research protocols approved by the Research Ethics Board (REB), and subjects who are willing and able to consent will be enrolled on a rolling basis.

[0240] Baseline in vivo imaging: Baseline DNIRA images are acquired using a commercially available confocal laser scanning ophthalmoscope (cSLO) (e.g., Heidelberg Retinal Angiography, HRA-2, Heidelberg Engineering, Germany). Images are acquired using cSLO before systemic injection, and images with significant optical opacity are excluded from further imaging (unless clinically indicated). Among subjects with adequate optical clarity, additional images are acquired in red-free, FAF (488 / 500 nm excitation / emission), IR reflectance channel (830 nm), and ICG fluorescence channel (795 / 810 nm excitation / emission), along with optical coherence tomography (OCT) and fundus color photography. Within an appropriate short period of time, such as between 1 day and 28 days, subjects receive an ICG injection that may or may not include an angiogram.

[0241] ICG injection: ICG dye (Cardiogreen, Sigma) is freshly prepared to a final stock concentration of 5.0 mg / ml in sterile water prior to the experiment. A fine gauge catheter is inserted intravenously and ICG dye is injected. If angiography is performed, images are acquired prior to injection (baseline), during dye circulation, and at various intervals thereafter, usually up to 40 min or 2 h.

[0242] DNIRA Imaging: Within an appropriate time frame, for example 1 to 5 days (but potentially from 24 hours to 120 days post-ICG), subjects return for DNIRA images obtained using ICG excitation / emission filters or a prescribed (795 / 810 nm) laser system, but without injection of additional dye. Multiple images of each eye are acquired, including images centered on the fovea and optic nerve head. Images are also taken in the four quadrants of the macula, thereby allowing visualization of a larger area and post-mortem analysis of "pseudo-composite" images.

[0243] Repeat DNIRA Imaging: To determine temporal changes between DNIRA images over time, subjects return for repeat studies at intervals of, for example, 1 month, 2 months, 3 months, 4 months, 6 months, 8 months, or 12 months, or longer.

[0244] Logical processes can be used for the temporal analysis of areas of expanding low-fluorescent DNIRA signal. Logical processes can also be used for the cross-modality analysis of areas of expanding low-fluorescent DNIRA signal.

[0245] The left pair of images shows a patient with a relatively small hypofluorescent DNIRA total area. The right pair of images shows a patient with a relatively larger hypofluorescent DNIRA total area. By comparing the two, a ratio or percentage can be established. In this case, the patient on the right was identified as more than three times more hypofluorescent using DNIRA. Such total areas can also be measured over time to provide a measure of gradual or dynamic changes.

[0246] Example 3: Low Fluorescent DNIRA Total Area as a Marker of Progressive and Dynamic Changes (Monitoring, Predictive, or Prognostic Biomarker) 17A and 17B are graphic representations of this embodiment.

[0247] "Total low-fluorescence DNIRA area(s)" reflects the overall burden of disease at a given time point. Low-fluorescence DNIRA signals are easily segmentable and quantifiable, providing a first and easily tractable measure of disease in addition to GA (determined by FAF, not shown here).

[0248] Subject consent will be used for all clinical research protocols approved by the Research Ethics Board (REB), and subjects who are willing and able to consent will be enrolled on a rolling basis.

[0249] Baseline in vivo imaging: Baseline DNIRA images are acquired using a commercially available confocal laser scanning ophthalmoscope (cSLO) (e.g., Heidelberg Retinal Angiography, HRA-2, Heidelberg Engineering, Germany). Images are acquired using cSLO before systemic injection, and images with significant optical opacity are excluded from further imaging (unless clinically indicated). Among subjects with adequate optical clarity, additional images are acquired in red-free, FAF (488 / 500 nm excitation / emission), IR reflectance channel (830 nm), and ICG fluorescence channel (795 / 810 nm excitation / emission), along with optical coherence tomography (OCT) and fundus color photography. Within an appropriate short period of time, such as between 1 day and 28 days, subjects receive an ICG injection that may or may not include an angiogram.

[0250] ICG injection: ICG dye (Cardiogreen, Sigma) is freshly prepared to a final stock concentration of 5.0 mg / ml in sterile water prior to the experiment. A fine gauge catheter is inserted intravenously and ICG dye is injected. If angiography is performed, images are acquired prior to injection (baseline), during dye circulation, and at various intervals thereafter, usually up to 40 min or 2 h.

[0251] DNIRA Imaging: Within an appropriate time frame, for example 1 to 5 days (but potentially from 24 hours to 120 days post-ICG), subjects return for DNIRA images obtained using ICG excitation / emission filters or a prescribed (795 / 810 nm) laser system, but without injection of additional dye. Multiple images of each eye are acquired, including images centered on the fovea and optic nerve head. Images are also taken in the four quadrants of the macula, thereby allowing visualization of a larger area and post-mortem analysis of "pseudo-composite" images.

[0252] Repeat DNIRA Imaging: To determine temporal changes between DNIRA images over time, subjects return for repeat studies at intervals of, for example, 1 month, 2 months, 3 months, 4 months, 6 months, 8 months, or 12 months, or longer.

[0253] Logical processes can be used for the temporal analysis of areas of expanding low-fluorescent DNIRA signal. Logical processes can also be used for the cross-modality analysis of areas of expanding low-fluorescent DNIRA signal.

[0254] In FIG. 17A, the "total hypofluorescent DNIRA area" can be monitored over time and repeatedly compared to a specified starting time (baseline). The extensive hypofluorescent DNIRA can be segmented and features extracted and compared over time. Other embedded metadata in the image can also be used to extract other features. An example shows three time points over time where areas of hypofluorescence are segmented (green trace) and compared against each other to detect changes at a given time point.

[0255] In FIG. 17B, the top panel shows a green trace showing the boundaries of the extensively hypofluorescent DNIRA (from the image in the middle panel) areas that may decrease in size over time, unlike the observation of GA (detected using FAF or OCT). These data suggest a much more dynamic and changing aspect of the disease that was previously unrecognized, and show that areas outside of GA or macular atrophy (MA) can be easily quantified. The DNIRA images also reveal complex unsegmentable patterns that represent unstructured image-based data suitable for cognitive analytics or artificial intelligence (AI)-based analytics. Other forms of metadata may also be used to describe the images. The bottom right image shows that dynamic changes can be compared over time to provide a time-dependent baseline. Such data can be used over time for disease prognosis, prediction, patient monitoring, and to help identify treatment strategies.

[0256] Example 4: Rate of change in hypofluorescent DNIRA total area as a marker of prognosis or response to intervention Figure 25 is a graphical representation of this process.

[0257] With subject consent, all clinical diagnostic procedures were performed as described in the examples above.

[0258] Logical processes were used for the temporal analysis of areas of expanding hypofluorescent AMDI signals. Logical processes were also used for the cross-modality analysis of areas of expanding hypofluorescent AMDI signals.

[0259] The leftmost panel shows that this patient had early dry AMD, and FAF was used to confirm the absence of geographic atrophy (GA) throughout the course of the study.

[0260] The center and right panels show DNIRA, with the green trace showing the boundaries of areas of extensive hypofluorescent DNIRA, which may become smaller in size over time, unlike the observations of GA (detected using FAF or OCT). Compared to the patient shown in the previous example (Example 3, Figures 17A and 17B), the rate of change was more rapid, thus allowing comparisons between groups of individuals and potentially predicting a more rapid transition to late AMD.

[0261] These features may be particularly useful in identifying not only the prognosis of disease progression but also responses to interventions that occur in this time window.

[0262] Example 5: DNIRA identifies sites of widespread hypofluorescent / black signals that extend beyond sites of hypofluorescent / black FAF - Calculation of "Delta" as a measure of disease burden FIG. 14 is a graphical illustration of this concept.

[0263] DNIRA can identify areas of widespread hypofluorescent / black signal beyond that observed using FAF, the current gold standard for the evaluation of subjects with late dry AMD. FAF identifies areas of RPE / photoreceptor loss known as GA (in the absence of endogenous fluorophores such as lipofuscin). When present, areas of GA are always within areas of absent DNIRA signal. Thus, logic and software algorithms have been developed to identify and quantify the extent of absent DNIRA signal beyond the absent FAF signal. This may represent a measure of disease burden rather than areas of tissue loss. Figures 15A and 15B are graphical illustrations of this example.

[0264] Subject consent: All clinical study protocols are Research Ethics Board (REB) approved, and subjects who are willing and able to consent will be enrolled on a rolling basis.

[0265] Baseline in vivo imaging: Baseline DNIRA images are acquired using a commercially available confocal laser scanning ophthalmoscope (cSLO) (e.g., Heidelberg Retinal Angiography, HRA-2, Heidelberg Engineering, Germany). Images are acquired using cSLO before systemic injection, and images with significant optical opacity are excluded from further imaging (unless clinically indicated). Among subjects with adequate optical clarity, additional images are acquired in red-free, FAF (488 / 500 nm excitation / emission), IR reflectance channel (830 nm), and ICG fluorescence channel (795 / 810 nm excitation / emission), along with optical coherence tomography (OCT) and fundus color photography. Within an appropriate short period of time, such as between 1 day and 28 days, subjects receive an ICG injection that may or may not include an angiogram.

[0266] ICG injection: ICG dye (Cardiogreen, Sigma) is freshly prepared to a final stock concentration of 5.0 mg / ml in sterile water prior to the experiment. A fine gauge catheter is inserted intravenously and ICG dye is injected. If angiography is performed, images are acquired prior to injection (baseline), during dye circulation, and at various intervals thereafter, usually up to 40 min or 2 h.

[0267] DNIRA Imaging: Within an appropriate time frame, for example 1 to 5 days (but potentially from 24 hours to 120 days post-ICG), subjects return for DNIRA images obtained using ICG excitation / emission filters or a prescribed (795 / 810 nm) laser system, but without injection of additional dye. Multiple images of each eye are acquired, including images centered on the fovea and optic nerve head. Images are also taken in the four quadrants of the macula, thereby allowing visualization of a larger area and post-mortem analysis of "pseudo-composite" images.

[0268] A logical process was used for the elimination of background signals: in some cases where autofluorescence signals are obtained in baseline DNIRA images (e.g., before dye injection), such baseline signals are mathematically eliminated before further image analysis.

[0269] A logical process was used to calculate the "delta", the difference between the low fluorescence DNIRA image and the FAF image. When GA is present, DNIRA uses FAF to identify sites of extensive low fluorescence / black signal that are equal to or greater than the area of ​​extensive low fluorescence / black signal obtained.

[0270] A software and hardware unit for unprocessed high-resolution DNIRA and FAF images undergoing registration, alignment, and segmentation. If the baseline (pre-ICG injection) image reveals a clear signal prior to ICG injection, this signal is removed from all other images. Registered DNIRA and FAF images assessed at a single time point are compared, and sites of hypofluorescent DNIRA that extend beyond the boundaries of the dark FAF or are in different sites from the dark FAF are identified (both in separate images and as part of a two-layer overlay) and quantified (following segmentation). In this example, DNIRA identifies multiple sites outside of the FAF hypofluorescence that contain sites of black DNIRA signal, increasing the total number of black hypofluorescent signals, the total area of ​​the black signals, and the perimeter or edge of the black sites.

[0271] Example 6: DNIRA signatures (diagnostic biomarkers) to identify early stage disease and distinct phenotypes of early stage disease 26A and 26B are graphic representations of this embodiment.

[0272] DNIRA can distinguish different phenotypes of early AMD. These may represent fundamental biology (genetic or environmental) differences that are potentially modifiable with different and targeted therapies. Although AMD has a strong genetic component, AMD accounts for only about half of disease susceptibility at best. Among patients with a genetic background, the statistical probability that they will or will not develop the disease can be calculated for the population, but not for individuals. In contrast, early changes in DNIRA can confirm positive genetic traits, e.g., very early disease, even before drusen formation.

[0273] With subject consent, all clinical diagnostic procedures were performed as described in the previous example. DNIRA imaging procedures were performed on patients who had no personal or known family history at the time of presentation. However, a positive family history was obtained at a later time during the clinical visit upon questioning.

[0274] In FIG. 26A, the left panel shows a patient with a family history of AMD but no personal diagnosis (absence of drusen confirmed). The DNIRA image shows a faint but complex hypofluorescent pattern. The right panel shows a decade-matched control patient with no family or personal history of AMD and no found drusen. Absence of AMD (absence of drusen) using the current clinical definition was confirmed in a clinical trial. This illustrates the ability of DNIRA to detect subtle changes in patients who may not otherwise present with any identifiable features of the disease.

[0275] In FIG. 26B, the top panel shows two patients with a diagnosis of early AMD based on clinical testing. On the left is a DNIRA image of a patient with early AMD showing a pronounced dark gray gradual pattern of markings. On the right is a DNIRA image showing a subtle gradual pattern of markings. The corresponding FAF images confirm that these two patients have early dry AMD, but show relatively little difference with this modality. The DNIRA and FAF images were obtained at the same time points.

[0276] Thus, the DNIRA patterns observed in early stage AMD may aid in personalized diagnosis of AMD earlier than other methods, allowing for unique and targeted therapy.

[0277] Example 7: DNIRA signatures (predictive biomarkers) to distinguish disease subtypes among patients with a known diagnosis of AMD and correlate with disease etiology 21A through 21E are graphic representations of this embodiment.

[0278] It is known that blinding eye diseases such as AMD are complex and arise from multigenic inheritance (more than 30 genes and 200 single nucleotide polymorphisms) that are subject to significant environmental additives, making successful treatment difficult. Novel groupings of diseases and disease features correlate with complex inherited and epigenetic influences.

[0279] DNIRA and related methods can identify features not seen before the disease and complex new phenotypes. For example, several complex two-dimensional patterns have been identified that have been labeled as "tight patterns," "loose patterns," "gray stains," "oil stains," "bullet holes," etc. Thus, DNIRA can be used to define new phenotypes and, subsequently, novel genotype-phenotype correlations.

[0280] With subject consent, all clinical diagnostic procedures were performed as described in previous examples.

[0281] Figures 21A-21E show examples of DNIRA uniquely identifying novel image-based phenotypes, exemplified by currently identified phenotypes named Figure 21A, "Loose Pattern" pattern, Figure 21B, "Leopard Print", Figure 21C, "Gray Stain", Figure 21D, "Fingering Potato", and Figure 21F, "Very Large Bright Spot".

[0282] These kinds of complex phenotypes were previously unknown and have never been identified using any other imaging modality, demonstrating the unique ability of DNIRA to capture them.

[0283] Example 8: DNIRA signatures (diagnostic biomarkers) to differentiate diseases with similar clinical presentations and aid in diagnostic accuracy 27A and 27B are graphic representations of this embodiment.

[0284] Complex DNIRA patterns observed in association with other disorders, without a personal history of AMD, may provide an early diagnostic biomarker. For example, early vitelliform disease, especially adult-onset (AVD), is characterized by foveal involvement that is often mistaken for AMD. In some cases, the diagnosis is made following non-response to anti-VEGF agents (prescribed for late neovascular "wet" AMD). However, as a potential diagnostic biomarker, DNIRA identifies complex two-dimensional patterns of hypo / hyperfluorescence that are not observed in patients with AVD. Another non-AMD disease that may be confused with AMD is central serous retinopathy (CSR).

[0285] In the cases described here, with the subject's consent, all clinical diagnostic procedures were performed as described in the previous examples.

[0286] In FIG. 27A, the top panel shows an image of the right eye of a patient with a provisional diagnosis of AMD because of vision complaints and the presence of a small yellowish material in the fovea (color image not shown). IR and FAF are mediocre, while DNIRA shows a faint variable appearance. The bottom panel shows an image of this patient's left eye, confirming the diagnosis of AVD with the obvious appearance of a "fluid line" across the central lesion. The bottom center panel shows a DNIRA image with hypofluorescence in the area of ​​the central lesion and almost no background contrast (as noted in the fellow eye). This is in stark contrast to the DNIRA image obtained from a patient with early AMD (bottom left), where a streaked pattern is evident.

[0287] In Figure 27B, a non-AMD patient with CSR in the left eye showed a complex "loose pattern" pattern in both eyes, despite no family or personal history of AMD and no drusen in either eye. During the study, drusen were detected in the non-CSR eye (not assessable in the affected eye).

[0288] The top panel in B shows the right eye to be normal using infrared and FAF, but DNIRA shows a "loose pattern" pattern. The middle panel shows the left eye with both CSR features and patterns. The bottom panel shows the eye exam also confirming normal color, but OCT confirmed the onset of drusen at the follow-up evaluation when the patent completed the study (approximately 2 years after the initial exam). The first OCT image taken in 2015 demonstrates the absence of drusen in a patient with no personal or family history of AMD, while the lower image of the same patient in 2017 demonstrates the onset of classic drusen, confirming the diagnosis of AMD, which was inferred in 2015 due to the presence of the "pattern" pattern.

[0289] This illustrates that DNIRA reveals several complex image-based patterns that are absent in patients with ocular conditions other than AMD or that are frequently misdiagnosed as AMD, such as, but not limited to, adult vitelliform disease, Bietti crystalline corneal retinal dystrophy, pattern dystrophies, and others.

[0290] Thus, DNIRA can be used to identify patients with disorders that clinically mimic AMD and may be used as a diagnostic biomarker to detect early ocular disease, allowing for early therapeutic, lifestyle, or environmental intervention.

[0291] Example 9: DNIRA signature identifies inherited single gene disorders and not yet diagnosed (NYD) diseases 28A and 28B are graphic representations of this embodiment.

[0292] Complex DNIRA patterns may provide early diagnostic biomarkers for inherited monogenic diseases such as Stargardt's disease, retinitis pigmentosa, Leber's congenital amaurosis (LCA), or Bardet-Biedl syndrome (BBS).

[0293] In this example, patients with an inherited monogenic disease (Stargardt disease) demonstrate the ability of DNIRA to detect disease and detect potential risk areas for disease progression from early to late stages in these types of diseases.

[0294] With subject consent, all clinical diagnostic procedures were performed as described in previous examples.

[0295] In Figure 28A (left panel), the left and center images show IR and FAF images of the left fundus of a patient with Stargardt's disease, where bright hyperfluorescent lesions with a characteristic "fish-shaped" shape can be seen. The right image shows that widespread areas of hypofluorescence can be seen that are segmentable and presumably represent areas of poor dye uptake or accumulation and / or potentially reduced choroidal flow. These differ from methods used in the assessment of Stargardt's disease, which utilizes near-infrared autofluorescence (NIR-AF), i.e., NIR light but without the early delivery of NIR dye (and thus detects autofluorescence from pigments such as melanin).

[0296] The magnified yellow box (right-most image) shows DNIRA and identifies small hyperfluorescent bright dots in association with areas of hypofluorescent DNIRA. These dots may represent actively phagocytic macrophages that ultimately lead to macular atrophy (MA) in this disease. Throughout the image, they are found specifically in association with sites of hypofluorescent DNIRA signal, potentially attributing macrophage activity to disease progression, thus linking DNIRA to potential therapeutic strategies (e.g., modulation or inhibition of macrophages).

[0297] In FIG. 26B, DNIRA images show a putative inherited eye disease that shares some aspects of genetic susceptibility with AMD, but is not yet diagnosed (NYD).

[0298] Thus, DNIRA can be used to diagnose patients with disorders other than AMD and may be used as a diagnostic biomarker to detect early eye disease and enable early therapeutic, lifestyle, or environmental intervention.

[0299] Example 10: DNIRA signatures for monitoring disease progression or efficacy of treatment over time (monitoring biomarkers) FIG. 29 is a graphical representation of this embodiment.

[0300] With subject consent, all clinical diagnostic procedures were performed as described in previous examples.

[0301] Hypofluorescent features identified with DNIRA can be observed to expand over time, similar to observations made using en face OCT and more recently using fundus autoluminescence (FAF) to quantify the rate of expansion of geographic atrophy (GA).

[0302] Currently, in the absence of other useful markers to drive clinical trial design for potential treatments for dry AMD, the rate of GA expansion is an acceptable Phase III clinical trial endpoint. An example of this is shown in the top panel, which shows how FAF, when followed over time, identifies sites of hypofluorescent signal that are indicative of GA expansion. The bottom panel shows that DNIRA identifies a significantly more hypofluorescent signal that changes over time and identifies more disease.

[0303] A faint purple line has been used to trace the periphery of the abnormal hypofluorescent FAF in the top panel, and a faint green line has been used to do the same in the lower DNIRA panel. These extracted image elements are shown in the bottom two panels below the fundus image.

[0304] Thus, DNIRA can be used to monitor disease progression and in many cases provides more information than current methods such as FAF.

[0305] Example 11: DNIRA signatures (monitoring and prognostic biomarkers) for monitoring disease progression and comparing with other imaging biomarkers over time within and across modalities FIG. 30 is a graphical representation of this embodiment.

[0306] With subject consent, all clinical diagnostic procedures were performed as described in previous examples.

[0307] Shown in the figure are cases observed to date where DNIRA was compared to other imaging modalities across time points to identify features that may serve as monitoring and prognostic biomarkers.

[0308] FAF and DNIRA features have been traced using purple and green lines, respectively, and used to trace the periphery of the abnormal hypofluorescent FAF or DNIRA signal. These extracted image elements are shown alongside the fundus image.

[0309] This patient had central serous retinopathy in the left eye and no drusen in either eye at the first visit. However, a striking complex two-dimensional "pattern" was observed in both eyes. During the study, drusen were detected in the non-CSR eye (not assessable in the affected eye). The leftmost panel shows the extensive area of ​​hypofluorescent DNIRA observed at baseline (DNIRA S1) in this patient.

[0310] The two top panel images on the right show baseline and 8-month follow-up FAF images.

[0311] The bottom panels show registered and segmented DNIRA (green outline) and FAF (purple outline) images taken over time, illustrating the progression of GA towards areas of low DNIRA signal.

[0312] When compared across time points, DNIRA session 1 (S1) was observed to detect features that were only evident in FAF session 3 (S3), indicating that DNIRA can be used to monitor disease progression.

[0313] The spread of GA to hypofluorescent sites using DNIRA has been observed in all patients to date. In contrast, not all areas of hypofluorescent DNIRA become areas of GA (at least during this study).

[0314] Thus, GA expands only into regions of pre-existing low-fluorescence DNIRA. In this way, several regions of low-fluorescence DNIRA signal can be used to predict sites of GA expansion.

[0315] Example 12: DNIRA signature as a biomarker to identify patients likely to progress to late stage disease (prognostic biomarker) FIG. 31 is a graphical representation of this embodiment.

[0316] With subject consent, all clinical diagnostic procedures were performed as described in previous examples.

[0317] Shown in the figure is where DNIRA was used and compared to other imaging modalities across time points to identify specific features and complex patterns observed that may serve as prognostic biomarkers that may predict disease progression.

[0318] In some cases, these features may not have been previously described, while in other cases, the features may correlate with previously known risk factors for disease progression, such as the presence of large RPE detachments.

[0319] Illustrated is a patient with early dry AMD. The left panel shows early AMD using FAF without GA as indicated by the absence of areas of abnormal widespread hypofluorescence. The right panel shows the same area imaged at the same time point as DNIRA, where DNIRA identifies multiple features of the disease, including both widespread quantifiable areas of hypofluorescence and complex three-dimensional patterns.

[0320] The center panel shows a second case in which heatmap analysis performed at three time points (4 months apart) shows that the corresponding FAF image (red) is smaller when compared against the DNIRA image (green) in session 1 (S1, left) but progressively "fills in" the site identified as vulnerable using DNIRA. By 8 months, the site of GA identified using FAF expands beyond the site originally predicted by DNIRA.

[0321] The bottom panel confirms that (8 months later) simultaneous FAF and DNIRA imaging confirms that the DNIRA site also expanded over time and that the site of GA identified using FAF is contained within its boundary (the apparent expansion of the FAF signal, inferior to the merged signal, reflects the "fuzzy" boundary of the area of ​​GA obtained using FAF).

[0322] Therefore, the DNIRA signal can be used to predict sites of GA expansion.

[0323] Example 13: DNIRA signatures as biomarkers to quantify known disease aspects and predict progression (prognostic biomarkers) FIG. 32 is a graphical representation of this embodiment.

[0324] DNIRA can easily identify, quantify and follow over time spatiotemporal changes in disease features that are predicted to correlate with disease but are currently not quantifiable and therefore not explicit. For example, drusen have been identified as an important prognostic factor predicting the 10-year risk of a patient progressing to blinding late AMD (both wet and dry). The clinical utility of large and soft and soft confluent drusen precludes their application in prevention studies due to the large sample sizes required to adequately facilitate the studies. Large and medium drusen confer disease risk and progression risk, but their dynamic changes over time have only recently been described. As they are not easily quantifiable, it is not currently possible to provide a measure of their changes.

[0325] With subject consent, all clinical diagnostic procedures were performed as described in previous examples.

[0326] The figure shows in the top panel large soft and confluent soft drusen identified as yellowish round / oval deposits observed on color fundus photography (CFP). The bottom panel correlates the same area on an image using DNIRA, clearly showing the black defined area where the drusen is localized.

[0327] Thus, DNIRA can readily identify and quantify known high-risk prognostic features of the disease, such as large soft and confluent drusen.

[0328] Therewith, we describe the Dynamic Drusen Index (DDI), which can calculate the total drusen burden at a single or multiple time points, or calculate the change in a specific drusen subtype. Relying on the current classification of drusen and the utility of the novel DNIRA images to identify drusen, we propose the Total DDI (for all drusen types), the Confluent Dynamic Large Drusen Index (LC-DDI), the Dynamic Large Drusen Index (L-DDI), the Dynamic Medium Drusen Index (M-DDI), and the Dynamic Small Drusen Index (S-DDI). The DDI can direct which patients should be prioritized for inclusion in a particular clinical trial, thereby providing a predictive biomarker.

[0329] Example 14: Using DNIRA to correlate with disease pathogenesis (predictive biomarker) FIG. 33 is a graphical representation of this embodiment.

[0330] Drusen have been identified as an important prognostic factor predicting the 10-year risk of a patient progressing to blinding late AMD (both wet and dry). The clinical usefulness of large soft and soft confluent drusen precludes their application in prevention studies due to the large sample sizes required to adequately power the studies.

[0331] With subject consent, all clinical diagnostic procedures were performed as described in the examples above.

[0332] To correlate DNIRA in patients with a valid model representing macrophages in the eye, retinal atrophy was induced in rabbits using systemic injection of oxidizing agents and imaged using the DNIRA method. DNIRA images were acquired using a commercially available confocal laser scanning ophthalmoscope (cSLO) (e.g., Heidelberg Retinal Angiography, HRA-2, Heidelberg Engineering, Germany). Images were obtained using cSLO in red-free, FAF (488 / 500 nm excitation / emission), IR reflectance channel (830 nm), and ICG fluorescence channel (795 / 810 nm excitation / emission) before systemic injection. ICG dye (cardiogreen, Sigma) was freshly prepared to a final stock concentration of 5.0 mg / ml in sterile water prior to the experiment. A fine gauge catheter was inserted intravenously into the marginal ear vein and ICG dye was injected. Within 5-14 days after toxin injection and 2-3 days after dye injection, DNIRA images are obtained using ICG excitation / emission filters or a prescribed laser system (795 / 810 nm), but without additional dye injection. Multiple images of each eye, including central and peripheral, are acquired.

[0333] The image on the left shows a DNIRA image of the patient demonstrating a ring of hyperfluorescent dots consistent with putative macrophages located just beyond the area of ​​hypofluorescence.

[0334] The center image shows that DNIRA images of rabbit eyes after acute oxidative stress reveal similar rows of bright hyperfluorescent dots surrounding areas of hypofluorescence.

[0335] After imaging, the rabbit eyes are removed for immunofluorescence histology using a marker for macrophages, such as Iba-1.

[0336] The right panel shows immunofluorescently labeled rabbit retinal tissue (blue nuclear staining) with the presence of Iba-1+ macrophages (red) surrounding sites of outer retinal damage and RPE loss.

[0337] Additionally, the bottom panel confirms the distribution of Iba-1+ macrophages surrounding small newly formed areas of outer retinal damage (indicated by the ring and convolution of the outer nuclear layer) and RPE loss (indicated by sparse staining of the RPE marker RPE65 in green) resembling geographic atrophy.

[0338] In this manner, DNIRA can identify putative macrophages in the eyes of patients with a disease, e.g., AMD or an ocular tumor, and identify individuals who are more likely to experience favorable or unfavorable effects from exposure to a pharmaceutical or environmental agent, e.g., a macrophage modulating therapy, than similar individuals without the use of the biomarker.

[0339] Example 15: Analysis of Tumor-Associated Macrophages (TAM) Using DNIRA (TAMI) as a Predictor of Disease Severity in Ocular Melanoma Figures 18 and 19A-19J are graphic representations of this embodiment.

[0340] With subject consent, all clinical diagnostic procedures were performed as described in the examples above.

[0341] Multimodal and cross-modal analysis for identification of regions of interest (ROI) containing hyperfluorescent TAMI signals. A logical process for the temporal analysis of discrete, dot, or speckled hyperfluorescent TAMI signals. A logical process for the temporal analysis of clusters or accumulations of dot or speckled TAMI signals. A logical process for the cross-modality analysis of hyperfluorescent dots or speckles.

[0342] FIG. 18 shows a comparison of DNIRA versus color photography and clinical grading of an example of uveal melanoma. The two left images show a uveal nevus, the center right image shows an indeterminate lesion, and the right image shows a melanoma image. In the top panel, the DNIRA images show the absence or presence of bright hyperfluorescent dots within the tumor area or areas of associated fluid. These images were used to perform the analysis described in FIG. 19A through FIG.

[0343] Figures 19A to 19G show TAMI composite images for correlation of TAMI imaging with currently available clinical imaging modalities. The collection of TAMi composite images allows correlation of the novel imaging method with current clinical modalities. Fundus autofluorescence (FAF, Figure 19B) and TAMi composite images (Figure 19C) were collected as overlays on a color photograph (Figure 19A). Multimodal analysis allowed identification of regions of interest including tumor site (i), areas of current or former fluid (ii), and surrounding (additional damaging) sites (iii). Figure 19D shows multimodal analysis allowing identification of tumor site. Figure 19E shows multimodal analysis allowing identification of fluid regions. Figure 19F shows identified regions of interest including tumor site (i), areas of current or former fluid (ii), and surrounding (additional damaging) features (iii). Figure 19G shows identification of regions of interest from the composite images. mm 2 Areas of the lesions at 100 nm were determined using ImageJ measurements and used in the hyperfluorescent dot density calculation (Figure 19H). These were compared across patients with melanoma, indeterminate lesions, or benign nevi, as determined by clinician ground truth.

[0344] Comparisons were made to subject risk characteristics including the presence of orange pigment, lesion area on color fundus photography, lesion thickness, lesion elevation, proximity to the optic nerve head, presence of subretinal fluid, presence of hot spots, and degree of pigmentation. Risk characteristics were analyzed in subjects with benign nevi, ocular melanoma, and control subjects and compared to hyperfluorescent dot counts observed on TAMI / DNIRA imaging. Mean site dot density was calculated as dots / mm 2 Shown as ±SEM.

[0345] (Figure 19I) One-way ANOVA and post hoc multiple comparisons show significantly higher dot density in melanomas when considering lesion and fluid areas, but not within the periphery. (Figure 19J) Multiple comparisons of dot density at each site within the risk groups show that melanomas had significantly higher dot density in lesion and fluid areas when compared to the periphery. This was not observed in the other risk groups.

[0346] The presence of hyperfluorescent dots correlated with the presence of ocular melanoma but not benign nevi, i.e., in control subjects, suggesting that the presence of hyperfluorescent dots can predict the severity of ocular tumors.

[0347] Example 16: DNIRA (Predictive Biomarkers) for detecting multiple types of hyperfluorescent dot signals Figures 34A through 34D are graphic representations of this embodiment.

[0348] DNIRA has the potential to identify patients with expanded intraocular macrophage presence, serve as a predictive biomarker, and provide a unique link between imaging biomarkers and disease pathogenesis. Analysis of these features has previously been correlated to preclinical studies, as confirmatory ocular biopsies from human subjects are often not possible.

[0349] With subject consent, all clinical diagnostic procedures were performed as described in the examples above.

[0350] Repeated analyses exemplified here from just two patients show a size distribution of hyperfluorescent dots that were designated as macrophages in correlative preclinical studies.

[0351] In Figure 34A, the left panel shows small hyperfluorescent dots that can be seen in some patients using DNIRA. These have a high degree of circularity and are approximately the same size. The middle panel shows another type of bright dot with a dark center. The right panel shows a very large bright dot that is observed in some forms of the disease.

[0352] In FIG. 34B, Kruskal-Wallis analysis shows that the size and uniformity of the dots is consistent with their cellular identity. The distribution of dot sizes is quite similar - both qualitatively (histograms) and quantitatively (Kruskal-Wallis p-value of 0.7577 indicates no significant difference). With a mean diameter of 22.5 μM, the mean size of pt05 is 399 μm. 2 with an average diameter of 24.8 μM, and the average size of pt010 is 483 μm 2 Macrophages in vivo are predicted to range from 15 to 40 µm in diameter, depending on their activity status and other influences, and these dots are consistent with macrophage size.

[0353] In Figure 34C, the distribution of dot circularity - both qualitatively (histogram) and quantitatively (Kruskal-Wallis p-value of 0.1055 indicates no significant difference) - is similar across patients. The mean circularity of pt005 is 0.926. The mean circularity of pt010 is 0.888. These values ​​are consistent with measurements from preclinical studies of rabbit eyes.

[0354] Finally, in Figure 34D, we demonstrate that in preclinical studies using macrophage modulators [such as, but not limited to, bindarit and its derivatives, methotrexate, and others], the number, size, and location of macrophages are modified in treated animals compared to untreated controls, suggesting the utility of DNIRA for monitoring drug response as well as for identifying patients for clinical trials. Thus, DNIRA is a useful pharmacodynamic marker that demonstrates a significant signal reduction following macrophage modulating therapy in preclinical studies in rabbit eyes. It is expected that a similar reduction in hyperfluorescent signal may also be observed in patients receiving macrophage and immunomodulatory agents.

[0355] Example 17: DNIRA (Predictive Biomarker) to detect static and dynamic hyperfluorescent dot signals FIG. 35 is a graphical representation of this embodiment.

[0356] DNIRA images show hyperfluorescent (bright) dots that are transient or transient in nature. It is not known whether these dots move, disappear and reappear, or whether new dots have arrived.

[0357] With subject consent, all clinical diagnostic procedures were performed as described in the examples above.

[0358] The top panel on the left shows a pair of images comparing DNIRA to FAF. DNIRA shows additional areas of hyperfluorescent signal plus bright hyperfluorescent dots found in association with these areas. In the pair of images on the right, the same set obtained at least 3 months later shows subtle changes.

[0359] The center panel shows two DNIRA fundus images of a patient with Stargardt's disease taken at least 3 months apart. Enlargements of these images show bright dots present at both time points (blue arrows). At other sites, bright dots appeared at the second visit that were not noticed at the first visit (yellow arrows). The blue outlined arrows in the right image indicate sites where dots that were previously there have moved, disappeared, or potentially become less phagocytic and therefore no longer take up dye.

[0360] Example 18: DNIRA detects sites of tissue damage and macrophage activity in central serous retinopathy (predictive biomarker, prognostic biomarker) FIG. 36 is a graphical representation of this embodiment.

[0361] Central serous retinopathy (CSR) is a disease characterized by the shift of fluid from the choroid into the subretinal and intraretinal spaces. Over time, the poorly broken down fluid is associated with irreversible chorioretinal atrophy (MA).

[0362] With subject consent, all clinical diagnostic procedures were performed as described in the examples above.

[0363] Images of a patient with CSR in the top panels show, from left to right, IR, FAF, and DNIRA images, respectively. FAF shows hyperfluorescence within areas of previous (and potentially current) fluid. DNIRA similarly identifies areas associated with fluid and areas with reduced functional ability to take up dye, which appear black (hypofluorescent). These areas suggest vulnerability to future tissue loss (using standard methods of FAF, not yet evident as areas of GA).

[0364] In the bottom panel, the left image taken in the ICG channel before systemic dye injection is devoid of signal, while the middle and right images show that hyperfluorescent dots are found spatiotemporally within areas of CSR-associated fluid and tissue damage.

[0365] These data suggest that macrophages contribute to tissue injury and that both macrophages and tissue injury can be detected using DNIRA, allowing for early detection and prognosis.

[0366] Example 19: DNIRA detects two populations of bright hyperfluorescent dots in diffuse trich AMD (diagnostic biomarker, prognostic biomarker) FIG. 37 is a graphical representation of this embodiment.

[0367] The diffuse trickling form of late dry AMD is usually defined by a unique FAF pattern. Here, we identify novel biomarkers for the diffuse trickling subtype of AMD.

[0368] With subject consent, all clinical diagnostic procedures were performed as described in the examples above.

[0369] In the top panel, the left and center images show IR and FAF, respectively, identifying extensive areas of GA, while the right image shows DNIRA with areas of hypofluorescence roughly the same area as FAF, but also with enlarged boundaries of hyperfluorescent dots.

[0370] The bottom panel shows NIR imaging before systemic delivery of ICG dye on the left, which shows no detectable signal. The middle image shows the same DNIRA image as above and a magnification of the particular area. On the right, DNIRA reveals two populations of hyperfluorescent (bright) dots at the border and in the surrounding so-called "junctional zone." Some are of a size typical of putative macrophages (as seen, for example, in the context of tumors), while others are much larger, potentially pointing to accumulation of dye in RPE or other cells.

[0371] Example 20: DNIRA to detect reticular pseudodrusen FIG. 38 is a graphical representation of this embodiment.

[0372] Reticular pseudodrusen (RPD) are a particularly aggressive subtype of dry AMD and are associated with a faster rate of disease progression.

[0373] With subject consent, all clinical diagnostic procedures were performed as described in the examples above.

[0374] DNIRA images of the peripheral part of the fundus and subsequent logic processing are used to show the presence of discrete, dot-like, or speckled hyperfluorescent signals, which can be used for temporal analysis of the aggregation or accumulation of dot-like or speckled DNIRA signals (indicated by arrows) and for intermodality analysis of the hyperfluorescent dots or speckles.

[0375] This signal is then used to correlate with reticular pseudodrusen (RPD) burden and thus disease burden, allowing for a potentially better predictive biomarker of disease progression.

[0376] Example 21: DNIRA for detecting inflammatory eye disease FIG. 39 is a graphical representation of this embodiment.

[0377] DNIRA can be applied to the analysis of inflammatory eye diseases. In this example of presumed acute multifocal posterior plaque-like epitheliopathy (AMPPE), DNIRA reveals abnormal pigment uptake in the RPE / outer retina.

[0378] With subject consent, all clinical diagnostic procedures were performed as described in the examples above.

[0379] In the left panel, IR imaging shows variable reflectivity, including areas of hyper- and hypo-reflectivity, as well as punctate signal. In the center panel, FAF similarly reveals areas of hyper- and hypofluorescence, along with areas of hyperfluorescence that are presumed to correlate with areas of potential disease activity. In the right panel, DNIRA imaging shows multiple areas of widespread hypofluorescence that likely correlate with areas of diseased tissue that fail to accumulate dye or potentially block the signal. Additionally, as shown in the box, a small area of ​​hypofluorescence forms (oval area) that is only present in the DNIRA image with little change in other channels. Within this area are two large hyperfluorescent dots. Based on preclinical studies, these dots correlate with phagocytic macrophages and are of a size and shape consistent with dots seen in other conditions, such as ocular melanoma.

[0380] Thus, DNIRA delineates areas of damage greater than either FAF or IR imaging in inflammatory eye diseases and can be used as a diagnostic or prognostic biomarker.

[0381] Example 22: DNIRA as a marker for high-risk AMD characteristics (predictive biomarker, prognostic biomarker) FIG. 40 is a graphical representation of this embodiment.

[0382] DNIRA can be used, for the first time, to quantify and better characterize image-based features of diseases such as AMD, some of which are already known to confer a high risk of progression from early to late disease (either atrophic or angiographic). Some of these features may have a correlation with histological features, such as basement membrane deposits, basal linear deposits, or deposits of extracellular, immune, and poorly defined material that are deposited within the subretinal or sub-RPE space. Furthermore, some deposits, such as hyperreflective material (within the subretinal space, it is known as subretinal hyperreflective material SRHM), are associated with a high risk. However, many other materials are observed using OCT, which are not hyperreflective and therefore not seen en face and are not clearly defined. It is therefore not possible to quantify or compare these features over time.

[0383] With subject consent, all clinical diagnostic procedures were performed as described in the examples above.

[0384] In the top row, the left image shows an infrared image where a variable signal can be observed along with hyperreflective material that is known to fluoresce in the IR spectrum and may correlate with the accumulation of pigments such as melamine. The top right image shows that the FAF confirms that this patient has early AMD without obvious GA, but where hyperreflective material is also present.

[0385] The center row shows DNIRA images of the sample before dye injection, where no signal is visible. The right image shows a DNIRA obtained in the same session as the FAF and IR images in the top panel, showing widespread areas of hypofluorescence. In addition, small punctate hyper- and hypofluorescence are also observed, the former of which may correlate with dye-labeled macrophages or other phagocytes capable of ingesting the dye delivered systemically.

[0386] The bottom row shows an OCT image in which shallow drusen, subretinal material, and shallow sub-RPE clefts are observed associated with widespread hypofluorescent DNIRA signal.

[0387] Thus, DNIRA has the ability to highlight features of early AMD that may correlate with tissue elements associated with early disease and with high risk of disease progression, and serves as a predictive or prognostic biomarker.

[0388] equivalent While the disclosure has been described in relation to exemplary embodiments thereof, it will be understood that it is susceptible of additional modifications, and that this application is intended to cover any variations, uses, or adaptations of the invention which generally follow the principles of the invention, including such departures from the disclosure as come within known or customary practice within the art to which the invention pertains, and as may be applied to the basic features hereinbefore described and which follow within the scope of the appended claims.

[0389] Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific embodiments specifically described herein which equivalents are intended to be encompassed by the scope of the following claims.

[0390] Incorporation by Reference All patents and publications referenced herein are hereby incorporated by reference in their entirety.

[0391] The publications discussed herein are provided solely for their disclosure prior to the filing date of the present application, and nothing herein should be construed as an admission that the present invention is not entitled to antedate such publication by virtue of prior invention.

[0392] As used herein, all headings are for organizational purposes only and are not intended to limit the disclosure in any way. The content of any individual section may be equally applicable to all sections.

Claims

[Claim 1] The invention as depicted in the drawings of this application.

Images