A quantitative system for the reflectance of pigment epithelial detachment (PED) and subretinal high reflectance material (SHRM).

Abstract

This invention provides a system and method for quantifying biomarkers associated with ophthalmic features identified via OCT imaging. A set of optical coherence tomography (OCT) volumes is received, each OCT volume corresponding to a different time point in a set of time points, and each OCT volume includes a set of OCT B scans. A set of feature images is generated for each of the OCT volumes in the set of OCT volumes, the set of feature images visually identifying the ophthalmic features using ophthalmic feature markers. The ophthalmic feature markers assign different groups of pixels to each of the plurality of ophthalmic features. The ophthalmic features include target retinal features, the retinal pigment epithelium (RPE) layer, and the vitreous body. A reflectance score for the target retinal feature identified in each OCT volume is calculated using the ophthalmic features identified by the ophthalmic feature markers in each OCT volume, thereby forming a set of reflectance scores.

Description

[0001] Inventor: Isabel Wilma BACHMEIER, Carl-Gustav Olsen GLITTENBERG, Ian LloydJONES, Sayedali SHETAB BOUSHEHRI, Siqing YU, Mahnaz AMIRI PARIAN, AndreasMAUNZ Cross-references to related applications

[0002] This application relates to and claims the benefit of U.S. Provisional Application 63 / 566,349, filed March 17, 2024, entitled “Pigment Epithelial Detachment (PED) and Subretinal Hyperreflective Material (SHRM) Reflectivity Distribution and Dynamics,” which is incorporated herein by reference in its entirety. Technical Field

[0003] This application relates to the quantification of biomarkers using optical coherence tomography (OCT) images, and more specifically to the quantification (e.g., one or more reflectance scores) of one or more biomarkers (e.g., pigment epithelial detachment (PED) and / or subretinal high reflectance material (SHRM)) identified using OCT images of the retina for the purpose of assessing changes in retinal conditions (e.g., neovascular age-related macular degeneration (nAMD)) over time (e.g., treatment response / treatment outcome). Background Technology

[0004] Age-related macular degeneration (AMD) is a leading cause of vision loss in subjects aged 50 and older. AMD initially presents as dry AMD and can progress to wet AMD. In dry AMD, small deposits (drusen) form under the macula of the retina, leading to progressive retinal degeneration. In wet AMD, also known as neovascular AMD (nAMD), abnormal blood vessels originating from the choroid of the eye grow into the retina and leak fluid from the bloodstream into the retina. Once inside the retina, the fluid can immediately distort the subject's vision and, over time, may damage the retina itself, for example, by causing loss of photoreceptors in the retina. The fluid can cause the macula to separate from its base, resulting in severe and rapid vision loss.

[0005] Optical coherence tomography (OCT) can provide detailed scans of the macula, helping to detect macular degeneration, diabetic macular edema, and other macular and / or general retinal problems earlier than previously possible. To investigate the extent of retinal degeneration in patients with nAMD, OCT images of the retina (e.g., temporal-domain optical coherence tomography (TD-OCT), spectral-domain optical coherence tomography (SD-OCT), or swept-frequency optical coherence tomography (SS-OCT) images can be acquired and used to label features that can be associated with different levels of degenerative nAMD. SD-OCT is an imaging technique in which light is incident on the retina at various optical frequencies, and the reflected light is collected to capture high-resolution two-dimensional or three-dimensional cross-sectional images of the retina via an interferometric signal detected as a function of frequency. OCT imaging can be used to capture the corresponding intensities of the labeled features of the retina.

[0006] OCT imaging data can visually identify or allow for the identification of various biomarkers. Some techniques currently available for quantifying such biomarkers (e.g., the reflectance of retinal elements such as pigment epithelium detachment (PED) and / or subretinal high reflectance material (SHRM)) may be difficult or impossible to adjust for noise in OCT imaging data and / or the background reflectance of other retinal elements. Therefore, one or more methods and / or systems that recognize and take into account the aforementioned problems are desirable. Summary of the Invention

[0007] In one or more embodiments, a method is provided for quantifying a target retinal element as a biomarker using reflectivity dynamics. A set of optical coherence tomography (OCT) volumes of a subject's retina can be received, wherein each OCT volume in the set corresponds to a different time point in a set of time points, and includes a set of two-dimensional OCT B scans. A set of feature images can be generated for each OCT volume in the set of OCT volumes, wherein the set of feature images visually identifies multiple ophthalmic elements using multiple ophthalmic element markers; wherein the multiple ophthalmic element markers assign different groups of pixels to each of the multiple ophthalmic elements; and wherein the multiple ophthalmic elements include a target retinal element, the retinal pigment epithelium (RPE) layer, and the vitreous body. The reflectivity score of the target retinal element identified in each OCT volume in the set of OCT volumes can be identified using the multiple ophthalmic elements identified by the multiple ophthalmic element markers in each OCT volume in the set of OCT volumes, thereby forming a set of reflectivity scores.

[0008] In one or more embodiments, a system includes one or more data processors; and a non-transitory computer-readable medium containing instructions that, when executed on the one or more data processors, cause the one or more data processors to perform part or all or a portion of the one or more methods described herein.

[0009] In one or more embodiments, a computer program product is provided. The computer program product is tangibly embodied in a non-transitory machine-readable storage medium and includes instructions configured to cause one or more data processors to perform part or all or a portion of the one or more methods described herein. Attached Figure Description

[0010] To more fully understand the principles and advantages disclosed herein, please refer to the following description in conjunction with the accompanying drawings: Figure 1 is a block diagram of a biomarker quantification system according to various embodiments.

[0011] Figure 2 is a block diagram of a portion of the biomarker quantification system from Figure 1, which is described in further detail according to one or more embodiments.

[0012] Figure 3 is an example image used and / or generated by the biomarker quantification system from Figure 1 according to one or more embodiments.

[0013] Figure 4 is a flowchart for biomarker quantification according to various embodiments.

[0014] Figure 5 is a graph showing the example reflectance score distribution according to various embodiments.

[0015] Figures 6 and 7 are graphs illustrating the changes in reflex scores calculated for a group of patients who received treatment at a set of time points according to various embodiments.

[0016] Figure 8 is a block diagram of a computer system according to various embodiments.

[0017] It should be understood that the accompanying drawings are not necessarily drawn to scale, and the objects in the drawings are not necessarily drawn to scale relative to each other. The drawings are depictions intended to clearly and understand the various embodiments of the devices, systems, and methods disclosed herein. Where possible, the same reference numerals will be used throughout the drawings to refer to the same or similar parts. Furthermore, it should be understood that the drawings are not intended to limit the scope of this teaching in any way. Detailed Implementation

[0018] I. Overview Detailed retinal scanning can be used to detect, diagnose, and / or treat various types of ophthalmic diseases (or conditions) to assess retinal elements that may change over time in response to disease progression and / or treatment. As an example, detailed retinal scanning can be used to detect, diagnose, and / or treat neovascular age-related macular degeneration (nAMD) to assess the reflectance of retinal elements present in the retina, such as pigment epithelial detachment (PED) and / or subretinal high-reflectivity material (SHRM). Fibrous or scar tissue in the retina that may occur due to the vascular growth characteristics of nAMD can also be detected, diagnosed, and / or treated in a similar manner. The embodiments described herein provide an improved technique for assessing the reflectance of such retinal elements that is more accurate and reliable than existing methods for assessing retinal element reflectance. More accurate and reliable assessment of retinal element reflectance can help ensure more accurate and comprehensive diagnostic and / or treatment solutions for patients suffering from ophthalmic diseases such as, but not limited to, nAMD, and the resulting scar tissue (e.g., fibrosis).

[0019] Retinal elements (such as SHRM) may be associated with subtype-based fibrotic development. As an example, an undefined SHRM with low reflectivity and no clearly defined boundaries may be associated with immature vessels in the macular neovascularization complex, which may have more solubility or more fluid components, while a well-defined SHRM with higher reflectivity and more clearly defined boundaries may be associated with less fluid and more fibrotic components.

[0020] Retinal elements can consist of at least one of retinal layer elements or retinal pathological elements. The detection and labeling of one or more retinal layer elements can be referred to as layer element (or retinal layer element) segmentation. The detection and labeling of one or more retinal pathological elements can be referred to as pathological element (or retinal pathological element) segmentation.

[0021] In some cases, assessing retinal element reflectance using current methods and systems to accurately adjust for noise, other retinal features, and / or background reflectance of the retinal layer, or both, in OCT imaging data can be difficult. For example, when analyzing the reflectance of selected retinal features (e.g., PED / SHRM), some currently available methods and systems may fail to adjust for noise and / or background reflectance of other retinal features and / or the retinal layer in OCT imaging data, potentially distorting such analyses. Furthermore, manual annotation of each SHRM subtype (e.g., undefined, explicitly defined, etc.) for each B-scan of an OCT volume is overly cumbersome for human graders; this is especially true for a set of OCT volumes within a set of time points.

[0022] Therefore, the embodiments described herein provide improved methods and systems for quantifying the reflectance of ophthalmic elements, such as retinal elements, at one or more time points. For example, embodiments may provide a method for quantifying reflectance using a reflectance score that adjusts for the reflectance of noise and / or other ophthalmic elements, thereby enabling more accurate and reliable quantification of reflectance. Thus, this type of reflectance score can be used more reliably and confidently as a marker of retinal conditions to assess changes in retinal health over time, treatment response, treatment outcomes, etc.

[0023] Recognizing and considering the importance and utility of methods and systems that can provide the aforementioned improvements, this disclosure describes various embodiments for quantifying the reflectance of selected ophthalmic elements (e.g., calculating reflectance scores), which may include segmenting and labeling multiple selected ophthalmic elements using ML-based algorithms. The embodiments described herein enable more accurate and reliable assessment of ophthalmic element reflectance, which can improve the accuracy and reliability of any detection, diagnosis, and / or treatment methods that rely on the results of such reflectance assessment. In some embodiments, full automation makes reflectance calculation unbiased and unaffected, or at least less biased and less affected by any underlying pathology, compared to manual grading.

[0024] The embodiments described herein for quantifying the reflectance of ophthalmic elements also provide improved techniques for three-dimensional measurement of such retinal elements, which are more accurate and reliable than existing methods. More accurate and reliable three-dimensional measurements can help ensure more accurate and comprehensive diagnostic and / or treatment solutions for patients suffering from ophthalmic diseases such as, but not limited to, nAMD and the resulting scar tissue (e.g., fibrosis).

[0025] II. Reflectance Score Calculation Based on Machine Learning (ML) II.A. Example system for reflectivity scoring calculation Figure 1 is a block diagram of a biomarker quantification system 100 according to various embodiments. The biomarker quantification system 100 is used to automatically assess the retinal condition of a subject by automatically quantifying the reflectance of ophthalmic elements captured in OCT imaging data of a subject suffering from various ophthalmic diseases (or conditions), such as, but not limited to, nAMD, using image input 102 that can be received or accessed via network 104. In some embodiments, the retina is a healthy retina. In other embodiments, the retina is a retina that has been diagnosed with or is suspected of having a retinal disease. For example, the diagnosis may be one of the following: age-related macular degeneration (AMD), diabetic macular edema (DME), or some other type of retinal disease.

[0026] Typically, the biomarker quantification system 100 is used to quantify biomarkers identified by OCT imaging data. For example, biomarkers may take the form of OCT features. OCT features can be features that can be visualized on OCT images, identified from OCT images, or otherwise measured using OCT images to provide indications about normal biological processes, pathogenic processes, or pharmacological responses to therapeutic interventions in a subject's retina.

[0027] As shown in Figure 1, the biomarker quantification system 100 includes a computing platform 106 configured to store and execute an image processor 108, an ophthalmic feature recognition system 110, and a score generator 112. Typically, the image processor 108 receives or accesses image input 102 and generates a processed image 114. The processed image 114 serves as input to the ophthalmic feature recognition system 110, which uses the processed image 114 to generate a set of feature images 116. As shown, the ophthalmic feature recognition system 110 may include a retinal segmentation model 118 and an ophthalmic feature algorithm 120. The retinal segmentation model 118 and the ophthalmic feature algorithm 120 are used to generate a set of feature images 116, which are sent to the score generator 112 to generate a set of reflectance scores 122. The set of reflectance scores 122 can be used to generate an output 126 using an output generator 124, which can be transmitted via a network 104 to a remote device 128.

[0028] While the image processor 108, ophthalmic feature recognition system 110, score generator 112, and output generator 124 are shown to be stored and executed using the same computing platform (i.e., computing platform 106), in some embodiments, one or more of these components may be stored and executed using a different computing platform than computing platform 106. In some embodiments, the biomarker quantification system 100 also includes a data storage device 130 and a display system 132. The data storage device 130 and display system 132 each communicate with computing platform 106. In some instances, the data storage device 130, display system 132, or both may be considered part of or otherwise integrated with computing platform 106. Therefore, in some instances, computing platform 106, data storage device 130, and display system 132 may be separate components communicating with each other, but in other instances, combinations of these components may be integrated together.

[0029] As shown, image input 102 may include OCT imaging data 134, which can be generated using an OCT imaging system 136 or an OCT scanner. The OCT imaging system 136 may be a large desktop configuration for a clinical setting, a portable or handheld dedicated system, or a "smart" OCT system incorporated into a user's personal device, such as a smartphone. In some cases, the OCT imaging system 136 may include an image denoiser configured to remove noise and other artifacts from the raw OCT volume images to generate OCT volumes. In one or more embodiments, the OCT imaging data 134 includes a set of OCT volumes 138 of the subject's retina. In some embodiments, a set of OCT volumes 138 corresponds to the patient's retina at a given (i.e., one) time point. In one or more embodiments, a set of OCT volumes 138 corresponds to the patient's retina at a set of time points. For example, each OCT volume 140 in a set of OCT volumes 138 may correspond to a specific time point in a set of time points (e.g., the baseline time point, the number of hours, days, or months following the baseline time point, or some other type of time point). Each OCT volume 140 in a set of OCT volumes 138 may consist of a set of OCT B scans 142 of the subject's retina. A set of OCT B scans 142 may include, for example, but not limited to, dozens, hundreds, thousands, tens of thousands, or some other number of OCT B scans. OCT B scans may also be referred to as OCT slice images or cross-sectional OCT images.

[0030] Although only one of each OCT imaging system 136 and biomarker quantification system 100 is shown, in other embodiments, there may be more than one of each. Furthermore, although Figure 1 shows the OCT imaging system 136 and biomarker quantification system 100 as two separate components, in some embodiments, the OCT imaging system 136 and biomarker quantification system 100 may be part of the same system (e.g., maintained by the same entity, such as a healthcare provider or clinical trial administrator). In some cases, a portion of the biomarker quantification system 100 may be implemented as part of the OCT imaging system 136. For example, the biomarker quantification system 100 may be configured to operate as a module implemented using the processor, microprocessor, or some other hardware component of the OCT imaging system 136. In still other embodiments, the biomarker quantification system 100 may be implemented within a cloud computing system that can be accessed by or otherwise communicated with by the OCT imaging system 136.

[0031] In one embodiment, image processor 108 is configured or programmed to receive a set of OCT B scans 142 (i.e., image input 102) of a set of OCT volumes 138 and perform a set of processing operations thereon to form a processed image 114. The set of OCT volumes 138 may be sent as input to image processor 108, retrieved by image processor 108 from a storage device, or accessed in some other way. The set of processing operations may include, for example, but not limited to, normalization, scaling, resizing, horizontal flipping, vertical flipping, cropping, rotation, noise filtering, or at least one of some other type of preprocessing operations.

[0032] Image processor 108 may be implemented using hardware, software, firmware, or a combination thereof. In one or more embodiments, image processor 108 may be implemented within computing platform 106, but in other embodiments, at least a portion (e.g., a module) of image processor 108 may be implemented within OCT imaging system 136.

[0033] The ophthalmic feature recognition system 110 can be composed of any number of models, algorithms, formulas, and / or equations that can be used to process image input 102 and / or processed image 114. When the ophthalmic feature recognition system 110 processes image input 102, image input 102 may include a set of OCT volumes 138; it can be formed by preprocessing a set of OCT volumes 138; it may include OCT volume 140 and its corresponding set of OCT B scans 142; or it can be formed by preprocessing OCT volume 140 and its corresponding set of OCT B scans 142. In some embodiments, the ophthalmic feature recognition system 110 may process image input 102 and / or processed image 114 to generate, for example, a biomarker score, which may then be processed by output generator 124 to form output 126. The ophthalmic feature recognition system 110 processes image input 102 and / or processed image 114 to generate a set of feature images 116 comprising each OCT B scan in a set of OCT B scans ...

[0034] In one or more embodiments, the ophthalmic feature recognition system 110 includes a retinal segmentation model 118 and an ophthalmic feature algorithm 120. In some embodiments, the retinal segmentation model 118 may generate a first segmented image (e.g., a first segmented image 144) of each OCT B scan in a set of OCT B scans 142 for each OCT volume 140 in a set of OCT volumes 138 of the image input 102 and / or the processed image 114. For example, one or more of the first segmented images may be generated from OCT imaging data according to one or more OCT segmentation techniques as described in International Publication No. WO2023205511A1, which is incorporated herein by reference in its entirety. In some embodiments, a set of feature images 116 visually identifies multiple ophthalmic features using multiple ophthalmic feature markers. Ophthalmic features may consist of retinal layer features or retinal pathology features, or may be associated with at least one of retinal layer features or retinal pathology features.

[0035] Retinal layer elements can be, for example, the retinal layer itself or boundaries associated with it. Examples of retinal layers include, but are not limited to, the internal limiting membrane (ILM), external limiting membrane (ELM), external retinal loop-Henler fiber layer (OPL-HFL), retinal pigment epithelium (RPE), RPE detachment layer, Bruch's membrane (BM), ellipsoidal zone (EZ), and other types of retinal layers. Boundaries associated with the retinal layer can be, for example, the inner boundary of the retinal layer, the outer boundary of the retinal layer, boundaries associated with pathological features of the retinal layer (e.g., the inner or outer boundary of retinal detachment), or some other type of boundary. For example, a boundary can be the inner boundary of an RPE (IB-RPE) detachment layer, the outer boundary of an RPE (OB-RPE) detachment layer, or another type of boundary.

[0036] Retinal pathological elements can include, for example, effusions, cells, solid materials, or combinations thereof that demonstrate retinal pathology associated with an ophthalmic disease or condition. For instance, the presence of certain retinal effusions may be an indication of retinal vascular leakage, which could be a sign of nAMD. Examples of retinal pathological elements include, but are not limited to, intraretinal effusion (IRF), subretinal effusion (SRF), effusion associated with pigment epithelium detachment (PED), high reflectivity material (HRM), subretinal high reflectivity material (SHRM), intraretinal high reflectivity material (IHRM), high reflectivity foci (HRF), hard exudates (HE), retinal effusion pockets, and destruction. In some cases, retinal pathological elements may be destruction of a retinal layer or retinal region (e.g., discontinuity, delamination, loss, etc.). For example, destruction may be of the ellipsoidal region, ELM, RPE, or another layer or region. Destruction may indicate damage or loss of cells (e.g., photoreceptors) in the destroyed area.

[0037] In some cases, retinal layer elements are associated with retinal pathology elements. For example, the RPE detachment layer (which is a retinal layer element) is associated with PED (which is a retinal pathology element). In other cases, ophthalmic elements are associated with either retinal layer elements or retinal pathology elements. For example, the vitreous body (which is an ophthalmic element) is associated with the ILM layer. In some embodiments, the vitreous body may be alternatively referred to as the vitreous matrix or vitreous chamber.

[0038] In one or more embodiments, the retinal segmentation model 118 may additionally detect and label ophthalmic elements that are further classified as subtypes of ophthalmic elements. As a non-limiting example, the retinal segmentation model 118 may additionally detect and label one or more subtypes of SHRM, including, for example, but not limited to, undefined SHRM and / or defined SHRM. Undefined SHRM may be more common in the early stages of retinal diseases (e.g., nAMD, etc.) and may be completely eliminated or converted to defined SHRM through treatment.

[0039] In some embodiments, the retinal segmentation model 118 may include, but is not limited to, a convolutional neural network (such as the U-net model) trained to identify regions of intraretinal fluid effusion (IRF), subretinal fluid effusion (SRF), SHRM, and PED on pixel-level OCT B scans using OCT training data. In one or more embodiments, the retinal segmentation model 118 is trained using training data comprising multiple OCT B scan images annotated by a human grader to classify IRF, SRF, SHRM, and PED within the OCT images. In one or more embodiments, the retinal segmentation model 118 is trained using training data comprising multiple OCT B scan images annotated by a human grader to delineate the contours of IRF, SRF, PED, and SHRM. Based on whether ≤50% or >50% of SHRM can be depicted from the neurosensory retina, OCT B scans included in the training data including SHRM annotations are further manually labeled as including undefined SHRM or explicitly defined SHRM. The labels “unqualified SHRM” and “explicitly qualified SHRM” are used together with the background labels for training. IRF, SRF, and PED annotations are also included to allow learning a wider range of features.

[0040] Therefore, the retinal segmentation model 118 can be used to perform layer feature segmentation to detect and label retinal layer features and / or retinal pathology features, thereby detecting and labeling ophthalmic features. In some embodiments, ophthalmic features can be subtypes of retinal layer features or retinal pathology features.

[0041] In some embodiments, the ophthalmic feature algorithm 120 may use a first segmented image (e.g., first segmented image 144) generated by the retinal segmentation model 118 to generate a second segmented image (e.g., second segmented image 146) to identify ophthalmic features associated with the retinal features (i.e., retinal layer features and / or retinal pathology features) identified in the first segmented image. The second generated image may include multiple ophthalmic feature labels that identify additional ophthalmic features relative to those identified in the first generated image by assigning pixel groups to additional ophthalmic features.

[0042] In one or more embodiments, the ophthalmic feature algorithm 120 may use one or more ophthalmic features marked on the first segmented image to mark other ophthalmic features. For example, but not limited to, the ophthalmic feature algorithm 120 may use an ILM layer, such as that marked by the retinal segmentation model 118, on the first segmented image to mark the vitreous body as another ophthalmic feature.

[0043] The outputs of retinal segmentation model 118 and ophthalmic feature algorithm 120 (i.e., the first segmented image and the second segmented image) are used to generate a set of feature images 116 (e.g., a first feature image 148 and / or a second feature image 150). In one or more embodiments, the set of feature images 116 includes a first feature image that at least identifies target retinal features (e.g., PED or SHRM) and a second feature image that at least identifies the RPE layer and vitreous body. In some embodiments, the feature images in the set of feature images 116 are post-processed by image processor 108 by performing a pixel erosion operation. In one or more embodiments, the pixel erosion operation may include, for example, but not limited to, eroding multiple pixels from each direction. For example, but not limited to, one, two, three, five, eight, ten, twenty, or some other number of pixels may be removed from each direction in the pixel erosion operation. In some embodiments, a secondary cleaning operation may be performed after the pixel erosion operation. For example, but not limited to, one, two, three, five, eight, ten, twenty, or some other number of pixels may be removed from only one, two, or three directions. The following discussion will refer to Figure 2 for a more detailed analysis of a set of feature images 116.

[0044] As shown, a set of feature images 116 is sent to a scoring generator 112, which uses at least one feature image from the set of feature images 116 to generate one or more scores for each corresponding OCT B scan. For example, the scoring generator 112 may generate a reflectance score for the target retinal feature identified in each OCT B scan in a set of OCT B scans 142, thereby forming a set of reflectance scores 122. Each reflectance score in the set of reflectance scores 122 is a quantification of the reflectance of the target retinal feature, which takes into account noise in the OCT B scan as well as the reflectance of the RPE layer and vitreous body and has been adjusted accordingly.

[0045] Output generator 124 can generate output 126 using a set of reflectance scores 122. Output 126 can include a classification of retinal symptoms at a given time point or a set of time points based on the set of reflectance scores 122. Output 126 can include a report based on the set of reflectance scores 122. The report may be an indication of disease progression, treatment response, or treatment outcome.

[0046] Figure 2 is a block diagram of a portion of the biomarker quantification system 100 of Figure 1, described in further detail according to one or more embodiments. After image input 102 has been received by the biomarker quantification system 100 and used to generate a processed image 114, the biomarker quantification system 100 processes the processed image 114. As described above and in one or more embodiments, image input 102 and / or processed image 114 may comprise a set of OCT B scans 142, or may be formed by preprocessing a set of OCT B scans 142.

[0047] As discussed above, the biomarker quantification system 100 may include an ophthalmic feature recognition system 110, which may include a retinal segmentation model 118 and an ophthalmic feature algorithm 120, which are used to generate a first segmented image 144 and a second segmented image 146, respectively.

[0048] As shown, a first segmented image 144 generated by retinal segmentation model 118 may include multiple ophthalmic feature markers, which identify multiple ophthalmic features (e.g., ophthalmic features 202, 204, and 206) by assigning pixel groups to each of the multiple ophthalmic features. In one or more embodiments, the multiple ophthalmic features include at least one target retinal feature, such as detached pigment epithelium (PED) or subretinal high reflectivity material (SHRM), internal limiting membrane (ILM) layer, and retinal pigment epithelium (RPE) layer. In FIG2, the first segmented image 144 includes the ILM layer (ophthalmic feature 202), PED (ophthalmic feature 204), and RPE layer (ophthalmic feature 206). In some embodiments, the entire RPE across each B-scan is used instead of the RPE in the SHRM region across each B-scan because the RPE can be better preserved around the macular neovascularization (MNV) lesion.

[0049] As further shown, the second segmented image 146 can be generated by the ophthalmic feature algorithm 120 using the first segmented image 144 to identify additional ophthalmic features using multiple ophthalmic feature labels by assigning pixel groups to additional ophthalmic features. In Figure 2, the second segmented image 146 identifies the vitreous body as an additional ophthalmic feature (ophthalmic feature 208) using the ILM layer (ophthalmic feature 202) as indicated by the retinal segmentation model 118 on the first segmented image 144.

[0050] Then, the first segmented image 144 and the second segmented image 146 are used to generate a set of feature images 116, which include a first feature image 148 and a second feature image 150. The first feature image 148 at least identifies the target retinal feature (e.g., PED or SHRM), while the second feature image 150 at least identifies the RPE layer and other ophthalmic features as identified by the ophthalmic feature algorithm 120. In FIG2, the first feature image 148 identifies the PED as the target retinal feature (ophthalmic feature 204), while the second feature image 150 identifies the RPE layer (ophthalmic feature 206) and the vitreous body (ophthalmic feature 208).

[0051] A set of feature images 116 (e.g., at least one feature image of the set of feature images 116) generated for each OCT B scan of image input 102 can be used by score generator 112 to generate one or more scores for each corresponding OCT volume in image input 102. For example, score generator 112 can generate reflectance scores for the target retinal features marked in each OCT volume, thereby forming a set of reflectance scores 122. Each reflectance score in the set of reflectance scores 122 is a quantification of the reflectance of the target retinal feature in the OCT volume at a certain time point, which takes into account the noise in the OCT volume as well as the reflectance of the RPE layer and vitreous body and has been adjusted accordingly.

[0052] Reflectance scores can be calculated in various ways, such as by quantifying the intensity measure of the target retinal element and adjusting the intensity measure to take into account the reflectance of the RPE layer and vitreous body (e.g., intensity measures of the RPE layer and vitreous body).

[0053] In some embodiments, the intensity measure of an ophthalmic feature can be quantified by first labeling the pixel intensity values ​​assigned to groups of pixels of an ophthalmic feature, such as those labeled in a set of feature images 116, using one or more ophthalmic feature markers. For example, the intensity measure of a PED, such as that labeled in the first feature image 148 (ophthalmic feature 204), can be quantified by first labeling the pixel intensity values ​​assigned to groups of pixels of ophthalmic feature 204.

[0054] In some embodiments, for each OCT B scan in which a set of feature images 116 has been generated, the pixel intensity value of the ophthalmic feature can be identified. In other embodiments, the pixel intensity value of the ophthalmic feature can be further processed by the scoring generator 112 to normalize the pixel intensity value of the ophthalmic feature in each OCT B scan of the OCT volume in which a set of feature images 116 has been generated.

[0055] The intensity measure of ophthalmic features in an OCT volume can be indicated by the pixel intensity values ​​of ophthalmic features that meet selected criteria in each OCT B scan of the OCT volume. In some embodiments, the selected criteria can be the median pixel intensity value. In some embodiments, the selected criteria can be the pixel intensity value at a selected percentile. The selected percentile can be any percentile between a given range (such as, for example, between the 90th and 99th percentiles).

[0056] In one or more embodiments, the rating generator 112 calculates intensity measures of a plurality of ophthalmic features and then uses them as a reference to adjust the intensity measures of a target retinal feature (e.g., when the reflectance rating of the target retinal feature is indicated) to take into account a first reflectance of a reference that is darker than the target retinal feature and a second reflectance of a reference that is brighter than the target retinal feature.

[0057] The reflectance score of a target retinal element can be represented by the following equation, where An intensity measure for identifying target retinal features. An intensity measure indicated for a reference that is darker than the target retinal element, and An intensity measure for a reference element that is brighter than the target retinal element: For example, the rating generator 112 can calculate intensity measures of at least the target retinal element (e.g., PED or SHRM), the RPE layer, and the vitreous body. These intensity measures can then be used to calculate a reflectance rating of the target retinal element, which uses the reflectance of the RPE layer as a reference that is brighter than the target retinal element and the vitreous body as a reference that is darker than the target retinal element.

[0058] For example, the reflectance score of a target retinal element can be represented by the following equation, where An intensity measure for identifying target retinal features. For the intensity measure indicated for the vitreous body, and For the strength metric indicated for the RPE layer: In one or more embodiments, It can be calculated as the median intensity of the target retinal feature marked on each OCT B scan within a set of OCT volumes. It can be calculated as the median intensity of the vitreous body indicated on each OCT B scan within a set of OCT volumes, and It can be calculated as the pixel intensity value of RPE at the 95th percentile of each OCT B scan in a set of OCT volumes.

[0059] In one or more embodiments, a set of reflectance scores 122 is a set of biomarkers for disease progression. In some embodiments, a set of reflectance scores 122 is a set of biomarkers for at least a treatment outcome or treatment response to a selected treatment of the retina. For example, but not limited to, a set of reflectance scores 122 may be a set of biomarkers for at least a treatment outcome or treatment response to an anti-angiogenic therapy of the retina (such as ranibizumab, aflibercept, bevacizumab, fareximab, or broolizumab).

[0060] In some embodiments, clinical trials may be modified, redesigned, replicated, or evaluated in response to a set of reflectance scores 122. For example, but not limited to, clinical trials investigating the therapeutic effects of retinal diseases (e.g., nAMD or fibrosis) may be modified, redesigned, replicated, or evaluated in response to a set of reflectance scores generated for target retinal elements that serve as biomarkers of retinal diseases.

[0061] As discussed above, a set of reflectance scores 122 can then be used by output generator 124 to generate output 126. Output 126 may include indications of a classification of retinal symptoms, disease progression, treatment response, or treatment outcome at a time point within a set of time points. In one or more embodiments, clinical trials may be modified, redesigned, replicated, or evaluated in response to output 126. For example, but not limited to, clinical trials investigating the therapeutic effects of retinal diseases (e.g., nAMD or fibrosis) may be modified, redesigned, replicated, or evaluated in response to output 126 generated using a set of reflectance scores 122 generated for target retinal features that are biomarkers of retinal diseases. In some embodiments, using image processor 108 and / or ophthalmic feature recognition system 110 to process image input 102 may reduce the overall computational resources required to generate such reflectance scores in other ways. In some embodiments, ophthalmic feature recognition system 110 generating a set of feature images 116 allows score generator 112 to generate more accurate output more efficiently, thereby saving computational resources.

[0062] In some embodiments, the biomarker quantification system 100 assesses and characterizes ophthalmic elements with greater accuracy and consistency than expert human graders. Therefore, and in some embodiments, the biomarker quantification system 100 provides the technical effects of improved accuracy, reduced overall computational resources, and / or the provision of previously non-automated fully automated biomarker quantification.

[0063] In some embodiments, the biomarker quantification system 100 provides a technological improvement for the field of three-dimensional SHRM research. For example, the use of the biomarker quantification system 100 allows or enables the study of SHRM and its subtypes in a three-dimensional manner, which allows not only point measurements but also frontal area and volume measurements. Furthermore, the use of the biomarker quantification system 100 allows or enables the restriction of measurements to specific regions of interest for ETDRS.

[0064] II.B. Example image used for biomarker quantification Figure 3 includes example images 300 used and / or generated by the biomarker quantification system 100 according to various embodiments. Image 302 is an example OCT B scan (e.g., an OCT B scan in a set of OCT B scans 142) that can be used by the biomarker quantification system 100 as image input 102. Image 304 is an example of a first generated image 144 such as that generated by the retinal segmentation model 118 of the ophthalmic feature recognition system 110. On image 304, undefined and explicitly defined SHRMs have been detected and labeled as ophthalmic features. In some embodiments, a set of reflectance scores 122 may be calculated separately by the biomarker quantification system 100 for undefined SHRMs and explicitly defined SHRMs such as those labeled by the retinal segmentation model 118.

[0065] II.C. Example methods for calculating reflectance scores Figure 4 is a flowchart of a process 400 for quantifying a target retinal element as a biomarker using reflectivity dynamics, according to various embodiments. The target retinal element may be a biomarker associated with retinal diseases such as, for example, nAMD or fibrosis. The target retinal element can be quantified by identifying reflectivity dynamics (such as, for example, a set of reflectivity scores at a set of time points). In various embodiments, process 400 is implemented using the biomarker quantification system 100 described in Figure 1.

[0066] Step 402 of process 400 includes receiving a set of optical coherence tomography (OCT) volumes of the subject's retina, wherein each OCT volume in the set corresponds to a different time point in a set of time points, and includes a set of two-dimensional OCT B scans. The set of OCT volumes can be, for example, a set of OCT volumes 138 as shown in FIG. 1. The set of two-dimensional OCT B scans can be, for example, a set of OCT B scans 142 as shown in FIG. 1. The set of OCT volumes can be used as an image input for process 400, or can be preprocessed to form that image input. For example, a set of OCT volumes 138 and the corresponding set of OCT B scans 142 can be used as an image input 102 of the biomarker quantification system 100 as described in FIG. 1, or preprocessed thereto. Examples of image input 102 are further discussed above with reference to FIG. 3.

[0067] Step 404 of process 400 includes generating a set of feature images for each OCT volume in a set of OCT volumes, wherein multiple ophthalmic features are visually identified in the set of feature images using multiple ophthalmic feature markers. The multiple ophthalmic feature markers can assign different groups of pixels to each of the multiple ophthalmic features, and the multiple ophthalmic features can include at least one target retinal feature, the retinal pigment epithelium (RPE) layer, and the vitreous body. A set of retinal images can be, for example, a set of feature images 116 as shown in Figure 1.

[0068] In one or more embodiments, generating a set of feature images for each OCT volume in a set of OCT volumes may include generating a first segmentation image by a retinal segmentation model for each OCT B scan in a set of OCT B scans for each OCT volume in a set of OCT volumes. The first segmentation image may include a plurality of ophthalmic feature markers that identify multiple ophthalmic features by assigning pixel groups to each of the multiple ophthalmic features. The first segmentation image may be, for example, a first segmentation image 144 generated by the retinal segmentation model 118 in FIG. 1. An example of the first segmentation image 144 is further discussed above with reference to FIG. 3.

[0069] In some embodiments, a first feature image is generated using a first segmented image, at least identifying the target retinal feature. In some embodiments, a second feature image is generated using the first segmented image, at least identifying the RPE layer and additional ophthalmic features. The first and second feature images can be, for example, first feature image 148 and second feature image 150 in Figures 1 and 2. The target retinal feature can be a retinal layer feature or a retinal pathology feature, as described above with reference to Figure 1. For example, the target retinal feature can be a retinal pathology feature, such as pigment epithelial detachment (PED) or subretinal high reflectivity material (SHRM). The target retinal feature can be, for example, ophthalmic feature 204 (i.e., PED) in Figure 2. The RPE layer can be, for example, ophthalmic feature 206 in Figure 2. In some embodiments, the target retinal feature can be one or more subtypes of retinal features, such as, for example, but not limited to, explicitly defined and / or undefined SHRM (i.e., subtypes of SHRM).

[0070] In one or more embodiments, generating the second feature image includes generating a second segmentation image to identify additional ophthalmic features using a first segmentation image and an ophthalmic feature algorithm. The second segmentation image may include multiple ophthalmic feature markers that identify additional ophthalmic features by assigning pixel groups to them. The second segmentation image may be, for example, a second segmentation image 146 generated by the ophthalmic feature algorithm 120 in FIG1.

[0071] Generating a second segmented image may include using one or more ophthalmic features marked on the first segmented image to mark additional ophthalmic features. As a non-limiting example, in one or more embodiments, the additional ophthalmic feature may be the vitreous body and may be marked using the internal limiting membrane (ILM) layer marked on the first segmented image. The ILM layer may be, for example, ophthalmic feature 208 in FIG. 2. The vitreous body may be, for example, ophthalmic feature 202 in FIG. 2.

[0072] In one or more embodiments, a set of feature images can be post-processed by performing one or more pixel erosion operations to remove multiple pixels from each feature image in the set of feature images from various directions, removing multiple pixels from each feature image in the set of feature images from one or more directions, or a combination thereof. In some embodiments, one or more pixel erosion operations can be performed to reduce the overlap between the target retinal feature (such as that identified by the ophthalmic feature recognition system 100) and the RPE layer.

[0073] For example, but not limited to, a pixel erosion operation can be performed on a first feature image (e.g., a feature image that identifies the target retinal feature) to remove one pixel from each direction. A second pixel erosion operation can be performed on the first feature image to remove three pixels above the target retinal feature (e.g., where the target retinal feature is PED) or eight pixels below the target retinal feature (e.g., where the target retinal feature is SHRM).

[0074] Step 406 of process 400 includes labeling a target retinal element labeled in each of a set of OCT volumes with multiple ophthalmic element markers in each of the set of OCT volumes, thereby forming a set of reflectance scores. In some embodiments, determining the reflectance score of the target retinal element includes quantifying an intensity measure of the target retinal element and adjusting the intensity measure of the target retinal element to take into account a first reflectance of the vitreous body and a second reflectance of the RPE layer. A set of reflectance scores may be, for example, a set of reflectance scores 122 generated by the score generator 112 in FIG. 1.

[0075] In one or more embodiments, a set of reflectance scores for a target retinal element includes baseline reflectance scores at a set of time points, and thresholds for separating reflectance scores indicating low reflectance and reflectance scores indicating high reflectance can be selected based on an analysis of the distribution of multiple baseline reflectance scores calculated for multiple subjects.

[0076] In some embodiments, a set of reflectance scores may include a baseline reflectance score associated with a baseline time point and a first reflectance score associated with a first time point after the baseline time point. In some embodiments, a set of reflectance scores may include the reflectance variation between the baseline reflectance score associated with the baseline time point and the first reflectance score associated with the first time point after the baseline time point.

[0077] In one or more embodiments, a set of reflectance scores is a set of biomarkers for disease progression. In some embodiments, a set of reflectance scores is a set of biomarkers for at least a treatment outcome or treatment response to a selected treatment of the retina. For example, but not limited to, a set of reflectance scores may be a set of biomarkers for at least a treatment outcome or treatment response to an anti-angiogenic therapy of the retina (such as ranibizumab, aflibercept, bevacizumab, fareximab, or broolizumab).

[0078] In some embodiments, clinical trials may be modified, redesigned, replicated, or evaluated in response to a set of reflectance scores. For example, but not limited to, clinical trials investigating the therapeutic effects of retinal diseases (e.g., nAMD or fibrosis) may be modified, redesigned, replicated, or evaluated in response to a set of reflectance scores generated for target retinal elements that serve as biomarkers of retinal diseases.

[0079] Process 400 may optionally include step 408 of generating output using a set of reflectance scores. The output may include indications of a classification of retinal symptoms, disease progression, treatment response, or treatment outcome at time points within a set of time points. In one or more embodiments, clinical trials may be modified, redesigned, replicated, or evaluated in response to the generated output. For example, but not limited to, clinical trials investigating the therapeutic effects of retinal diseases (e.g., nAMD or fibrosis) may be modified, redesigned, replicated, or evaluated in response to the output generated using a set of reflectance scores targeting target retinal elements as biomarkers of retinal diseases. The generated output may be, for example, output 126 generated by output generator 124 in FIG. 1.

[0080] In some embodiments, process 400 measures and characterizes ophthalmic elements with greater accuracy and consistency than expert human graders. In some embodiments, process 400 provides the technical effects of improved accuracy, reduced overall computational resources, and / or the provision of fully automated biomarker quantification that conventional systems cannot provide.

[0081] In some embodiments, process 400 provides a technological improvement for the field of three-dimensional SHRM research. For example, the use of biomarker quantification system 100 can allow or enable the study of SHRM and its subtypes in a three-dimensional manner, allowing not only point measurements but also frontal area and volume measurements. Furthermore, the use of biomarker quantification system 100 can allow or enable the restriction of measurements to specific ETDRS regions of interest. In some embodiments, process 400 includes novel combinations of steps that achieve technological improvements relative to conventional SHRM grading and characterization methods.

[0082] In some embodiments, process 400 provides a technological improvement in the field of fibrosis detection and / or prediction. Traditionally, fibrosis is graded using color and fluorescein angiography (FA) characteristics, which is difficult for manual graders to consistently achieve, leading to variability in fibrosis detection and / or prevalence. In some embodiments, process 400 provides an automated and objective method for detecting fibrosis by using a well-defined SHRM as an alternative to the presence and development of fibrosis. For example, and as discussed below, a well-defined SHRM has a higher internal reflectance compared to an undefined SHRM, and the higher reflectance and well-defined boundaries of the SHRM may indicate the presence of a fibrotic process. Therefore, in some embodiments, the biomarker quantification system 100 and / or process 400 can use image input 102 associated with a subject to identify whether the subject has fibrosis or is at high risk of developing fibrosis. In some embodiments, process 400 further includes administering a therapeutic or preventative therapy to subjects identified as having a fibrotic process or at high risk of developing fibrosis to prevent or minimize further fibrotic development.

[0083] In some embodiments, process 400 provides technological improvements in the field of SHRM subtypes regarding their distribution, characteristics, and association with visual function in nAMD eyes. In some embodiments, process 400 can automate, volumetrically, and objectively compare different mechanisms of action in future nAMD treatments, and, due to a more comprehensive understanding of the disease, can provide a reference for better treatment strategies.

[0084] II.D. Example results of reflectivity score calculation Figure 5 is a diagram 500 illustrating an example distribution of reflectance scores calculated in the manner described above using the biomarker quantification system 100 of Figure 1 or the process 400 of Figure 4, to automatically assess retinal lesions by automatically quantifying the reflectance of target retinal elements. In one or more embodiments, the distribution map (such as the one depicted in Figure 5) can be used to assess a set of OCT volumes for multiple patients at a given time point. For example, the distribution map at the baseline time point can be used to analyze the subsequent reflectance distribution of target retinal elements at subsequent time points.

[0085] Figures 6 and 7, respectively (Figures 600 and 700), illustrate the changes in PED and SHRM reflectance scores calculated as target retinal elements in a set of OCT volumes for patients diagnosed with nAMD (a total of 108 study eyes) who received aflibercept treatment (i.e., anti-VEGF therapy for nAMD) at weeks 0, 4, 8, 16, and 24. OCT volumes for patients were generated at weeks 0, 8, and 24 to determine the baseline time point and two additional time points, thereby examining the correlation between PED and SHRM volumes and reflectance two and six months after treatment. Reflectance scores can be examples of a set of reflectance scores 122 calculated as described above using the biomarker quantification system 100 in Figure 1 or the procedure 400 in Figure 4.

[0086] Figure 6 shows that PED reflectivity increases relative to the baseline over time, while PED volume decreases relative to the baseline over time. Similarly, Figure 7 shows that SHRM reflectivity increases relative to the baseline over time, while SHRM volume decreases over time.

[0087] II.E. Example analysis of reflectivity score calculation using the example system. In some embodiments, and during an example post-hoc analysis using the biomarker quantification system 100, at least some medical data from patients involved in the clinical trial are also received. In some embodiments, the medical data includes a treatment-naïve state, which provides an indication of whether the subject has previously received nAMD treatment (i.e., is treatment-naïve). In some embodiments, the medical data includes demographic data, such as age, sex, one or more other types of demographic variables, or combinations thereof. In some embodiments, the medical data also includes clinical data, such as visual acuity measurements or another type of clinical parameter or measurement. Visual acuity measurements can be, for example, baseline (relative to the start of treatment within the trial) and BCVA at one or more reference time points. In some cases, reference time points are time points after baseline and / or after the first dose of treatment, such as, for example, 4 weeks, 6 weeks, 18 weeks, 24 weeks, 3 months, 6 months, 9 months, 1 year, 2 years, or some other time period after the first dose of treatment. Typically, clinical data corresponding to reference time points can include data generated at the reference time point, data generated within a selected range of the reference time point (e.g., within a selected number of days, weeks, or months), or both.

[0088] In one example analysis, a total of 1094 eligible, untreated study eyes from 1094 patients were included (out of all 1326 TENAYA / LUCERNE study eyes). Of these, 655 were female and 439 were male, with a mean age of 75.1 years (range: 50–89; SD: 7.9) and a mean BCVA of 60.3 ETDRS letters (range: 24–78; SD: 13.1). Based on the segmentation performed and reflectivity scores indicated by the biomarker quantification system 100, most eyes (i.e., 79%) had some degree of SHRM; undefined SHRM was significantly higher in terms of volume, frontal area, and maximum height within the 3 mm diameter of the ETDRS, including eyes with zero values ​​for the corresponding SHRM subtype, compared to defined SHRM; and undefined zero values ​​were also fewer than defined SHRM (12% vs. 25%). Reflectance for each SHRM subtype was also assessed by including only eyes with non-zero segmentation for each subtype. The mean reflectance for the defined SHRM volumes was 0.59 (n=794), ranging from 0.28 to 0.89, and significantly higher than the mean reflectance for the undefined SHRM, which was 0.55 (n=952) ranging from 0.21 to 0.98 (Difference: 0.04, 95% confidence interval (CI) 0.03 to 0.04, P<.001). Reflectance heterogeneity, measured by the standard deviation across all pixels in the SHRM volume, was 0.11 in the defined SHRM and 0.12 in the undefined SHRM, ranging from 0 to 0.28 and 0 to 0.26, respectively (Difference: -0.007, 95% CI -0.01 to -0.004, p<0.001). Both the mean and SD reflectance variables exhibit a normal distribution.

[0089] When MNV categories are differentiated by central reading center grading of fluorescein angiography (FA), the median volume of undefined and defined SHRMs is highest in classic and major classic MNVs, gradually decreasing to more latent MNV categories. Within the classic and major classic MNV groups, the proportion of defined SHRMs is higher than in the more latent categories. Detailed summary statistics of all dimensions of undefined and defined SHRMs for each MNV category reveal trends regarding frontal area and highly repetitive volume.

[0090] Using least squares regression analysis, a negative statistically significant association between BCVA and SHRM was observed in the ETDRS 1 mm diameter for both undefined and defined SHRM, as well as for all SHRM dimensions at baseline. The association was stronger in defined SHRM compared to undefined SHRM. For example, for a volume with a central 1 mm diameter, an additional 100 nl of undefined SHRM was associated with a loss of 6.0 ETDRS letters, while for defined SHRM, it was associated with a loss of 18.3 ETDRS letters. The ETDRS 1 mm diameter measurement showed the strongest association with BCVA, particularly in terms of volume and frontal area. Notably, the ETDRS 1 mm ring volume is affected by a ceiling effect (due to the smaller radius limiting the maximum possible SHRM volume or frontal area), which may make the regression slope steeper than without such a ceiling. Standardized multiple regression analysis showed that for each of the well-defined SHRM dimensional features, the correlation between 3 or 6 mm ETDRS diameter and BCVA was stronger than that between 1 mm ETDRS diameter and BCVA. This standardized regression analysis was performed to compare the partial correlations between BCVA and SHRM subtypes across different dimensions and ETDRS diameters. For well-defined SHRM (-0.14 to -0.34), the partial correlations were again stronger than those for undefined SHRM (-0.03 to -0.11). Within well-defined SHRM, frontal area had the highest partial correlation within each ETDRS diameter, followed by maximum height and volume. Within undefined SHRM, the partial correlations were closer together, with no clear preference for either dimension or ETDRS area.

[0091] In this example, it was demonstrated in the untreated eye that: (1) when volume assessment was performed using the Biomarker Quantification System 100, the undefined SHRM was the dominant subtype; (2) as indicated by the Biomarker Quantification System 100, the defined SHRM had a higher internal reflectance than the undefined SHRM; (3) compared to other MNV categories, the classic and major classic MNVs showed a greater amount of SHRM and a higher proportion of defined SHRM; (4) both SHRM subtypes were negatively associated with BCVA, with a more pronounced effect on defined SHRM; and (5) frontal area had the most adverse effect on BCVA, with the largest effect observed in the 3 and 6 mm ETDRS rings.

[0092] In some embodiments, the biomarker quantification system 100 and / or process 400 allow for the study of SHRMs and their subtypes in a three-dimensional manner, enabling not only point measurements but also frontal area and volume measurements. This is a technical advantage over other systems and processes that limit SHRM measurements to single scans or foveal scans, or assess only the height and width of the SHRM. In addition to being able to measure frontal area and volume, the biomarker quantification system 100 and / or process 400 also allow for the restriction of measurements to specific ETDRS regions of interest.

[0093] In the example analysis, and when comparing the overall means across various dimensions between undefined and defined SHRM, higher baseline values ​​and fewer zero values ​​were observed for undefined SHRM compared to defined SHRM. Although fibrosis was an exclusion criterion for TENAYA / LUCERNE at enrollment and was detected only in six eyes by the reading center, a high prevalence of defined SHRM was observed, which is often considered a substitute for fibrosis if it persists. Therefore, a much lower prevalence of defined SHRM was expected in this example analysis. However, the large variation in fibrosis prevalence reported in the literature may have occurred because this characteristic is traditionally graded based on color and FA properties, which are more difficult to grade consistently. Nevertheless, even defined SHRM may regress or decrease with treatment, and in some embodiments, the biomarker quantification system 100 was configured to study SHRM evolution patterns in the example dataset, which should, in turn, allow for a better understanding of dynamic changes within subtypes.

[0094] Furthermore, as demonstrated in the example analysis, a well-defined SHRM exhibits higher internal reflectance compared to an undefined SHRM, and the higher reflectance and well-defined boundaries of the SHRM may indicate the presence of a fibrotic process. Therefore, the biomarker quantification system 100 and / or process 400 can use image input 102 associated with the subject to indicate whether the subject has a fibrotic process or is at high risk of having a fibrotic process.

[0095] The example analysis also applied quantitative assessments of undefined and defined SHRMs to differentiate MNV categories. The analysis showed that SHRM size was largest in both subtypes of classic and major classic MNVs, consistent with the OCT correlation of SHRM representing the subretinal focal neovascularization complex. At least one study using DL segmentation of the overall SHRM on Cirrus images from the HARBOR trial reached the same conclusion, with MNV type 2 lesions showing the largest SHRM volume, and other studies have previously shown that high-reflectivity material (HRM) is most frequently observed in type 2 and mixed choroidal neovascularization (CNV). The example analysis further showed that the ratio of defined to undefined SHRMs was higher in eyes with classic or major classic MNVs than in other MNV categories before treatment initiation. This supports the hypothesis that in classic MNVs with a subretinal neovascularization complex, the natural process involves partial transformation into more mature components. In contrast, this occurs less frequently in other MNV categories because SHRM represents more fluid and regressible components in these subtypes. The example analysis differs from other analyses (based on SHRM that has not been volumetally quantified, eyes were classified using a human grader, and no attempt was made to determine the relative proportions of defined or undefined SHRMs) and indicates that at baseline, the largest proportion of eyes with defined SHRMs were classified as CNV type 1, polypoid choroidal vascular lesions (PCV), or a mixed type.

[0096] In the example analysis, both undefined and defined SHRM were negatively associated with baseline BCVA. The example analysis revealed that defined SHRM had a more pronounced adverse effect on visual acuity on a per-unit-volume basis, supporting the idea that fibrotic transformation may occur within defined SHRM. Nevertheless, undefined SHRM also showed a negative association with visual acuity as it is a biomarker for exudation and active MNV. Importantly, the overall visual burden associated with undefined SHRM is considerable, although its regression coefficient is small considering the larger overall undefined SHRM volume in untreated eyes. However, longitudinally, undefined SHRM plays a negligible role, as complete regression upon initiation of anti-angiogenic therapy is possible. The example analysis showed that all SHRM dimensions (height, frontal area, and volume) were negatively associated with baseline BCVA, contrasting with other studies where only width, or height, was found to have an adverse effect on visual acuity. This contrast is likely a result of the more robust and unbiased measurement properties of the biomarker quantification system 100 compared to human readings, some of which even use measurements based solely on median thresholds from individually selected scans rather than continuous thresholds. In our example analysis, frontal area was the measure with the greatest adverse effect on visual acuity, consistent with the view that it is most correlated with the level of lesion severity, as it determines the extent to which photoreceptors are affected. On a per-unit basis, measurements within 1 mm of the center were most strongly associated with BCVA, as this area covers the fovea, as expected from a biological perspective. In some embodiments, the effects of 3 mm and 6 mm diameters were even greater when compared using standardized regression. This is a surprising finding, but on the one hand, it may reflect that BCVA is not merely a measure of a patient's ability to form eccentrically fixed central visual acuity. On the other hand, due to the inclusion criteria of the TENAYA / LUCERNE trial, this dataset contains only central lesions, not just extrafoveal lesions that are associated with better visual acuity but are captured only within larger ETDRS rings.

[0097] Example analyses demonstrate several significant advantages of characterizing SHRMs in nAMD. First, the Biomarker Quantification System 100 offers technological improvements and effectiveness in its application of deep learning to differentiate between undefined and defined SHRMs. By distinguishing these two subtypes of SHRM and labeling their reflectance, the Biomarker Quantification System 100 provides a more nuanced understanding of their corresponding roles in visual impairment and potential fibrosis development. Automated analysis allows for consistent, objective, and scalable analysis, overcoming the limitations of manual segmentation, which is both time-consuming and subject to inter-rater variability. In some embodiments, continuous quantification via the Biomarker Quantification System 100 using volumetric and three-dimensional measurements instead of traditional binary readings or threshold measurements allows for a more comprehensive and accurate representation of SHRMs, which is crucial for a full understanding of their clinical significance. Furthermore, the Biomarker Quantification System 100's ability to continuously quantify SHRM reflectance enables the labeling of subtle differences in the composition of SHRMs that may be clinically relevant and potentially helpful in differentiating underlying biological components. Secondly, the example analysis utilized large and well-characterized datasets from the TENAYA and LUCERNE Phase 3 clinical trials, which enhanced the generality of our findings and provided strong statistical power for our analysis.

[0098] The results of this example analysis deepen our understanding of the distribution, characteristics, and associations of different SHRM subtypes with visual function in untreated nAMD eyes. The results also provide further evidence that undefined SHRMs are biomarkers of exudative and active MNVs, and that defined SHRMs are biomarkers of fibrotic changes within MNVs. The ability of the Biomarker Quantification System 100 to characterize SHRMs allows for studies on large phase 3 datasets and overcomes the limitations of human binary or semi-quantitative grading.

[0099] In some embodiments, the biomarker quantification system 100 and / or process 400 are applied to longitudinal data to further establish the dynamics and correlations of undefined (UD) and defined (WD) SHRMs. For example, and in one embodiment, a biomarker quantification system including a U-NET model was used in a retrospective study to examine the dynamics of UD and WD SHRMs under anti-angiogenic therapy and their association with best-corrected visual acuity (BCVA). In this example, the retinal segmentation model 110 was trained on AVENUE (NCT02484690) phase 2 Spectralis SD-OCT scans to segment SHRM subtypes and applied in a treatment-agnostic manner to all available 97-line Spectralis SD-OCT volumetric scans from the TENAYA / LUCERNE (NCT03823287 / NCT03823300) phase 3 trial. In this example, 1,499 B scans spanning 334 volumes were pixel-level annotated, and scans with SHRM annotations were subsequently classified as SHRM-UD or SHRM-WD based on SHRM annotations and depiction of the neurosensory retina (≤50% or >50% SHRM depictability, respectively). The U-Net model was trained over 150 epochs with 1,337 scans. The Dice scores (sensitivity, specificity) for SHRM-UD and SHRM-WD segmentation on the retain set (162 scans) were 0.711 (0.860, 0.998) and 0.839 (0.976, 0.999), respectively. SHRM-UD and SHRM-WD volumes were calculated within ETDRS 3-mm diameter circles for morphology and ETDRS 1-mm diameter circles for BCVA analysis. For the summary statistics, median and interquartile range (IQR) were used due to the skewed distribution. Least square regression was used to analyze the association with BCVA. Regarding the example results, on day 1 (n eyes = 1095), the median volumes (Q1–Q3) of SHRM-UD and SHRM-WD were 72.5 (3.9–237.4) nl and 2.2 (0–20.3) nl, respectively.By week 12 (n eyes = 1072), SHRM-UD dropped to 0 (0-10) nl and remained at a low level until week 48 (0, 0-2.2 nl; n eyes = 970) and week 112 (0, 0-1.9 nl; n eyes = 923), while SHRM-WD remained relatively stable (week 12: 1, 0-11.3 nl; week 48: 0.9, 0-11.8 nl; week 112: 0.9, 0-10.0 nl). On day 1, an additional 100 nl of SHRM volume was associated with a 6.0 letter (95% CI: –7.7; –4.4; P<.001) decrease in BCVA for SHRM-UD and an 18.3 letter (95% CI: –22.6; –14.0; P<.001) decrease for SHRM-WD. In these example results, the decrease in BCVA was exacerbated over time for each SHRM subtype volume increase. Regarding the example conclusions based on the example results, a trained biomarker quantification system 100 robustly segmented and quantified SHRM-UD and SHRM-WD in nAMD. SHRM-UD showed larger initial volume and treatment-induced changes, while SHRM-WD had a stronger negative BCVA association. This supports the hypothesis that SHRM-UD represents an active neovascular membrane, while SHRM-WD indicates a fibrotic process. In some embodiments, the results of this example analysis can enable automated, volumetric, and objective comparisons of different mechanisms of action in future nAMD treatments, and, due to a more comprehensive understanding of the disease, can provide a reference for better treatment strategies.

[0100] III. Computer-implemented systems Figure 8 is a block diagram of a computer system according to various embodiments. The computer system 800 may be an example of an embodiment of the computing platform 106 described in Figure 1 above. In one or more embodiments, the computer system 800 may include a bus 802 or other communication mechanism for transmitting information, and a processor 804 coupled to the bus 802 for processing information. In various embodiments, the computer system 800 may also include a memory (which may be random access memory (RAM) 806 or other dynamic storage device) coupled to the bus 802 for determining instructions to be executed by the processor 804. The memory may also be used to store temporary variables or other intermediate information during the execution of instructions to be executed by the processor 804. In various embodiments, the computer system 800 may further include a read-only memory (ROM) 808 or other static storage device coupled to the bus 802 for storing static information and instructions for the processor 804. A storage device 810 (such as a magnetic disk or optical disk) may be provided and coupled to the bus 802 for storing information and instructions.

[0101] In various embodiments, computer system 800 may be coupled to display 812 (such as a cathode ray tube (CRT) or liquid crystal display (LCD)) via bus 802 for displaying information to a computer user. Input device 814, including alphanumeric keys and other keys, may be coupled to bus 802 for transmitting information and command selections to processor 804. Another type of user input device is cursor control 816 (such as a mouse, joystick, trackball, gesture input device, gaze-based input device, or cursor direction keys) for transmitting directional information and command selections to processor 804 and for controlling cursor movement on display 812. Input device 814 typically has two degrees of freedom on two axes (a first axis (e.g., x) and a second axis (e.g., y)), allowing the device to specify a position in a plane. However, it should be understood that input device 814 that allows three-dimensional (e.g., x, y, and z) cursor movement is also contemplated herein.

[0102] Consistent with certain implementations of this teaching, results may be provided by computer system 800 in response to processor 804 executing one or more sequences of one or more instructions contained in RAM 806. Such instructions may be read into RAM 806 from another computer-readable medium or computer-readable storage medium, such as storage device 810. Execution of the sequence of instructions contained in RAM 806 may cause processor 804 to perform the processes described herein. Alternatively, this teaching may be implemented using hardwired circuitry in place of or in combination with software instructions. Therefore, implementation of this teaching is not limited to any particular combination of hardware circuitry and software.

[0103] As used herein, the terms "computer-readable medium" (e.g., data repository, data storage device, storage device, data storage unit, etc.) or "computer-readable storage medium" refer to any medium that participates in providing instructions to processor 804 for execution. Such media can take many forms, including but not limited to non-volatile media, volatile media, and transmission media. Examples of non-volatile media may include, but are not limited to, optical discs, solid-state drives, and magnetic disks (such as storage device 810). Examples of volatile media may include, but are not limited to, dynamic memory, such as RAM 806. Examples of transmission media may include, but are not limited to, coaxial cables, copper wires, and optical fibers, including the wires constituting bus 802.

[0104] Common forms of computer-readable media include, for example, floppy disks, collapsible disks, hard disks, magnetic tapes or any other magnetic media; CD-ROMs, any other optical media; punched cards, paper tapes, any other physical media with a perforated pattern; RAM, PROM and EPROM, FLASH-EPROM, any other memory chips or cartridges; or any other tangible media that a computer can read.

[0105] In addition to computer-readable media, instructions or data may also be provided as signals on a transmission medium included in a communication device or system to provide a sequence of one or more instructions to a processor 804 of a computer system 800 for execution. For example, a communication device may include a transceiver having signals indicating instructions and data. The instructions and data are configured to cause one or more processors to perform the functions outlined in this disclosure. Representative examples of data communication transmission connections may include, but are not limited to, telephone modem connections, wide area networks (WANs), local area networks (LANs), infrared data connections, NFC connections, optical communication connections, etc.

[0106] It should be recognized that the methods, flowcharts, diagrams and accompanying disclosures described herein can be implemented using computer system 800 as a standalone device or on a distributed network (such as a cloud computing network) sharing computer processing resources.

[0107] Depending on the application, the methods described herein can be implemented in various ways. For example, these methods can be implemented in hardware, firmware, software, or any combination thereof. For hardware implementation, the processing unit can be implemented within one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, electronic devices, other electronic units designed to perform the functions described herein, or combinations thereof.

[0108] In various embodiments, the methods of this teaching can be implemented as firmware and / or software programs and applications written in conventional programming languages ​​such as C, C++, Python, etc. If implemented as firmware and / or software, the embodiments described herein can be implemented on a non-transitory computer-readable medium, wherein a stored program enables a computer to perform the methods described above. It should be understood that the various engines described herein can be provided on a computer system, such as computer system 800, wherein processor 804 will perform analysis and determination provided by these engines based on instructions provided by any memory component RAM 806, ROM 808, or storage device 810, or combinations thereof, and user input provided via input device 814.

[0109] IV. Example Definitions and Context This disclosure is not limited to these exemplary embodiments and applications, nor to the manner in which they are operated or as described herein. Furthermore, the accompanying drawings may show simplified or partial views, and the dimensions of elements in the drawings may be exaggerated or otherwise disproportionate.

[0110] Unless otherwise defined, scientific and technical terms used in conjunction with the teachings described herein shall have the meanings commonly understood by one of ordinary skill in the art. Furthermore, unless the context requires otherwise, singular terms shall include plural forms, and plural terms shall include singular forms. Generally, this document describes nomenclature and techniques used in conjunction with chemistry, biochemistry, molecular biology, pharmacology, and toxicology that are well-known and commonly used in the art.

[0111] Furthermore, when the terms “on,” “attached to,” “connected to,” “coupled to,” or similar words are used herein, an element (e.g., a component, material, layer, substrate, etc.) may be “on another element,” “attached to another element,” “connected to another element,” or “coupled to another element,” regardless of whether an element is directly on, directly attached to, directly connected to, or directly coupled to another element, or whether one or more intermediate elements exist between the element and the other. Additionally, when referring to a list of elements (e.g., elements a, b, c), such references are intended to include any single element listed, any combination of fewer than all listed elements, and / or any combination of all listed elements. The section divisions in this specification are for ease of review only and do not limit any combination of elements discussed.

[0112] The term "subject" may refer to a participant in a clinical trial, a person receiving treatment, a person receiving anticancer therapy, a person undergoing remission or recovery monitoring, a person undergoing preventative health analysis (e.g., due to their medical history), or any other person of interest or patient. In various contexts, "subject" and "patient" may be used interchangeably in this document.

[0113] As used herein, “substantially” means sufficient to achieve the intended purpose. Therefore, the term “substantially” allows for minor, insignificant variations relative to the absolute or ideal state, dimensions, measurements, results, etc., as expected by one of ordinary skill in the art, without significantly affecting overall performance. When used relative to numerical values ​​or parameters or categories that can be expressed as numerical values, “substantially” means within ten percent.

[0114] As used in this article, the term "about" when referring to numerical values ​​or parameters or categories that can be expressed as numerical values ​​means within ten percent of the numerical value. For example, "about 50" means a value in the range of 45 to 55, including the extreme values.

[0115] The term "ones" means more than one.

[0116] As used in this article, the term "multiple" can refer to 2, 3, 4, 5, 6, 7, 8, 9, 10 or more.

[0117] As used in this article, the term "set" refers to one or more items. For example, a set of items includes one or more items.

[0118] As used herein, the phrase “at least one of…” when used with a list of items indicates that different combinations of one or more of the listed items may be used, and possibly only one item from the list. An item can be a specific object, thing, step, action, process, or category. In other words, “at least one of…” means any combination or number of items that may be used in the list, but not all items in the list may be used. For example, but not limitingly, “at least one of item A, item B, or item C” means item A; item A and item B; item B; item A, item B, and item C; item B and item C; or item A and item C. In some cases, “at least one of item A, item B, or item C” means, but is not limited to, two of item A, one of item B, and ten of item C; four of item B and seven of item C; or some other suitable combination.

[0119] As used in this article, a “model” may include one or more algorithms, one or more mathematical techniques, one or more machine learning algorithms, or a combination thereof.

[0120] As used in this article, "machine learning" can encompass the practice of using algorithms to analyze data, learn from it, and then make determinations or predictions about something in the world. Machine learning can use algorithms that can learn from data without relying on rule-based programming. Deep learning can be a form of machine learning.

[0121] As used herein, "artificial neural network" or "neural network" (NN) can refer to a mathematical algorithm or computational model that simulates a set of interconnected artificial neurons, processing information based on connectionist computational methods. A neural network (also called a neural mesh) can use one or more layers of nonlinear units to predict the output of a received input. In addition to the output layer, some neural networks may also include one or more hidden layers. The output of each hidden layer can be used as the input to the next layer in the network, i.e., the next hidden layer or output layer. Each layer of the network generates an output from the received input based on the current values ​​of its corresponding parameter set. In various embodiments, the reference to "neural network" can refer to one or more neural networks.

[0122] Neural networks can process information in two ways: in training mode when the network is being trained, and in inference (or prediction) mode when the network puts what it has learned into practice. Neural networks can learn through a feedback process (e.g., backpropagation), which allows the network to adjust the weights of individual nodes in intermediate hidden layers (modifying their behavior) so that the output matches the output of the training data. In other words, a neural network can learn and eventually learn how to obtain the correct output by being fed training data (learning instances), even if it is presented with a new range or set of inputs. Neural networks can include, for example, but not limited to, feedforward neural networks (FNNs), recurrent neural networks (RNNs), modular neural networks (MNNs), convolutional neural networks (CNNs), residual neural networks (ResNets), neural networks for ordinary differential equations (neural-ODEs), U-Nets, fully convolutional networks (FCNs), stacked FCNs, stacked FCNs using multi-channel learning, squeezed and excited embedded neural networks, MobileNets, or at least one of another type of neural network.

[0123] As used in this article, "deep learning" can refer to the use of multi-layered artificial neural networks to automatically learn representations from input data (such as images, videos, text, etc.) without human-provided knowledge, in order to provide highly accurate predictions in tasks such as object detection / labeling, speech recognition, language translation, etc.

[0124] As used herein, “target retinal element” can refer to a retinal element (e.g., a retinal layer element or a retinal pathology element) whose reflectance is quantified using the methods and procedures described herein.

[0125] V. Description of Example Implementations Example 1: A method comprising: receiving a set of optical coherence tomography (OCT) volumes of a subject's retina, wherein each OCT volume in the set of OCT volumes corresponds to a different time point in a set of time points, and includes a set of two-dimensional OCT B scans; generating a set of feature images for each OCT volume in the set of OCT volumes, wherein multiple ophthalmic feature markers are used to visually identify multiple ophthalmic features in the set of feature images; wherein the multiple ophthalmic feature markers assign different groups of pixels to each of the multiple ophthalmic features; and wherein the multiple ophthalmic features include a target retinal feature, a retinal pigment epithelium (RPE) layer, and a vitreous body; and determining a reflectance score of the target retinal feature identified in each OCT volume in the set of OCT volumes using the multiple ophthalmic feature markers in each OCT volume, thereby forming a set of reflectance scores. The foregoing embodiments may include one or more of the following embodiments, whether individually or in combination: Example 2: According to the method of Example 1, generating the set of feature images for each OCT volume in the set of OCT volumes includes: generating a first segmentation image for each OCT B scan in the set of OCT B scans for each OCT volume in the set of OCT volumes using a retinal segmentation model; wherein the first segmentation image includes the plurality of ophthalmic feature markers, the plurality of ophthalmic feature markers identifying the plurality of ophthalmic features by assigning pixel groups to each of the plurality of ophthalmic features; generating a first feature image using the first segmentation image that at least identifies the target retinal feature; and generating a second feature image using the first segmentation image that at least identifies the RPE layer and other ophthalmic features.

[0126] Example 3: According to the method of Example 2, generating the second element image includes: generating a second segmentation image using the first segmentation image and an ophthalmic element algorithm to identify the additional ophthalmic element, wherein the second segmentation image includes the plurality of ophthalmic element markers, the plurality of ophthalmic element markers identifying the additional ophthalmic element by assigning pixel groups to the additional ophthalmic element.

[0127] Example 4: According to the method of Example 3, generating the second segmented image includes using one or more ophthalmic features marked on the first segmented image to mark the additional ophthalmic features.

[0128] Example 5: The method according to Example 4, wherein the additional ophthalmic element includes the vitreous body.

[0129] Example 6: According to the method of Example 5, wherein the one or more ophthalmic elements marked on the first segmented image include an internal limiting membrane (ILM) layer; and wherein the vitreous body is marked using the ILM layer marked on the first segmented image.

[0130] Example 7: The method according to any one of Examples 1 to 6 further includes post-processing the set of feature images by performing a pixel erosion operation to remove multiple pixels from each feature image in the set of feature images from various directions, removing multiple pixels from each feature image in the set of feature images from one or more directions, or a combination thereof.

[0131] Example 8: The method according to any one of Examples 1 to 7, wherein determining the reflectance score of the target retinal element further comprises: quantifying the intensity measure of the target retinal element; and adjusting the intensity measure of the target retinal element to take into account the first reflectance of the vitreous body and the second reflectance of the RPE layer.

[0132] Example 9: According to the method of Example 1, the set of reflectance scores of the target retinal element includes the baseline reflectance scores of the baseline time points in the set of time points; and the method further includes selecting a threshold for separating reflectance scores indicating low reflectance and reflectance scores indicating high reflectance based on analyzing the distribution of multiple baseline reflectance scores calculated for multiple subjects.

[0133] Example 10: According to the method of Example 9, the set of time points includes the baseline time point and a first time point after the baseline time point; and the set of reflectance scores includes a baseline reflectance score associated with the baseline time point and a first reflectance score associated with the first time point.

[0134] Example 11: According to the method of Example 9, the set of time points includes the baseline time point and a first time point after the baseline time point; and the set of reflectance scores includes the reflectance change between the baseline reflectance score associated with the baseline time point and the first reflectance score associated with the first time point.

[0135] Example 12: The method according to Example 1, wherein the set of reflectance scores is a set of biomarkers for disease progression.

[0136] Example 13: The method according to Example 1, wherein the set of reflectance scores is a set of biomarkers for at least the treatment outcome or treatment response to a selected treatment of the retina.

[0137] Example 14: The method according to Example 1 further includes using the set of reflectance scores to generate an output; wherein the output includes an indication of the classification of the retinal symptom, disease progression, treatment response, or treatment outcome at a time point in the set of time points.

[0138] Example 15: The method according to any one of Examples 1 to 14, further comprising modifying, redesigning, replicating, or evaluating a clinical trial in response to the set of reflectivity scores.

[0139] Example 16: The method according to any one of Examples 1 to 15, wherein the target retinal element is a retinal layer element.

[0140] Example 17: The method according to any one of Examples 1 to 15, wherein the target retinal element is a retinal pathological element.

[0141] Example 18: The method according to any one of Examples 1 to 15, wherein the target retinal element is a subtype of a retinal layer element or a subtype of a retinal pathological element.

[0142] Example 19: The method according to Example 17, wherein the retinal pathological element is a fluid associated with pigment epithelial detachment (PED).

[0143] Example 20: The method according to Example 17 or Example 18, wherein the retinal pathological element is subretinal high reflectance material (SHRM).

[0144] Example 21: According to the method described in Example 1, the determination of the reflectance score of the target retinal element marked in each of the set of OCT volumes occurs automatically and in response to generating the set of element images for each of the set of OCT volumes.

[0145] Example 22: The method according to Example 1 includes: using the plurality of ophthalmic elements marked by the plurality of ophthalmic element markers in the OCT volume, determining the reflectance score of the target retinal element marked in the OCT volume in a set of OCT volumes at a given time point, wherein the OCT volume corresponds to the given time point.

[0146] Example 23: A system comprising: one or more data processors; and a non-transitory computer-readable storage medium containing instructions that, when executed on the one or more data processors, cause the one or more data processors to perform part or all of the methods disclosed in Examples 1 to 22.

[0147] Example 24: A computer program product tangibly embodied in a non-transitory machine-readable storage medium, comprising instructions configured to cause one or more data processors to perform part or all of the methods disclosed in Examples 1 to 22.

[0148] VI. Other considerations Although this teaching has been described in conjunction with various embodiments, it is not intended to be limited to such embodiments. Rather, this teaching encompasses various alternatives, modifications, and equivalents that will be understood by those skilled in the art.

[0149] For example, the flowcharts and block diagrams described above illustrate the architecture, functionality, and / or operation of various implementations of methods and systems. Each box in a flowchart or block diagram may represent a portion of a module, segment, function, operation, or step, or a combination thereof. In some alternative implementations of the embodiments, one or more functions labeled within a box may occur in a different order than those labeled in the diagram. For example, in some cases, two boxes shown consecutively may execute substantially simultaneously. In other cases, the boxes may execute in reverse order. Furthermore, in some cases, one or more boxes may be added to replace or supplement one or more other boxes in the flowchart or block diagram.

[0150] Therefore, in describing the various embodiments, the specification may have presented the methods and / or processes as a specific sequence of steps. However, if the method or process does not depend on the specific sequence of steps described herein, the method or process should not be limited to the specific sequence of steps listed, and those skilled in the art will readily understand that these sequences may be different and still remain within the spirit and scope of the various embodiments.

Claims

1. A method comprising: A set of optical coherence tomography (OCT) volumes of the subject's retina are received, wherein each OCT volume in the set of OCT volumes corresponds to a different time point in a set of time points, and includes a set of two-dimensional OCT B scans; A set of feature images is generated for each OCT volume in the set of OCT volumes, wherein the set of feature images uses multiple ophthalmic feature markers to visually identify multiple ophthalmic features; The plurality of ophthalmic feature markers assign different groups of pixels to each of the plurality of ophthalmic features; and The aforementioned multiple ophthalmic elements include target retinal elements, the retinal pigment epithelium (RPE) layer, and the vitreous body; as well as The reflectance score of the target retinal element marked in each of the multiple ophthalmic element markers in each of the set of OCT volumes is determined by using the multiple ophthalmic element markers in each of the set of OCT volumes, thereby forming a set of reflectance scores.

2. The method of claim 1, wherein generating the set of feature images for each OCT volume in the set of OCT volumes comprises: A first segmentation image is generated by the retinal segmentation model for each OCT B scan in the set of OCT B scans for each OCT volume in the set of OCT volumes; The first segmented image includes the plurality of ophthalmic feature markers, which identify the plurality of ophthalmic features by assigning pixel groups to each of the plurality of ophthalmic features; The first segmented image is used to generate a first feature image that at least identifies the target retinal features; as well as A second feature image is generated using the first segmented image, which identifies at least the RPE layer and other ophthalmic features.

3. The method of claim 2, wherein generating the second element image comprises: A second segmentation image is generated using the first segmentation image and the ophthalmic feature algorithm to identify the additional ophthalmic features. The second segmented image includes the plurality of ophthalmic feature markers, which identify the additional ophthalmic features by assigning pixel groups to them.

4. The method of claim 3, wherein generating the second segmented image includes using one or more ophthalmic features marked on the first segmented image to mark the additional ophthalmic features.

5. The method of claim 4, wherein the additional ophthalmic element comprises the vitreous body.

6. The method according to claim 5, The one or more ophthalmic elements marked on the first segmented image include the internal limiting membrane (ILM) layer; and The vitreous body is indicated using the ILM layer marked on the first segmented image.

7. The method according to any one of claims 1 to 6, further comprising post-processing the set of feature images by: performing a pixel erosion operation to remove a plurality of pixels from each feature image in the set of feature images from various directions, removing a plurality of pixels from each feature image in the set of feature images from one or more directions, or a combination thereof.

8. The method according to any one of claims 1 to 7, wherein determining the reflectance score of the target retinal element further comprises: The intensity measurement of the target retinal element is quantified; as well as The intensity measure of the target retinal element is adjusted to take into account the first reflectivity of the vitreous body and the second reflectivity of the RPE layer.

9. The method according to claim 1, The set of reflectance scores for the target retinal element includes the baseline reflectance scores for the baseline time point within the set of time points; and The method further includes selecting a threshold for separating reflectance scores that indicate low reflectance from reflectance scores that indicate high reflectance based on an analysis of the distribution of multiple baseline reflectance scores calculated for multiple subjects.

10. The method according to claim 9, The set of time points includes the baseline time point and a first time point after the baseline time point; and The set of reflectance scores includes a baseline reflectance score associated with the baseline time point and a first reflectance score associated with the first time point.

11. The method according to claim 9, The set of time points includes the baseline time point and a first time point after the baseline time point; and The set of reflectance scores includes the reflectance variation between the baseline reflectance score associated with the baseline time point and the first reflectance score associated with the first time point.

12. The method of claim 1, wherein the set of reflectance scores is a set of biomarkers for disease progression.

13. The method of claim 1, wherein the set of reflectance scores is a set of biomarkers for at least the treatment outcome or treatment response to a selected treatment of the retina.

14. The method of claim 1, further comprising using the set of reflectance scores to generate an output; The output includes an indication of the retinal symptom, disease progression, treatment response, or treatment outcome at a time point in the set of time points.

15. The method according to any one of claims 1 to 14, further comprising altering, redesigning, replicating, or evaluating a clinical trial in response to the set of reflectivity scores.

16. The method according to any one of claims 1 to 15, wherein the target retinal element is a retinal layer element.

17. The method according to any one of claims 1 to 15, wherein the target retinal element is a retinal pathological element.

18. The method according to any one of claims 1 to 15, wherein the target retinal element is a subtype of a retinal layer element or a subtype of a retinal pathological element.

19. The method of claim 17, wherein the retinal pathological element is a fluid associated with pigment epithelial detachment (PED).

20. The method of claim 17 or claim 18, wherein the retinal pathological element is subretinal high reflectance material (SHRM).

21. The method of claim 1, wherein the determination of the reflectance score of the target retinal element marked in each of the set of OCT volumes occurs automatically and in response to generating an image of the set of elements for each of the set of OCT volumes.

22. The method of claim 1, further comprising: Using the multiple ophthalmic elements marked by the multiple ophthalmic element markers in the OCT volume of the set of OCT volumes, determine the reflectance score of the target retinal element marked in the OCT volume at a given time point. The OCT volume corresponds to the given time point.

23. A system comprising: One or more data processors; as well as A non-transitory computer-readable storage medium comprising instructions that, when executed on the one or more data processors, cause the one or more data processors to perform part or all of the methods disclosed in claims 1 to 22.

24. A computer program product tangibly embodied in a non-transitory machine-readable storage medium, the non-transitory machine-readable storage medium comprising instructions configured to cause one or more data processors to perform part or all of the methods disclosed in claims 1 to 22.

Images