A system and method for use with optical coherence tomography (OCT) data to identify a target pathology extracts multiple pathology-characteristic images from the OCT data. The extracted pathology-characteristic images may include a mixture of OCT structural images (including retinal layer thickness information) and OCT angiography images. Optionally, other pathology-characteristic images and data maps (mapped to corresponding positions in the OCT data), such as fundus images and visual field test maps may be accessed as additional pathology-characteristic images. Each pathology-characteristic image defines a different image channel (e.g., “color channel”) per pixel in a composite, channel-coded image, which is then used to train a neural network to search for the target pathology in OCT data. The trained neural network may then receive new composite, channel-coded image and identify/segment the target pathology within the new channel-coded image.
1 . A method of analyzing optical coherence tomography (OCT) data, comprising:
collecting the OCT data with an OCT system, the OCT data including a plurality of A-scans;
using an electronic data processor:
extracting a series of metrics from each individual A-scan;
defining a set of images based on the extracted metrics, each image defining pathology-characteristic data;
defining a multi-channel image based on the set of images;
submitting the multi-channel image into a machine learning model trained to identify one or more pathologies based on the pathology-characteristic data; and
displaying or storing for future processing the identified pathology,
wherein the pathology is geographic atrophy,
wherein the series of metrics include two or more of sub-RPE (retinal pigment epithelium) reflectivity, inner RPE reflectivity, retinal thickness, choriocapillaris flow, and optical attenuation coefficient (OAC); and
wherein each metric defines a separate corresponding channel per pixel of the multi-channel image.
2 . The method of claim 1 , wherein the collected OCT data is volume data and each image in the set of images is a two-dimensional image.
3 . The method of claim 1 , wherein a pixel channel within the multiple channel image specifies a relative distance from a pixel's corresponding A-scan to a predefined ophthalmic landmark.
4 . The method of claim 3 , wherein the pixels are based on distances from each A-scan to the fovea.
5 . The method of claim 1 , further including accessing additional imaging data of one or more additional imaging modalities different than OCT, wherein the multi-channel image includes one or more image channels respectively based on the one or more additional imaging modalities.
6 . The method of claim 5 , wherein the one or more additional images are based on at least one of a fundus image, autofluorescence image, fluorescein angiography image, OCT angiography image, or visual field test map.
7 . The method of claim 5 , wherein the machine learning model is further trained using non-image data.
8 . The method of claim 7 , wherein the non-image data includes patient demographic data.
9 . The method of claim 1 , further including:
acquiring visual field functional data, the multi-channel image including at least one image channel based on the visual field functional data.
10 . The method of claim 1 , further including:
sorting the extracted metrics from each A-scan into corresponding metric groups with a one-to-one correspondence.
11 . The method of claim 10 , wherein each channel of the multi-channel image is based on a corresponding metric group.
12 . The method of claim 1 , wherein the extracted metrics from each A-scan are associated with the same pathology type.
13 . The method of claim 1 , wherein:
the OCT data includes OCT structural data and OCT angiography (OCTA) flow data; the series of metrics include OCT-based metrics extracted from the OCT structural data and OCTA-based metrics extracted from the OCTA flow data;
the set of images includes OCT-based images based on the OCT-based metrics and OCTA-based images based on OCTA-based metrics; and the multi-channel image is based on the OCT-based images and OCTA-based images.
14 . The method of claim 1 , wherein the machine learning model identifies a region of the multi-channel image where the pathology is present based on a combination of pathology-characteristic data provided by the individual channels of each pixel of the multi-channel image.
15 . The method of claim 14 , wherein each A-scan is mapped to a pixel in the multi-channel image, and the identified region where the pathology is present is mapped to the collected OCT data.
16 . The method of claim 1 , wherein the machine learning model is embodied by a neural network.
17 . The method of claim 16 , wherein the neural network is a U-Net type architecture.
18 . The method of claim 1 , wherein each channel in the multi-channel image is a color channel.