Machine learning-based optoretinography using phase-sensitive optical coherence tomography

A method of training a machine learning model for classifying a temporal phase signal obtained from optical coherence tomography (OCT) images produced for optoretinography using at least one processor, the method comprising:

• obtaining a series of OCT images produced by a phase-sensitive OCT system for optoretinography with respect to an outer retina, including a stimulus sub-series of OCT images produced in response to a visual stimulus to the outer retina;
• determining, for each OCT image of the series of OCT images and for each of a plurality of individual pixels in a target layer of the OCT image, a phase signal of the individual pixel in the target layer of the OCT image with respect to a reference layer of the OCT image to obtain, for each series of corresponding individual pixels of a plurality of series of corresponding individual pixels in the target layer across the series of OCT images, a temporal phase signal of the series of corresponding individual pixels, resulting in a set of temporal phase signals for the plurality of series of corresponding individual pixels, the target layer comprising a plurality of outer retinal bands;
• projecting a plurality of temporal phase signals of the set of temporal phase signals to a plurality of feature points in a feature space;
• labelling each of the plurality of feature points in the feature space;
• and
• training the machine learning model based on the plurality of labelled feature points in the feature space.