Summary The current technology provides a framework for a fully-automated deep-learning-based diabetic retinopathy screening that relies solely on optical coherence tomography (OCT) and its angiography, without the need for fundus photography or time-intensive clinical evaluation.
Technology Overview Diabetic retinopathy (DR) is a leading cause of preventable blindness globally but current diagnosis methods rely on fundus photography, which has low sensitivity and specificity. Optical coherence tomography (OCT) offers improved sensitivity and has been used to supplement fundus photography, but to-date OCT has not been used as stand-alone modality for diabetic retinopathy diagnosis. The current technology provides a framework for a deep-learning based diabetic retinopathy screening using OCT and OCT-angiography as the only imaging modality. This platform allows for fully-automated diagnosis of diabetic retinopathy without the need for clinician intervention or interpretation, which could provide more timely diagnoses and reduce dependency on trained specialists. In addition, by using volumetric OCT and OCT-angiography as inputs, the framework can avoid unstable preprocessing and use all of the features latent in the full data volume. Preliminary validation of the platform in a large cohort of over 300 patients showed specialist-level accuracy in detecting diabetic retinopathy. Ultimately, this software could allow for increased implementation of OCT systems in clinical practice by providing a faster and more accurate screening process for diabetic retinopathy
Publications Pengxiao Z, et al., “A Diabetic Retinopathy Classification Framework Based on Deep-Learning Analysis of OCT Angiography.” Trans. Vis. Sci. Tech. 2022;11(7):10. Link Pengxiao Z, et al., “DcardNet: Diabetic Retinopathy Classification at Multiple Levels Based on Structural and Angiographic Optical Coherence Tomography.” IEEE. Trans. Biomed. Eng. 2021;68(6):1859-1870. Link
Licensing Opportunity This technology is available for licensing.