MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202641064328 A) filed by Amrita Vishwa Vidyapeetham on May 21, 2026, for Fundus Image Processing System With Trainable Transformation.
Inventors include Pillai, Gopal S.; Mickhayel, Merin Dickson; Jacob, Rahul Binu; Mj, Anjana; and Ranjit, Amarnath Krishnamritam.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: The present disclosure provides a fundus image processing system (100) and a method (400) for a non-invasive concurrent prediction of a set of conditions from one or more fundus images of a single subject. The system (100) includes a localiser (106) that sets a radius of a disc region equal to a shorter side of a bounding rectangle of a largest bright connected region of a green channel, a trainable transformation (108) including a sequence of strided residual blocks (210, 214) and a sequence of token- mixing blocks (212, 216), an aggregator (110) that aggregates a set of pre-activation value sets through a mean and a bounded squashing function, and a spatial map computer (112) that computes an element-wise product within the disc region. Unlike conventional approaches relying on biochemical assays and invasive investigations, the system (100) enables non-invasive screening of multiple systemic conditions concurrently from retinal fundus images.
Disclaimer: Curated by HT Syndication.
