MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641091623 A) filed by Srm University- Ap on July 28, 2026, for System And Method For Entity-Aware Medical Image Captioning From Radiology Images.
Inventors include Dr. Shaik Rafi; P. Drutika, R; R. Srija; Ch. Surya Susmitha; and J. Sahasra.
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: The present invention provides a medical image captioning system (100) is a specialized framework designed to automatically generate precise textual descriptions directly from radiology images through an interconnected, five-part architecture. Initially, the image encoder (102) receives an input radiology image to extract its key visual features, while a tokenizer (104) simultaneously processes medical text to generate foundational token embeddings. Connected to the tokenizer, an entity extraction module (106) identifies critical medical entities, specifically isolating anatomical regions and clinical findings. Subsequently, a multimodal fusion module (108) integrates these visual features, token embeddings, and entity embeddings into a unified, fused multimodal representation. Finally, a decoder (110) analyzes this rich, fused representation to generate the final, detailed clinical caption, streamlining the radiology reporting workflow. (FIGURE 1)
Disclaimer: Curated by HT Syndication.
