Intellectual Property India Publishes Patent Application for 'A Multi-Agent Conversational Diagnostic System For Breast Ultrasound Analysis And Explainable Clinical Decision Support' Filed by Vellore Institute Of Technology

MUMBAI, India, May 29 -- Intellectual Property India has published a patent application (202641061211 A) filed by Vellore Institute Of Technology, Vellore, Tamil Nadu, on May 14, for 'a multi-agent conversational diagnostic system for breast ultrasound analysis and explainable clinical decision support.'

Inventor(s) include Dr Padmavathy T; Sharanya Acharya; Yami Sunil; and Poshika Reddy.

The application for the patent was published on May 29, under issue no. 22/2026.

According to the abstract released by the Intellectual Property India: "The present invention relates to a multi-agent conversational diagnostic system and method for breast ultrasound analysis and explainable clinical decision support is disclosed, wherein a coordinator agent receives and validates ultrasound images from a user interface and orchestrates interaction among specialized agents. An imaging agent processes the ultrasound image using a convolutional neural network, preferably a fine-tuned DenseNet model, to generate lesion class predictions, classification confidence scores, continuous malignancy probability values, and descriptive morphological characteristics derived from activation-responsive regions. A reasoning agent performs grounded semantic inference on structured imaging outputs and optional clinical text using a diagnostic knowledge graph, further leveraging a large language model to produce ranked diagnostic interpretations, reasoning confidence scores, and clinically interpretable justification narratives. An adaptive multimodal fusion engine integrates imaging-based and reasoning-based outputs using dynamically computed weights and concordance thresholds to generate a final diagnosis for clinician evaluation. The system further supports structured data exchange via API-based microservices, maintains audit logs of inter-agent communication, derives semantic tokens from high-activation image regions, and incorporates fallback degradation mechanisms to ensure robust and reliable diagnostic performance."

Disclaimer: Curated by HT Syndication.