MUMBAI, India, July 30 -- Intellectual Property India has published a patent application (202641087414 A) filed by Prof. Shreevatsa D S; Ms. Inchara K P; Dr. Rajanishree M; Dr. Saranya S N; Dr. Swetha M D; Dr. Madhumala R B; Dayananda Sagar Academy Of Technology And Management; B. M. S. College Of Engineering; and Bnmit, Bengaluru on July 17, 2026, for An Ai-Driven Medical Report Generation And Summarization System Using Multi-Modal Patient Data And Grounded Large Language Models.
Inventors include Prof. Shreevatsa D S; Ms. Inchara K P; Dr. Rajanishree M; Dr. Saranya S N; Dr. Swetha M D; Dr. Madhumala R B; Dayananda Sagar Academy Of Technology And Management; B. M. S. College Of Engineering; and Bnmit, Bengaluru.
The application for the patent was published on July 24, 2026, under issue no. 30/2026.
Abstract: The invention relates to an AI-driven medical report generation and summarization framework that overcomes the limitations of single-modality clinical documentation systems and of ungrounded clinical language models of the prior art. The system comprises a medical-imaging input (101), an electronic-health-record ingest (102), a physician-notes input (103) and a laboratory-results input (104) which jointly receive a patient's multi-modal data; a multi-modal data normalization and privacy layer (105); a vision transformer encoder (106), a clinical language model encoder (107) and a structured-data encoder (108) operating in parallel; a cross-modal fusion and context assembler (109) that aligns imaging patches, textual tokens and structured records at a fine-grained per-token, per-patch, per-record granularity; a medical knowledge base and guidelines store (110); a structured report generator (111) that emits a multi-section full-length medical report; a multi-level summarization engine (112) that emits at least three audience-specific summaries — a section-level summary for physician review, a plain-language patient-facing summary and a specialist hand-off summary; a hallucination-and-grounding verifier (113) that examines each clinical claim of each generated output for factual support; a citation-and-evidence attacher (114) that renders source pointers within the verified output; and a clinician review interface (115). The framework operates locally on hospital-premises infrastructure, emits its outputs within a sub-second latency, and reduces physician documentation burden while preserving factuality and full citation traceability suitable for clinical and medico-legal review.
Disclaimer: Curated by HT Syndication.
