MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641071369 A) filed by Sona College Of Technology on June 09, 2026, for Ai-Based System And Method For Reducing Hallucinations In Clinical Text Summarization.
Inventors include Hemalatha K; Kesavan K; Mohan Babu P; Mohit M; Vaishnavi Dali A; and Jayaprakash D.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: ABSTRACT The present invention relates to an AI-based system and method for reducing hallucinations in clinical text summarization generated from electronic health records (EHRs), discharge summaries, and physician notes. The system comprises a domain-specific fine-tuning module for adapting large language models using clinical corpora, a source-aware generation engine for grounding each generated statement to verifiable source text, and a multi-stage verification layer for token- level, sentence-level, and summary-level factual consistency checks. A factual consistency and source attribution module provides confidence scores, explicit citations, and traceable evidence mapping for each summary statement. A human-in-the-loop validation interface enables clinicians to review flagged or low-confidence outputs before final release. The invention prevents intrinsic and extrinsic hallucinations, improves interpretability, enhances clinician trust, and delivers reliable, clinically safe summaries suitable for real-world healthcare deployment.
Disclaimer: Curated by HT Syndication.
