A physical examination report intelligent generation method based on LLM and a knowledge base

Abstract

This invention provides an intelligent method for generating physical examination reports based on LLM and a knowledge base, aiming to better adapt LLM technology to the medical field of physical examinations, which lies between specialist diagnosis and routine health monitoring. The core of this method lies in constructing a hybrid architecture where a knowledge base and LLM work collaboratively to address the ambiguity in interpreting physical examination data caused by its "ambiguous intermediate zone." The specific method includes: identifying all abnormalities based on an indicator range knowledge base; performing precise matching and rule-based reasoning through a clinical physical examination diagnostic knowledge base, generating conclusions for unmatched abnormalities according to three preset rule categories; querying a health advice knowledge base and generating personalized suggestions through LLM; and performing global review and comprehensive verification by LLM to ensure the completeness and consistency of the conclusions. This invention combines a deterministic knowledge base with generative LLM reasoning, significantly improving the accuracy, personalization, and clinical interpretability of interpreting physical examination abnormalities, realizing the intelligent transformation from fragmented physical examination data to actionable health insights.

Images