RaTEScore: A Metric for Radiology Report Generation

TL;DR

RaTEScore is a new entity-aware metric designed to evaluate AI-generated radiology reports by focusing on medical entities and their clinical relevance, showing better alignment with human judgment than existing metrics.

Contribution

The paper introduces RaTEScore, a novel evaluation metric for radiology reports that leverages a specialized medical NER dataset and entity embedding similarity, improving assessment accuracy.

Findings

RaTEScore aligns more closely with human preferences than existing metrics.

The developed RaTE-NER dataset enables effective extraction of medical entities.

RaTEScore demonstrates robustness against medical synonyms and negation expressions.

Abstract

This paper introduces a novel, entity-aware metric, termed as Radiological Report (Text) Evaluation (RaTEScore), to assess the quality of medical reports generated by AI models. RaTEScore emphasizes crucial medical entities such as diagnostic outcomes and anatomical details, and is robust against complex medical synonyms and sensitive to negation expressions. Technically, we developed a comprehensive medical NER dataset, RaTE-NER, and trained an NER model specifically for this purpose. This model enables the decomposition of complex radiological reports into constituent medical entities. The metric itself is derived by comparing the similarity of entity embeddings, obtained from a language model, based on their types and relevance to clinical significance. Our evaluations demonstrate that RaTEScore aligns more closely with human preference than existing metrics, validated both on established public benchmarks and our newly proposed RaTE-Eval benchmark.