Leveraging Summary Guidance on Medical Report Summarization

TL;DR

This paper introduces three large medical report datasets and proposes a summary guidance method using BART to improve abstractive summarization, achieving better ROUGE and BERTScore results.

Contribution

The study provides new large-scale medical report datasets and a novel guidance technique that enhances summarization quality beyond existing models.

Findings

Guidelines improve ROUGE scores.

Guidance enhances BERTScore.

Method outperforms T5-large.

Abstract

This study presents three deidentified large medical text datasets, named DISCHARGE, ECHO and RADIOLOGY, which contain 50K, 16K and 378K pairs of report and summary that are derived from MIMIC-III, respectively. We implement convincing baselines of automated abstractive summarization on the proposed datasets with pre-trained encoder-decoder language models, including BERT2BERT, T5-large and BART. Further, based on the BART model, we leverage the sampled summaries from the train set as prior knowledge guidance, for encoding additional contextual representations of the guidance with the encoder and enhancing the decoding representations in the decoder. The experimental results confirm the improvement of ROUGE scores and BERTScore made by the proposed method, outperforming the larger model T5-large.