Joint Imbalance Adaptation for Radiology Report Generation

Anonymous

Abstract:

Radiology report generation, predicting text descriptions for radiological images, may face critical challenges due to data imbalance -- medical tokens appear less frequently than regular tokens, and normal labels of images may not equal to abnormal ones. However, existing studies mainly consider label imbalance without mitigating other factors, such as token imbalance.
In this study, we jointly consider two imbalance factors, label and token, determining distributions of radiology images and language, two fundamental modalities of the generation task.
We propose a Joint Imbalance Adaptation (JIMA) model to promote task robustness by leveraging token and label imbalance. Experiments on two standard evaluation data (IU X-ray and MIMIC-CXR) by automatic and human evaluations demonstrate our significant improvements over current state-of-the-art models. We conduct extensive ablation and case analyses to examine and present dual imbalance effects on the radiology report generation robustness.
While data imbalance remains challenging, our approach opens new task directions and shows promising results.

Paper Type: long

Research Area: Generation

Contribution Types: Data analysis

Languages Studied: English

Consent To Share Submission Details: On behalf of all authors, we agree to the terms above to share our submission details.