A Survey on Deep Learning and Explainability for Automatic Report Generation from Medical Images

TL;DR

This survey reviews deep learning methods for automatic medical report generation from images, highlighting recent advances, challenges in evaluation metrics, and the importance of explainability in clinical AI applications.

Contribution

It provides a comprehensive overview of datasets, architectures, explainability, and evaluation metrics in the field, identifying key challenges and future directions.

Findings

Current evaluation metrics are inadequate for medical correctness.

Deep neural networks are increasingly used for report generation.

Explainability remains a significant challenge in clinical AI.

Abstract

Every year physicians face an increasing demand of image-based diagnosis from patients, a problem that can be addressed with recent artificial intelligence methods. In this context, we survey works in the area of automatic report generation from medical images, with emphasis on methods using deep neural networks, with respect to: (1) Datasets, (2) Architecture Design, (3) Explainability and (4) Evaluation Metrics. Our survey identifies interesting developments, but also remaining challenges. Among them, the current evaluation of generated reports is especially weak, since it mostly relies on traditional Natural Language Processing (NLP) metrics, which do not accurately capture medical correctness.