# MetaChest: generalized few-shot learning of pathologies from chest X-rays

## TL;DR

This paper introduces MetaChest, a large dataset of chest X-rays, and evaluates few-shot learning methods for pathology classification, showing that standard transfer learning performs well.

## Contribution

The paper introduces MetaChest, a large-scale chest X-ray dataset, and evaluates few-shot learning methods in a generalized setting for medical image classification.

## Key findings

- Increasing the number of classes and examples per episode improves classification performance.
- Transfer learning outperformed ProtoNet in few-shot multi-label tasks.
- Higher-resolution images boost accuracy but require more computation.

## Abstract

The limited availability of annotated data presents a major challenge in applying deep learning methods to medical image analysis. Few-shot learning methods aim to recognize new classes from only a few labeled examples. These methods are typically investigated within a standard few-shot learning paradigm, in which all classes in a task are new. However, medical applications, such as pathology classification from chest X-rays, often require learning new classes while simultaneously leveraging the knowledge of previously known ones, a scenario more closely aligned with generalized few-shot classification. Despite its practical relevance, few-shot learning has rarely been investigated in this context. This study presents MetaChest, a large-scale dataset of 479,215 chest X-rays collected from four public databases. It includes a meta-set partition specifically designed for standard few-shot classification, as well as an algorithm for generating multi-label episodes. Extensive experiments were conducted to evaluate both the standard transfer learning (TL) approach and an extension of ProtoNet across a wide range of few-shot multi-label classification tasks. The results indicate that increasing the number of classes per episode and the number of training examples per class improves the classification performance. Notably, the TL approach consistently outperformed the ProtoNet extension, even though it was not specifically tailored for few-shot learning. Furthermore, higher-resolution images improved the accuracy at the cost of additional computation, whereas efficient model architectures achieved performances comparable to larger models with significantly reduced resource requirements.