Multi-instance Learning as Downstream Task of Self-Supervised Learning-based Pre-trained Model

TL;DR

This paper proposes using a self-supervised pre-trained model as a downstream task for deep multi-instance learning, significantly improving accuracy and F1 scores in brain hematoma CT image classification, especially with large instance bags.

Contribution

It introduces a novel approach of leveraging self-supervised pre-trained models to enhance multi-instance learning performance on large instance datasets.

Findings

Achieved 5-13% accuracy improvement.

Achieved 40-55% F1 score improvement.

Effective in addressing spurious correlation issues.

Abstract

In deep multi-instance learning, the number of applicable instances depends on the data set. In histopathology images, deep learning multi-instance learners usually assume there are hundreds to thousands instances in a bag. However, when the number of instances in a bag increases to 256 in brain hematoma CT, learning becomes extremely difficult. In this paper, we address this drawback. To overcome this problem, we propose using a pre-trained model with self-supervised learning for the multi-instance learner as a downstream task. With this method, even when the original target task suffers from the spurious correlation problem, we show improvements of 5% to 13% in accuracy and 40% to 55% in the F1 measure for the hypodensity marker classification of brain hematoma CT.