Efficient Medical Image Assessment via Self-supervised Learning

TL;DR

This paper introduces EXAMINE, a novel self-supervised learning-based method for assessing the quality of unlabeled medical images to optimize data labeling efforts.

Contribution

The paper proposes a new data assessment strategy using SSL embeddings and singular value analysis, enabling efficient selection of valuable data without prior labels.

Findings

Effective in ranking data quality for medical images

Reduces labeling costs by selecting high-value data

Validated on pathology dataset with promising results

Abstract

High-performance deep learning methods typically rely on large annotated training datasets, which are difficult to obtain in many clinical applications due to the high cost of medical image labeling. Existing data assessment methods commonly require knowing the labels in advance, which are not feasible to achieve our goal of 'knowing which data to label.' To this end, we formulate and propose a novel and efficient data assessment strategy, EXponentiAl Marginal sINgular valuE (EXAMINE) score, to rank the quality of unlabeled medical image data based on their useful latent representations extracted via Self-supervised Learning (SSL) networks. Motivated by theoretical implication of SSL embedding space, we leverage a Masked Autoencoder for feature extraction. Furthermore, we evaluate data quality based on the marginal change of the largest singular value after excluding the data point in the dataset. We conduct extensive experiments on a pathology dataset. Our results indicate the effectiveness and efficiency of our proposed methods for selecting the most valuable data to label.