Self-Supervised Learning as a Means To Reduce the Need for Labeled Data in Medical Image Analysis

Abstract

One of the largest problems in medical image processing is the lack of annotated data. Labeling medical images often requires highly trained experts and can be a time-consuming process. In this paper, we evaluate a method of reducing the need for labeled data in medical image object detection by using self-supervised neural network pretraining. We use a dataset of chest X-ray images with bounding box labels for 13 different classes of anomalies. The networks are pretrained on a percentage of the dataset without labels and then finetuned on the rest of the dataset. We show that it is possible to achieve similar performance to a fully supervised model in terms of mean average precision and accuracy with only 60% of the labeled data. We also show that it is possible to increase the maximum performance of a fully-supervised model by adding a self-supervised pretraining step, and this effect can be observed with even a small amount of unlabeled data for pretraining.

Influence of Intraoral Scanners, Operators, and Data Processing on Dimensional Accuracy of Dental Casts for Unsupervised Clinical Machine Learning: An In Vitro Comparative Study

“…To summarize the fundamental concept of supervised AI in healthcare [ 9 ], AI models require clinicians and operators to classify regions of interest in the training dataset of an AI model, which is then applied to test a dataset that the AI model has not seen before [ 10 , 11 ]. Recent advances in medical AI have introduced the possibility of unsupervised or self-supervised deep learning, which can learn from 3D scans without relying on human-labeled datasets [ 12 ]. Historically, human-labeled datasets in dentistry were a source of low reliability and high biases [ 2 ].…”