Self-supervised learning methods and applications in medical imaging analysis: a survey.

Area of Science:

• Medical Imaging Analysis
• Machine Learning
• Computer Vision

Background:

• High-quality annotated medical imaging datasets are scarce, hindering machine learning advancements in medical image analysis.
• Self-supervised learning (SSL) offers a solution by learning data representations without human annotation.

Purpose of the Study:

• To review state-of-the-art self-supervised learning approaches for image data.
• To focus on the applications of SSL in medical imaging analysis.
• To categorize recent SSL methods applicable to medical imaging.

Main Methods:

• Review of recent self-supervised learning methods from computer vision.
• Categorization of methods into predictive, generative, and contrastive approaches.
• Analysis of 40 recent research papers in self-supervised learning for medical imaging.

Main Results:

• Identified key self-supervised learning strategies relevant to medical imaging.
• Highlighted recent innovations and trends in the field.
• Provided a comprehensive overview of current research.

Conclusions:

• Self-supervised learning is a promising solution for overcoming data scarcity in medical imaging.
• Future research directions in SSL for medical imaging analysis are identified.