Towards more precise automatic analysis: a systematic review of deep learning-based multi-organ segmentation.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Computational Anatomy

Background:

• Accurate multi-organ segmentation is crucial for computer-aided diagnosis, surgical navigation, and radiation therapy.
• Deep learning methods have surpassed traditional approaches in automatic medical image segmentation.
• A systematic review is needed to summarize recent advancements in deep learning for multi-organ segmentation.

Purpose of the Study:

• To systematically review and summarize the latest research on deep learning-based multi-organ segmentation methods.
• To analyze existing datasets and categorize current segmentation approaches.
• To identify current trends and future directions in the field.

Main Methods:

• A systematic literature search was conducted on Google Scholar for papers from January 1, 2016, to December 31, 2023, using keywords 'multi-organ segmentation' and 'deep learning'.
• PRISMA guidelines were followed for study selection, resulting in 195 included studies.
• Methods were categorized into fully supervised, weakly supervised, and semi-supervised approaches based on label requirements.

Main Results:

• The review provides an overview and analysis of public datasets used for multi-organ segmentation.
• Existing methods are categorized and their segmentation accuracy achievements are summarized.
• Current trends in multi-organ segmentation are outlined and discussed.

Conclusions:

• Deep learning has become the state-of-the-art for multi-organ segmentation, offering significant improvements over traditional methods.
• The review highlights the importance of datasets and different supervised learning strategies in achieving high segmentation accuracy.
• Understanding current trends is essential for future research and development in medical image analysis.