U-Net-Based Medical Image Segmentation.

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Deep learning, particularly U-Net, excels in medical image segmentation.
• U-Net's scalable architecture and accuracy for small targets have led to over 2500 academic citations.
• Continuous development of U-Net variants addresses increasing performance demands in medical imaging.

Purpose of the Study:

• To summarize and categorize U-Net based medical image segmentation technologies.
• To review structural innovations and efficiency of U-Net variants.
• To provide a reference for future research by introducing common loss functions, evaluation parameters, and modules.

Main Methods:

• Literature review and categorization of U-Net variants in medical image segmentation.
• Analysis of U-Net architecture modifications, focusing on structure and innovation.
• Compilation of commonly used loss functions, evaluation metrics, and modules.

Main Results:

• A comprehensive overview of U-Net variants for medical image segmentation is presented.
• Key structural innovations and efficiency improvements in U-Net architectures are highlighted.
• Commonly applied methodologies, loss functions, and evaluation parameters are cataloged.

Conclusions:

• This review consolidates current U-Net based approaches in medical image segmentation.
• It serves as a valuable reference for researchers developing novel segmentation techniques.
• The findings will guide future advancements in deep learning for medical imaging analysis.