Review on Deep Learning Methodologies in Medical Image Restoration and Segmentation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Medical image processing is crucial for accurate diagnosis and therapy.
• Image restoration and segmentation are key tasks, essential for clinical decision-making.
• Conventional methods face challenges with large datasets and complex image features.

Approach:

• This review focuses on deep learning methodologies, particularly convolutional neural network architectures.
• It examines their application in medical image restoration and segmentation.
• Key deep learning models and their performance metrics are analyzed.

Key Points:

• For image restoration, TLR-CNN and Stat-CNN demonstrate effectiveness in noise/artifact suppression and enhancing image quality (PSNR).
• In segmentation, LCP net achieves high accuracy (98.12% Dice score) for cell contour segmentation.
• The 3D FCNN model is identified as optimal for brain tumor segmentation.

Conclusions:

• Deep learning provides a powerful alternative for medical image restoration and segmentation.
• The effectiveness of deep learning is particularly evident with the increasing size of medical imaging datasets.
• Specific CNN architectures offer state-of-the-art performance for distinct medical image processing tasks.