Modality specific U-Net variants for biomedical image segmentation: a survey.

Area of Science:

• Deep learning
• Medical image analysis
• Biomedical engineering

Background:

• U-Net architectures are pivotal in biomedical image segmentation for automated region identification.
• Deep learning models like CNNs, ResNets, and GANs underpin these advancements.
• U-Net based approaches achieve state-of-the-art performance in computer-aided diagnosis systems.

Purpose of the Study:

• To present the success of U-Net architectures in biomedical image segmentation.
• To comprehensively analyze U-Net variants through inter-modality and intra-modality categorization.
• To highlight U-Net's contribution to COVID-19 diagnosis and identify future research directions.

Main Methods:

• Describing the U-Net framework.
• Categorizing U-Net variants based on inter-modality and intra-modality analysis.
• Reviewing applications in various medical imaging modalities and diseases (e.g., brain tumors, lung cancer, COVID-19).

Main Results:

• U-Net variants demonstrate significant success in automated biomedical image segmentation.
• Analysis provides insights into challenges and solutions across different imaging modalities.
• U-Net frameworks have shown notable contributions in diagnosing severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2).

Conclusions:

• U-Net based approaches are highly effective for computer-aided diagnosis and disease detection.
• Further research into U-Net variants can enhance biomedical image segmentation accuracy and clinical utility.
• Addressing segmentation challenges will unlock new avenues for deep learning in medical imaging.