FM-Unet: Biomedical image segmentation based on feedback mechanism Unet

Area of Science:

• Computer Vision
• Deep Learning
• Medical Image Analysis

Background:

• Deep learning has significantly advanced medical image segmentation.
• The U-Net architecture is a foundational model in this field.
• Existing U-Net improvements primarily focus on backward propagation, neglecting forward propagation and information integration.

Purpose of the Study:

• To propose a novel Feedback mechanism U-Net (FM-Unet) model.
• To enhance information fusion in medical image segmentation networks.
• To address encoder information loss and decoder information shortage.

Main Methods:

• Introduced feedback paths to both the encoder and decoder of the U-Net architecture.
• Implemented a mechanism for fusing information from subsequent steps into current encoder and decoder stages.
• Evaluated the model on two public medical image datasets.

Main Results:

• The FM-Unet model effectively fuses information across network stages.
• Addressed challenges of encoder information loss and decoder information shortage.
• Demonstrated satisfactory performance on experimental datasets with moderate network parameters.

Conclusions:

• The proposed FM-Unet model offers an effective approach to medical image segmentation.
• Feedback mechanisms enhance information integration and network performance.
• FM-Unet provides a promising alternative for deep learning-based medical image analysis.