STCS-Net: a medical image segmentation network that fully utilizes multi-scale information.

Area of Science:

• Medical Image Analysis
• Deep Learning in Healthcare
• Computer Vision

Background:

• Deep learning significantly advances medical image segmentation.
• Current research often prioritizes encoder optimization.
• Decoders are crucial for refining image details and leveraging diverse information.

Purpose of the Study:

• To propose STCS-Net, a novel medical image segmentation architecture.
• To enhance feature extraction accuracy and inter-layer information interaction.
• To improve the performance of medical image segmentation models.

Main Methods:

• Developed STCS-Net with a specialized decoder for multi-scale filtering and correction.
• Introduced an information enhancement module in skip connections.
• Conducted comprehensive evaluations on ISIC2016, ISIC2018, and Lung datasets.

Main Results:

• STCS-Net demonstrated superior performance across multiple medical imaging datasets.
• Achieved outstanding accuracy and parameter efficiency compared to existing methods.
• Ablation studies confirmed the effectiveness of the proposed decoder and skip connection module.

Conclusions:

• STCS-Net offers a novel and effective approach to medical image segmentation.
• The proposed architecture enhances feature extraction and inter-layer communication.
• This research provides valuable insights for future medical image processing and analysis.