PL-Net: progressive learning network for medical image segmentation.

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Deep convolutional neural networks (CNNs) excel in medical image analysis.
• Current U-Net based methods often neglect the fusion of coarse-grained and fine-grained semantic information.

Purpose of the Study:

• To propose a novel 2D medical image segmentation framework, the Progressive Learning Network (PL-Net).
• To improve the complementation and fusion of multi-granularity semantic information in medical image segmentation.

Main Methods:

• Introduced Internal Progressive Learning (IPL) for hierarchical feature extraction.
• Developed External Progressive Learning (EPL) for staged training optimization.
• PL-Net integrates IPL and EPL to capture semantic information from coarse to fine granularity.

Main Results:

• PL-Net demonstrated competitive segmentation performance across five medical image datasets.
• The proposed method effectively fuses coarse-grained and fine-grained semantic features.
• Achieved state-of-the-art results without introducing additional learnable parameters compared to U-Net variants.

Conclusions:

• PL-Net offers an effective approach for 2D medical image segmentation.
• The framework successfully addresses the limitations of existing methods in semantic information fusion.
• PL-Net provides a parameter-efficient solution for enhanced medical image analysis.