Boundary-Aware Gradient Operator Network for Medical Image Segmentation.

Area of Science:

• Medical Image Analysis
• Computer-Aided Diagnosis
• Deep Learning

Background:

• Convolutional Neural Networks (CNNs) are vital for medical image segmentation but struggle with blurred boundaries and loss of fine structural details due to down-sampling.
• Current CNNs lack specificity in feature extraction as kernels are optimized without explicit gradient information.

Purpose of the Study:

• To introduce a novel Boundary-Aware Gradient Operator Network (BG-Net) for enhanced medical image segmentation.
• To improve the extraction of boundary and fine structural features in medical images.

Main Methods:

• Developed Gradient Convolution (GConv) to extract gradient and boundary features.
• Incorporated a Boundary-Aware Mechanism (BAM) to capture remote dependencies and global context.
• Utilized a multi-modal fusion mechanism to integrate shallow spatial details with deep contextual information.

Main Results:

• BG-Net demonstrated superior performance across eight diverse medical imaging datasets.
• The proposed GConv and BAM modules effectively enhanced boundary perception and feature specificity.
• The network successfully captured both global dependencies and low-level spatial details.

Conclusions:

• BG-Net offers a significant advancement in medical image segmentation, particularly for tasks requiring accurate boundary delineation.
• The network's design addresses limitations of traditional CNNs in handling subtle image features.
• BG-Net shows strong potential for improving computer-aided diagnosis systems.