CK-ATTnet: Medical image segmentation network based on convolutional kernel attention.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Medicine
• Computer Vision

Background:

• Medical image segmentation is crucial for diagnosis and treatment.
• Convolutional Neural Networks (CNNs) face limitations in feature extraction.
• Transformer models offer advancements but have high computational costs and structural rigidity.

Purpose of the Study:

• To develop an efficient and adaptable medical image segmentation model.
• To address the limitations of computational resources in clinical settings.
• To propose a novel attention mechanism for convolutional kernels.

Main Methods:

• Introduced CK-ATTnet, a network utilizing a convolutional kernel attention mechanism.
• Employed depthwise separable convolution for the attention mechanism.
• Applied attention directly to the convolutional kernel for feature extraction.

Main Results:

• CK-ATTnet enhances local feature acquisition and fine-grained feature extraction.
• The model demonstrates superior segmentation performance compared to other CNN and Transformer models.
• CK-ATTnet requires fewer learning parameters, making it suitable for clinical equipment.

Conclusions:

• CK-ATTnet offers a promising approach for medical image segmentation with improved accuracy and efficiency.
• The novel application of attention mechanisms to convolutional kernels presents a significant advancement.
• The model's reduced parameter count and strong performance indicate broad clinical applicability.