HDC-Net: Hierarchical Decoupled Convolution Network for Brain Tumor Segmentation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Accurate brain tumor segmentation from MRI is vital for clinical decisions and surgical planning.
• Deep learning models show promise but often require high computational resources.
• Existing methods face challenges with tumor diversity, complex boundaries, and computational cost.

Purpose of the Study:

• To develop a computationally efficient yet accurate method for brain tumor segmentation.
• To address the limitations of current deep learning models in resource-constrained clinical environments.

Main Methods:

• A novel Hierarchical Decoupled Convolution (HDC) module was developed to replace traditional 3D convolutions.
• The HDC module efficiently captures multi-scale and multi-view spatial contexts.
• The proposed HDC-Net segments 3D volumetric images in a single pass, reducing complexity.

Main Results:

• The HDC-Net demonstrated competitive accuracy compared to state-of-the-art methods on BraTS 2018 and 2017 datasets.
• The method achieved significantly reduced computational complexity and overhead.
• The pseudo-3D approach offers a balance between performance and efficiency.

Conclusions:

• HDC-Net provides an accurate and computationally efficient solution for brain tumor segmentation.
• The proposed HDC module is effective in exploring spatial contexts with high efficiency.
• This lightweight model is well-suited for clinical applications with limited computational budgets.