MRF-IUNet: A Multiresolution Fusion Brain Tumor Segmentation Network Based on Improved Inception U-Net.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Accurate segmentation of MRI brain tumors is crucial for clinical diagnosis and treatment planning.
• Existing methods may struggle with detailed feature recognition and information loss during downsampling.

Purpose of the Study:

• To propose a novel multiresolution fusion MRI brain tumor segmentation algorithm, MRF-IUNet, based on an improved inception U-Net.
• To enhance the segmentation accuracy of brain tumors by improving feature extraction and fusion capabilities.

Main Methods:

• Developed MRF-IUNet by replacing standard convolutional modules with inception modules to increase network width and depth.
• Incorporated atrous convolutions within inception modules to expand receptive fields and preserve detailed information.
• Implemented multiresolution feature fusion modules between the encoder and decoder to integrate semantic and spatial features.

Main Results:

• The MRF-IUNet achieved a Dice similarity coefficient (DSC) of 0.94 for enhanced tumor area, 0.83 for the whole tumor area, and 0.93 for the tumor core area on the BraTS dataset.
• The proposed method demonstrated improved recognition and segmentation of local detail features.
• Experimental results indicate a significant improvement in overall segmentation accuracy.

Conclusions:

• MRF-IUNet effectively segments MRI brain tumors by leveraging multiresolution fusion and improved inception U-Net architecture.
• The algorithm shows promising results for clinical applications, aiding in more precise brain tumor diagnosis and treatment.
• The integration of inception modules and multiresolution fusion modules enhances the model's ability to capture multi-scale features and detailed information.