Brain Tumor MRI Classification Using a Novel Deep Residual and Regional CNN.

Area of Science:

• Medical Imaging and Diagnostics
• Artificial Intelligence in Medicine
• Computational Neuroscience

Background:

• Brain tumor classification is critical for effective clinical diagnosis and treatment planning.
• Deep learning models offer potential for brain tumor classification but face challenges with tumor complexity and diversity.
• Accurate classification aids in determining appropriate therapeutic strategies and patient management.

Purpose of the Study:

• To introduce Res-BRNet, a novel deep residual and region-based convolutional neural network (CNN) for brain tumor classification using MRI.
• To enhance the extraction of crucial tumor features, including homogeneity, heterogeneity, and boundary information.
• To improve the accuracy and efficiency of automated brain tumor classification.

Main Methods:

• Developed Res-BRNet, a CNN architecture integrating regional and boundary-based operations within spatial and residual blocks.
• Spatial blocks were designed to extract homogeneity, heterogeneity, and boundary features.
• Residual blocks were utilized to capture local and global texture variations in MRI scans.

Main Results:

• Res-BRNet achieved high performance on a diverse brain tumor dataset (Br35H, figshare).
• The model demonstrated excellent accuracy (98.22%), sensitivity (0.9811), F1-score (0.9841), and precision (0.9822).
• Res-BRNet outperformed standard CNN models in brain tumor classification tasks.

Conclusions:

• Res-BRNet shows significant promise as a tool for accurate brain tumor classification from MRI scans.
• The proposed architecture effectively addresses the complexities of brain tumor imaging.
• This approach has the potential to enhance clinical diagnosis and streamline treatment planning.