S 2 A-RConvNet: standalone self-attention enabled deep learning model for brain tumor classification with MRI images

Area of Science:

• Medical imaging analysis
• Artificial intelligence in oncology
• Deep learning for diagnostics

Background:

• Brain tumors (BT) are a leading cause of mortality worldwide.
• Accurate BT classification is crucial for timely diagnosis and treatment.
• Existing classification models suffer from poor accuracy, high computational cost, and overfitting.

Purpose of the Study:

• To develop an accurate and efficient model for brain tumor classification.
• To address the limitations of conventional brain tumor classification approaches.
• To improve patient outcomes through enhanced diagnostic capabilities.

Main Methods:

• Developed the Standalone Self-Attention based Repeated Convolutional Network (S²A-RConvNet) model.
• Integrated a Standalone Self-Attention (S²A) module to enhance focus on tumor regions.
• Utilized Structured ResNet Attention Gray-level (SRAG) features for improved training efficiency and reduced complexity.

Main Results:

• The S²A-RConvNet model achieved high performance on the BraTS 2021 dataset.
• Achieved sensitivity of 97.61%, precision of 98.71%, F1-Score of 98.16%, specificity of 98.43%, and accuracy of 97.98% with 90% training data.
• Demonstrated improved accuracy and reduced computational complexity compared to conventional methods.

Conclusions:

• The S²A-RConvNet model offers a promising solution for accurate and efficient brain tumor classification.
• The proposed model effectively overcomes the limitations of existing approaches.
• This advancement has the potential to significantly improve patient survival rates through better diagnostics.