A novel approach to brain tumor detection using K-Means++, SGLDM, ResNet50, and synthetic data augmentation

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Brain tumors present complex diagnostic challenges requiring accurate, non-invasive detection methods.
• Early identification of brain tumors via MRI is critical for effective treatment planning and improved patient outcomes.

Purpose of the Study:

• To enhance early brain tumor detection in MRI images using advanced deep learning techniques.
• To identify the most effective deep learning model for classifying brain tumors from MRI data, thereby improving diagnostic accuracy and reliability.

Main Methods:

• A novel brain tumor classification method integrating K-means++ segmentation, Spatial Gray Level Dependence Matrix (SGLDM) feature extraction, and ResNet50 classification.
• Synthetic data augmentation was utilized to improve model robustness.
• Grad-CAM was employed for enhanced interpretability by visualizing influential regions in MRI scans.

Main Results:

• The proposed method demonstrated superior performance in accuracy, sensitivity, and specificity on the Br35H::BrainTumorDetection2020 dataset compared to existing state-of-the-art approaches.
• The evaluation confirmed the method's effectiveness in achieving higher precision for brain tumor identification and classification from MRI data.

Conclusions:

• The developed deep learning approach offers robust and accurate classification of brain tumors from MRI images, outperforming current methods.
• Enhanced sensitivity and specificity contribute to optimized clinical decision-making and patient care in neuro-oncology.