A Deep Learning Based Intelligent Decision Support System for Automatic Detection of Brain Tumor

Area of Science:

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

Background:

• Brain tumors (BT) are a significant cause of mortality, necessitating advanced diagnostic tools.
• Computer-aided diagnostic systems (CAD), particularly deep learning, are revolutionizing medical image analysis.
• Convolution Neural Networks (CNNs) excel at extracting features from medical images for disease detection.

Purpose of the Study:

• To develop and evaluate deep learning models for the automatic detection of brain tumors from medical images.
• To compare the performance of CNNs developed from scratch against transfer learning models (VGG-16, VGG-19, LeNet-5).
• To establish an intelligent decision support system for brain tumor diagnosis.

Main Methods:

• Utilized a deep learning approach to extract features from brain images for BT detection.
• Developed and tested CNN models, including VGG-16, VGG-19, and LeNet-5, for brain tumor classification.
• Employed data augmentation to increase dataset size and hyperparameter tuning for model optimization.

Main Results:

• VGG models demonstrated superior performance, achieving 99.24% accuracy, 99% precision, 99% recall, 99% specificity, and 99% F1-score.
• The proposed models outperformed existing state-of-the-art methods in accuracy, sensitivity, specificity, and F1-score.
• Comparative analysis confirmed the reliability and effectiveness of the developed models.

Conclusions:

• Deep learning models, especially VGG variants, show high efficacy in detecting brain tumors from medical images.
• The developed system serves as a reliable tool to aid medical practitioners in diagnosing brain tumors.
• This research highlights the potential of advanced AI in improving diagnostic accuracy and patient outcomes.