Brain Tumor Detection Based on Deep Learning Approaches and Magnetic Resonance Imaging

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Brain tumors pose significant health risks, necessitating accurate and efficient detection methods.
• Manual detection of brain tumors from MRI scans is challenging, time-consuming, and prone to errors.
• Advanced computational models are crucial for improving diagnostic accuracy and speed.

Purpose of the Study:

• To develop and evaluate a refined You Only Look Once version 7 (YOLOv7) model for accurate detection of brain tumors (meningioma, glioma, pituitary).
• To enhance the YOLOv7 model's feature extraction and fusion capabilities for improved tumor identification in MRI scans.
• To provide a robust AI-driven tool to assist medical professionals in brain tumor diagnosis.

Main Methods:

• Image enhancement techniques and data augmentation were applied to a diverse brain tumor dataset.
• The YOLOv7 model was modified with a Convolutional Block Attention Module (CBAM) and a Spatial Pyramid Pooling Fast+ (SPPF+) layer.
• Decoupled heads and a Bi-directional Feature Pyramid Network (BiFPN) were integrated to optimize feature learning and fusion.

Main Results:

• The proposed enhanced YOLOv7 model demonstrated superior performance in detecting meningioma, glioma, and pituitary tumors.
• The integration of CBAM and SPPF+ layers significantly improved feature extraction and model sensitivity.
• The model achieved higher overall accuracy compared to existing state-of-the-art brain tumor detection methods.

Conclusions:

• The enhanced YOLOv7 model offers a promising and accurate solution for automated brain tumor detection in MRI.
• The developed framework has the potential to serve as a valuable decision-support tool for radiologists and oncologists.
• Further research can explore broader applications and clinical integration of this AI-powered diagnostic system.