Automated Brain Tumor segmentation using hybrid YOLO and SAM

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Early-stage brain tumor detection is crucial for effective treatment and diagnosis.
• Deep learning models offer potential for improving diagnostic accuracy.

Purpose of the Study:

• To propose and evaluate a novel hybrid deep learning framework for early-stage brain tumor diagnosis.
• To integrate Convolutional Neural Network (CNN), YOLO, and Segment Anything Model (SAM) for enhanced tumor detection and segmentation.

Main Methods:

• A hybrid deep learning framework was developed, combining a CNN with YOLOv11 for object detection and SAM for precise segmentation.
• The CNN backbone was enhanced with deeper convolutional layers for robust feature extraction.
• YOLOv11 localized tumor regions, and SAM refined tumor boundaries via mask generation.

Main Results:

• The model was trained and validated on a dataset of 896 MRI brain images, including both tumorous and healthy scans.
• The CNN-based YOLO+SAM approach successfully segmented and diagnosed brain tumors.
• Achieved high performance metrics: 94.2% precision, 95.6% recall, and 96.5% mAP50(B).

Conclusions:

• The proposed hybrid deep learning model demonstrates significant effectiveness for early-stage brain tumor diagnosis.
• A comprehensive ablation study validated the model's robustness, suggesting suitability for clinical deployment.