Brain tumor detection on magnetic resonance imaging scans using the artificial intelligence-based You Only Look Once...

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Computer-Aided Diagnosis

Background:

• Brain tumor magnetic resonance imaging (MRI) scans present challenges including blurred boundaries and irregular shapes.
• Automated diagnosis systems require high accuracy and real-time processing capabilities for clinical application.
• Existing lightweight models may struggle with complex features and boundary definition in medical scans.

Purpose of the Study:

• To develop a lightweight detection framework for enhanced automated brain tumor diagnosis from MRI scans.
• To improve the accuracy and robustness of brain tumor detection, addressing issues like blurred boundaries.
• To meet real-time clinical requirements for efficient computer-aided diagnosis systems.

Main Methods:

• An improved You Only Look Once version 12 (YOLOv12n) model was developed, incorporating three novel modules.
• The Attention-based C2f with Frequency-domain Feed-Forward Network (A2C2f-DFFN) module enhances global context and feature reconstruction.
• The C2f with Token Statistics Self-Attention and Dynamic Tanh (C2TSSA-DYT) module and dynamic upsampling improve robustness and prevent detail loss.

Main Results:

• The proposed method achieved 93.2% precision, 88.4% recall, and 94.1% mAP@0.5 on the Kaggle brain tumor dataset.
• Performance surpassed the baseline YOLOv12n and other lightweight models, demonstrating effectiveness in glioma and pituitary tumor cases.
• The model operates efficiently with 6.0 GFLOPs and 2.76M parameters, enabling real-time inference.

Conclusions:

• The enhanced YOLOv12n framework offers a strong balance between accuracy and efficiency for brain tumor detection.
• The model exhibits robustness, making it suitable for clinical computer-aided diagnosis systems.
• This approach effectively tackles challenges in brain tumor MRI analysis, paving the way for improved diagnostic tools.