ContourTL-Net: Contour-Based Transfer Learning Algorithm for Early-Stage Brain Tumor Detection.

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Brain tumors are life-threatening neurological conditions requiring early detection.
• Diagnosing brain tumors is challenging due to brain tissue complexity.
• Automated tools are essential for aiding healthcare professionals in brain tumor diagnosis.

Purpose of the Study:

• To enhance the efficacy of computerized brain tumor detection in clinical settings.
• To introduce a novel deep learning model, ContourTL-Net, for early-phase brain malignancy detection.
• To improve the accuracy and efficiency of brain tumor diagnosis.

Main Methods:

• A novel thresholding-based MRI image segmentation approach was developed.
• A transfer learning model, ContourTL-Net, utilizing VGG-16 architecture was employed.
• Contour-based analysis was used for precise segmentation and capturing tumor morphology.

Main Results:

• The ContourTL-Net model achieved high accuracy on benchmark datasets, including unseen clinical data.
• Key performance metrics included 100% sensitivity and NPV, 98.60% specificity, 99.12% precision, 99.56% F1-score, and 99.46% accuracy.
• The model outperformed existing state-of-the-art methodologies in both seen and unseen data.

Conclusions:

• The proposed ContourTL-Net model demonstrates significant potential for improving brain tumor detection.
• Early and accurate diagnosis through this model can lead to improved patient outcomes.
• The model's validation on unseen data highlights its generalization capability and real-world applicability.