Smart brain tumor diagnosis system utilizing deep convolutional neural networks

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Early cancer diagnosis is critical for effective treatment and patient outcomes.
• Magnetic Resonance Imaging (MRI) is a primary diagnostic tool but has limitations causing delays.
• Computer-aided intelligent systems offer potential to enhance diagnostic accuracy and speed.

Purpose of the Study:

• To develop and evaluate a Convolutional Neural Network (CNN)-based system for brain tumor diagnosis.
• To improve upon existing deep learning architectures for medical image analysis.
• To assess the system's performance against state-of-the-art models.

Main Methods:

• Utilized EfficientNetv2s architecture for the CNN model, enhanced with Ranger optimization and extensive pre-processing.
• Compared the proposed model against ResNet18, ResNet200d, and InceptionV4.
• Evaluated performance on both augmented and original brain tumor imaging data.

Main Results:

• Achieved a micro-average test accuracy of 99.85% and Area Under the Curve (AUC) of 99.89%.
• Demonstrated high precision (98.16%), recall (98.17%), and F1-score (98.21%).
• The improved CNN model showed a significant impact on tumor categorization compared to other architectures.

Conclusions:

• The developed CNN system shows convincing performance in brain tumor detection and diagnosis.
• The AI-powered approach can assist physicians, potentially reducing diagnostic delays.
• This technology holds promise for improving early cancer management through accurate imaging analysis.