Leveraging Deep Learning Model for Computer Vision-Based Brain Tumor Classification in 3D MRI Brain Images.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Brain tumors are a significant health concern requiring accurate and early detection.
• Magnetic Resonance Imaging (MRI) is a key modality for visualizing brain structures.
• Deep learning offers potential for automating and improving diagnostic accuracy in medical imaging.

Purpose of the Study:

• To develop and evaluate a combined deep learning model for detecting brain tumors in MRI.
• To assess the performance of the integrated EfficientNet-3D and 3D Residual neural network (3DResnet) architecture.
• To investigate the utility of the model in aiding early brain tumor diagnosis and treatment planning.

Main Methods:

• Utilized computer vision techniques to integrate EfficientNet-3D and 3DResnet deep learning architectures.
• Employed a dataset of 586 brain MRI images from the 2021 RSNA Brain Tumor Challenge.
• Applied preprocessing steps including contrast enhancement, resampling, interpolation, and center alignment.

Main Results:

• Achieved high performance metrics: 97.32% accuracy, 96.55% sensitivity, and 98.15% specificity.
• Validated results using 5-fold cross-validation.
• The model demonstrated capability to classify uncertain cases as 'unknown' for expert review.

Conclusions:

• The combined deep learning model shows significant potential for early and accurate brain tumor detection in MRI.
• This approach can enhance diagnostic capabilities, improve patient outcomes, and support effective treatment strategies.
• The model's ability to flag uncertain cases facilitates expert clinical judgment and further investigation.