A stacked custom convolution neural network for voxel-based human brain morphometry classification.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Accurate automatic brain tumor identification remains a challenge in medical diagnostics.
• Existing methods often overlook the utility of voxel-based morphometry (VBM) in classification.
• There is a need for improved edge detection and classification accuracy in brain tumor analysis.

Purpose of the Study:

• To enhance brain tumor classification accuracy by integrating VBM with a stacked custom Convolutional Neural Network (CNN).
• To address limitations in current automatic brain tumor identification methods.
• To improve edge detection and overall classification performance.

Main Methods:

• Integration of voxel-based morphometry (VBM) for image normalization and segmentation.
• Development and application of a stacked custom Convolutional Neural Network (CNN) for tumor classification.
• Utilized ten-fold cross-validation and data augmentation for robust model training and testing.

Main Results:

• The proposed VBM and stacked custom CNN model demonstrated significant improvements in brain tumor classification.
• Achieved a high accuracy level of 98% in identifying brain tumors.
• Outperformed existing methods in brain tumor classification, as validated by ROC curves and other performance metrics.

Conclusions:

• The combined VBM and stacked custom CNN approach offers a superior method for automatic brain tumor classification.
• This novel integration significantly enhances diagnostic accuracy.
• The findings suggest a promising direction for improving automated neuroimaging analysis.