BT-CNN: a balanced binary tree architecture for classification of brain tumour using MRI imaging

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Deep learning for diagnostics

Background:

• Deep learning, particularly Convolutional Neural Networks (CNNs), is crucial for clinical diagnosis and therapy.
• Accurate brain tumor identification is vital for effective patient treatment.
• Existing CNN models face challenges in balancing computational time and memory usage.

Purpose of the Study:

• To propose a novel Balanced binary Tree CNN (BT-CNN) for enhanced brain tumor classification.
• To optimize performance by balancing convolution and depthwise separable convolution modules.
• To improve the accuracy and efficiency of automated brain tumor detection.

Main Methods:

• Development of a unique BT-CNN architecture inspired by binary trees.
• Implementation of distinct convolution and depthwise separable convolution groups within the BT-CNN.
• Pre-processing of medical images using CLAHE, denoising, cropping, and scaling.
• Utilizing 5-fold cross-validation for robust model training and testing.
• Comparative analysis against state-of-the-art models including CNN-KNN, Musallam et al., Saikat et al., and Amin et al.

Main Results:

• The proposed BT-CNN achieved an average training accuracy of 99.61%.
• The BT-CNN attained a test accuracy of 96.06%, significantly outperforming other models (68.86% to 90.41%).
• The model demonstrated the lowest standard deviation in accuracies across all folds, indicating high stability and reliability.

Conclusions:

• The BT-CNN offers a significant advancement in brain tumor classification accuracy and efficiency.
• The balanced architecture effectively optimizes computational time and memory usage.
• This deep learning approach shows promise for improving clinical diagnostic tools for neurological conditions.