Automatic Brain Tumor Classification via Lion Plus Dragonfly Algorithm.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Computational Neuroscience

Background:

• Accurate brain tumor classification is crucial for effective treatment planning.
• Existing methods often face challenges with image noise and feature extraction.
• Developing robust and automated classification models is an ongoing research priority.

Purpose of the Study:

• To develop an accurate and efficient brain tumor classification model.
• To introduce a novel hybrid optimization technique for improving classifier performance.
• To evaluate the proposed model's effectiveness against standard approaches.

Main Methods:

• Image preprocessing including denoising (entropy-based trilateral filter) and skull stripping (morphological partition, Otsu thresholding).
• Image segmentation using Adaptive Contrast Limited Fuzzy Adaptive Histogram Equalization (CLFAHE).
• Feature extraction via Gray-Level Co-occurrence Matrix (GLCM).
• A hybrid classifier combining Neural Network (NN) and Deep Belief Network (DBN) with optimized hidden neurons using the Lion with Dragonfly Separation Update (L-DSU) optimization technique.

Main Results:

• The proposed model demonstrates improved prediction accuracy through optimized hybrid classifiers.
• The L-DSU optimization effectively enhances the performance of the combined NN and DBN models.
• Comparative analysis shows the superiority of the developed model over standard methods.

Conclusions:

• The developed brain tumor classification model, incorporating advanced preprocessing and a novel L-DSU optimization, offers enhanced accuracy.
• This approach provides a promising tool for automated and reliable brain tumor diagnosis in medical imaging.
• Further research can explore broader applications of the L-DSU optimization in medical AI.