Innovative Approach for Brain Tumor Image Classification with Novel Optimized Feedforward Networks

Timely diagnosis and effective treatment planning rely on the accurate classification of brain tumor images. This study proposes a novel approach called Adaptive Moth Flame Optimized Feedforward Neural Network (AMFO-FNN) using Magnetic Resonance Imagin (MRI) for brain tumor classification. The method integrates Gaussian filtering for noise reduction and Gray-Level Co-occurrence Matrix (GLCM) for texture-based feature extraction. Classification is performed using a Feedforward Neural Network (FNN), whose parameters are optimized using an enhanced moth flame optimization algorithm. The model was evaluated on the publicly available Br35H dataset comprising 7,023 MRI images across four categories: meningioma, pituitary tumor, glioma, and no tumor. Applied preprocessing and data augmentation techniques to enhance generalization. Experimental results, validated through five-fold cross-validation, demonstrate the superior performance of AMFO-FNN, achieving 99.14% accuracy, 98.95% precision, 99.21% recall, and 99.08% F1-score. Comparative analysis with advanced models confirms the efficiency and robustness of the suggested approach. The model also shows minimal overfitting and high consistency, making it suitable for clinical application. Overall, AMFO-FNN offers a highly accurate and computationally efficient solution for automated brain tumor diagnosis