Brain Cancer Cataloguing using MRI with Optimization-based Deep Learning

Brain cancer remains a critical health challenge, requiring accurate diagnosis and classification for improved survival rates and treatment planning. In this work, we are discussing about deep learning system based on optimization for automatic brain cancer cataloguing using MRI data.. The methodology integrates preprocessing, data augmentation, deep convolutional neural networks (3D CNNs), and metaheuristic optimization (Particle Swarm Optimization, PSO) for hyperparameter tuning. In this approach we are using A collection of multimodal MRI images (T1, T1ce, T2, FLAIR) with identified tumor subregions was used for the experiments. When it came to tumor segmentation, our model's overall accuracy was 97.1%, its sensitivity was 96.3%, its specificity was 98.2%, and its Dice similarity coefficient was 0.92. Comparative findings demonstrate advantages over traditional CNN and SVM models. These results demonstrate the potential of optimization-driven deep learning models in clinical judgment when diagnosing brain cancer.