Fuzzy-Based Atrous Convolution for Brain Tumor Detection Using MRI

Early detection and diagnosis of brain tumors are crucial for improving patient outcomes. While convolutional neural networks (CNNs) have demonstrated promise in detecting tumors from magnetic resonance images (MRI), traditional meth-ods often suffer from overparameterization, resulting in lower accuracy improvements despite higher computational demands. To address this, we propose an efficient approach, Fuzzy Atrous Convolution (FAC), with two additional modules: Top of the funnel and middle of the funnel. In this approach, fewer trainable parameters are required while maintaining high classification accuracy. A novel convolutional operator is introduced in the FAC model that dilates the receptive field while preserving input data, enabling efficient feature map reduction and accurate tumor detection. In the propose model, fuzzy logic improves adaptability and robustness. Experiments on three datasets confirmed the efficiency of proposed approach (FAC) and it achieved accuracies of 98.83%, 99.67%, and 99.56% on Datasets I, II, and III, respectively. Compared with large transfer learning models, Proposed approach has over 300 times fewer parameters, making them efficient for early brain tumor detection.