Harnessing Deep Learning for Enhancing Brain Tumor Classification in a Sustainable Healthcare Environment

Brain tumor affect human health owing to their location. Early detection plays a pivotal role in improving treatment efficacy and patient outcomes. Computer-Aided detection systems, leveraging medical imaging techniques, have emerged as valuable tools for assisting clinicians in diagnosing cancer. Deep learning, particularly convolutional neural networks, has shown remarkable promise in medical image analysis, offering enhanced accuracy and efficiency. This study employed transfer learning and deep neural networks for the classification of brain tumors into its different classes. The pre-trained CNN models, used in the classification were EfficientNet, ResNet, DenseNet, and Xception and a series of image pre-processing and augmentation were carried out before feeding the images to the classification models to enhance classification performance. Four ensemble models were built after getting the results from the classification models, ICAVG for image classification average predictions, ICMAX for image classification maximum predictions, FCAVG for extracted features classification average predictions, and FCMAX for extracted features classification maximum predictions. The results from these four ensemble models were again combined into CAVG for the average results of both ICAVG and FCAVG, and CMAX for the maximum results from both ICMAX and FCMAX. The proposed model was tested on a brain tumor dataset of 3064 samples and the investigational validation of the proposed technique portrayed a superior accuracy value of 99.02% over existing models. Based on high predictive results this research work endeavors to contribute to the advancement of cancer diseases detection, facilitating early diagnosis and improving patient outcomes.