An efficient deep learning model for brain tumour detection with privacy preservation

Internet of medical things (IoMT) is becoming more prevalent in healthcare applicationsas a result of current AI advancements, helping to improve our quality of life and ensure asustainable health system. IoMT systems with cutting‐edge scientific capabilities arecapable of detecting, transmitting, learning and reasoning. As a result, these systemsproved tremendously useful in a range of healthcare applications, including brain tumourdetection. A deep learning‐based approach for identifying MRI images of brain tumourpatients and normal patients is suggested. The morphological‐based segmentationmethod is applied in this approach to separate tumour areas in MRI images. Convolu-tional neural networks, such as LeNET, MobileNetV2, Densenet and ResNet, are testedto be the most efficient ones in terms of detection performance. The suggested approachis applied to a dataset gathered from several hospitals. The effectiveness of the proposedapproach is assessed using a variety of metrics, including accuracy, specificity, sensitivity,recall and F‐score. According to the performance evaluation, the accuracy of LeNET,MobileNetV2, Densenet, ResNet and EfficientNet is 98.7%, 93.6%, 92.8%, 91.6% and91.9%, respectively. When compared to the existing approaches, LeNET has the bestperformance, with an average of 98.7% accuracy.