Enhancement of medical images diagnosis using fuzzy convolutional neural network

Abstract

Brain diseases are primarily brought on by abnormal brain cell growth, which can harm the structure of the brain and eventually result in malignant brain cancer. Major challenges exist when using a computer aided diagnosis (CAD) system for an early diagnosis that enables decisive treatment, particularly when it comes to the accurate detection of various diseases in the pictures for magnetic resonance imaging (MRI). In this study, the fuzzy convolutional neural networks (FCNN) were proposed for accurate diagnosis of brain tumors (glioma, meningioma, pituitary and non-tumor) which is implemented using Keras and TensorFlow. This approach follows three steps, training, testing, and evaluation. In training process, it builds a smart model and the structure consists of seven blocks (convolution, rectified linear unit (ReLU), batch normalization, and max pooling) then use flatten, fuzzy inferences layer, and dense layer with dropout. An international dataset with 7,022 brain tumor MRI images, was tested. The evaluation model attained a high performance with training accuracy of 99.84% and validation accuracy is 98.63% with low complexity and time is 58 s per epoch. The suggested approach performs better than the other known algorithms and may be quickly and accurately used for medical picture diagnosis.