Efficient Method for Early Detection of Brain Tumor

Abstract

In this paper, a three-step preprocessing algorithm is proposed to enhance the quality of Magnetic Resonance Imaging (MRI) scans for more accurate detection of various brain diseases. This preprocessing algorithm is combined with a new deep convolutional neural network (DCNN) structure for effective diagnosis of glioma, meningioma, and pituitary. The proposed model is computationally lightweight with a small number of convolutional and max-pooling layers, as well as simple initialization of the layer weights. This allows for faster training with a higher learning rate. To compare the proposed architecture to other models discussed in the paper, a demonstrative contrast was performed. The results of this experiment showed an impressive accuracy of 98.22% for glioma detection, 99.13% for meningioma detection, 97.3% for pituitary detection, and 97.14% for normal images. This was tested on a dataset with 3394 MRI images. The results of this experiment prove the robustness of the proposed architecture, allowing for increased accuracy in the diagnosis of a variety of brain illnesses in a short amount of time.