Multiple sclerosis identification in brain MRI images using wavelet convolutional neural networks

Abstract

Multiple sclerosis (MS) is a degenerative disease of the covering around the nerves in the central nervous system. It damages the immune cells and causes small lesions in the patient s brain. Automated image recognition techniques can be employed for increasing the accuracy of detection. The use of convolutional neural networks (CNN) is the most common deep learning method for detecting lesions in image. Due to the specific features of MS lesions, the use of spectral features especially multiresolution enables the highlighting of images lesions and leads to a more accurate diagnosis. In the present study, the Haar wavelet transform was applied to make use of the spectral information. The proposed method is a combination of the two‐dimensional discrete Haar wavelet transform and the CNN network. Experiments on the image data of 38 patients and 20 healthy individuals revealed accuracy, precision, and sensitivity of 99.05%, 98.43%, and 99.14%, respectively.