Multiple Sclerosis (MS) is a neurodegenerative disease that occurs because of demyelination in nerve cells. Early treatment can be provided, and its progression can be prevented with an early diagnosis of the disease. The most remarkable finding in the identifying of MS disease is white matter lesions in the brain, which can be detected by magnetic resonance imaging (MRI). In this study, the identification of MS was performed with the proposed Convolutional Neural Network (CNN) model by detecting the presence of lesions from the brain Fluid-Attenuated Inversion Recovery (FLAIR) Magnetic Resonance (MR) images. The features of MS lesions in MR images are extracted with the proposed CNN model as an efficient and useful model with a low number of trainable parameters. The proposed CNN model has been compared with the traditional machine learning and state-of-the-art DL methods on a 5-fold cross-validation procedure. All methods are implemented on the same dataset. The results were obtained with both slice-level and patient-level data splitting methods. According to the results of slice-level splitting, the proposed CNN model achieved better success with the accuracy of 98.0% (± 0.02), the sensitivity of 97.9% (± 0.03), specificity of 98.3% (± 0.03), precision of 98.2% (± 0.03) values. In the results obtained with patient-level splitting, the accuracy of 90.3% (± 0.05), the sensitivity of 90.5% (± 0.05), the specificity of 90.1% (± 0.09), and the precision of 91.1% (± 0.09). The proposed CNN model obtained high and consistent performance in both splitting methods compared to other methods. © 2001 Elsevier Science. All rights reserved.