Document Type : Original Paper
School of Control Science and Engineering, Shandong University, Jinan 250061, China
School of Control Science and Engineering, Shandong University, Jinan, China
School of Biological Science, Islamic Azad University of Najafabad, Isfahan, Iran
College of Automation and Electronic Engineering, Qingdao University of Science & Technology, Qingdao 260061, China
Introduction: Diagnosing the brain lesions through MRI images has always been a challenging issue for the doctors due to the brightness levels of the images. This study aimed to segment MS lesions using artificial intelligence and a deep learning process. Therefore, affected and healthy people were distinguished by the convolutional network in this process.
Material and Methods: The images of dataset were randomly imported to the system with healthy and affected labels to train the system. Then, the test images were classified based on the features extracted under supervision. In the next stage, the lesion was isolated for detection by the regression neural network and the particle swarm optimization process based on two 2D parts. Finally, the best possible matrix for convolutional layers was extracted and used by the regression network. The second stage consists of two parts of 2D to modify the lesion location and size and 3D to detect the lesion volume.
Results: This method could identify more than 96% of the patients and detect the lesion location and size up to 83%.
Conclusion: According to the results, this method is appropriate for segmenting MS lesions and can be used as an auxiliary tool alongside the doctor.