Magnetic Resonance Image Feature Analysis under Deep Learning in Diagnosis of Neurological Rehabilitation in Patients with Cerebrovascular Diseases

Abstract

To explore the impact of magnetic resonance imaging (MRI) image features based on deep learning algorithms on the neurological rehabilitation of patients with cerebrovascular diseases, eighty patients with acute cerebrovascular disease were selected as the research objects. According to whether the patients were with vascular cognitive impairment (VCI), they were divided into VCI group (34 cases) and non-VCI group (46 cases). In addition, based on the convolutional neural network (CNN), a new multimodal CNN image segmentation algorithm was proposed. The algorithm was applied to the segmentation of MRI images of patients with vascular cognitive impairment (VCI) and compared with the segmentation results of CNN and fully CNN (FCN). As a result, the segmentation results of the three different algorithms showed that the Dice coefficient, accuracy, and recall of the multimodal CNN algorithm were 0.78 ± 0.24, 0.81 ± 0.28, and 0.88 ± 0.32, respectively, which were significantly increased compared to those of other two algorithms ( $P$  < 0.05). The neurological evaluation results showed that the MMSE and MoCA scores of VCI patients were 15.4 and 14.6 ± 5.31, respectively, which were significantly lower than those of the non-VCI group ( $P$  < 0.05). The TMT-a and TMT-b scores of VCI patients were 221.7 and 385.9, respectively, which were significantly higher than those of the non-VCI group ( $P$  < 0.05). The FA and MD values of nerve function-related fibers shown in the MRI images of the VCI group were significantly different from those of the non-VCI group ( $P$  < 0.05). Therefore, the neurological recovery process of VCI patients was affected by multiple neurocognitive-related fiber structures. To sum up, the multimodal CNN algorithm can sensitively and accurately reflect the degree of neurological impairment in patients with cerebrovascular disease and can be applied to disease diagnosis and neurological evaluation of VCI patients.