Deep-learning-based MRI in the diagnosis of cerebral infarction and its correlation with the neutrophil to lymphocyte ratio.

Abstract

Dizziness is a common symptom in clinic, but there lacks an effective treatment method. This study sought to examine the efficiency of deep learning (DL)-based magnetic resonance imaging (MRI) in the diagnosis of cerebral infarction mainly manifesting as vertigo using the neutrophil to lymphocyte ratio (NLR) and other routine blood indexes. An improved multiscale U-Net [MS (U-Net)] model, based on the U-net model, was proposed and applied in the segmentation of MRI of the brain. One hundred and fifteen vertiginous cerebral infarction (VCI) patients, admitted to the Department of Neurology at Huizhou Central People's Hospital from January 2016 to December 2020, were chosen as the research subjects. Based on the MRI segmentation results for the brain, the patients were allocated to the benign paroxysmal positional vertigo (BPPV) group or acute cerebral infarction (ACI) group. Additionally, 50 healthy individuals, whose venous blood was collected for routine blood analyses, were allocated to the control group. The MS (U-Net) model accomplishes MRI segmentation of the brain, and its segmentation results were much closer to the real results than those of the U-Net model. Compared to the control group, the monocyte count (MC), low-density lipoprotein/high-density lipoprotein (LDL/HDL) ratio, and NLR of patients in the BPPV and ACI groups showed an obvious increase (P<0.05), as did the white blood cell count, triglyceride (TG) level, and other indexes of ACI patients (P<0.05). In relation to the diagnosis, the areas under the curve for the TG level, LDL/HDL ratio, and NLR of the BPPV and ACI groups were 0.930 and 0.760, 0.900, and 0.770, 0.945 and 0.855, respectively (P<0.05). DL can accomplish MRI segmentation in cerebral infarction patients, and the TG level, LDL/HDL ratio and NLR can be used in the diagnosis of VCI.