Radiology: Machine learning helps you find and calculate acute ischemic brain infarction

Backgroundto be able to identify acute infarction and its extent at the onset of the disease is very important for the treatment of patients with acute ischemic stroke (AIS)Patients with large-scale infarction are less likely to benefit from thrombotic or mechanical thrombosispurpose
this study aims to use non-enhanced CT examination to establish an automatic calculation method that can evaluate acute terrier lesions and their volume in AIS patientsmaterials and methods
This study collected non-enhanced CT images of AIS patients (from onset to CT for 6 hours), and all patients were tested for protosconous-weighted (DW) MRI within 1 hour of The onset of AISThe volume of ischemic cerebral infarction is artificially sketched as the reference standardEstablish an automatic segmentation method that includes machine learning (ML) to identify a brain infarctionThe ML model was trained and validated in 157 cases with non-enhanced images of the artificially sketched DW MRI image infarction lesions, and the remaining 100 patients were tested as separate groupsThe difference between the quantitative comparison of ML algorithms and reference standards (DW MRI) was analyzed using Bland-Altman plots and pearson correlationsresultsin 100 test groups, the baseline non-enhanced CT test time was 48 minutes after the onset of the median, the baseline MRI test time was 38 minutes after the CT testThe algorithm to check the disease volume is related to the artificial sketching of the disease volume by reference standardThe average difference between the segmented volume and the DW MRI in the algorithm was 11 ml conclusion
use machine learning algorithms to automatically segment the volume of cerebral infarction in non-enhanced CT images and the volume of cerebral infarction in diffusion-weighted MRI examination