ResNet-152在脑胶质瘤MRI诊断中的应用研究OA

Objective:This study used artificial neural network to learn the cerebral MRI images of glioma patients to es-tablish auxiliary diagnosis model of glioma.Methods:A total of 241 patients from Tianjin Huanhu Hospital were included in-to this study.For them,there were 4 456 T1WI images and 4 451 T2WI images.For T1WI and T2WI images,we devided thd image data into train set and test set according to the ratio of 7∶3.ResNet-152 was used to establish the classification model.Accuracy,positive predic-tive value(PPV),sensitivity,specificity,negative predictive value(NPV),the area under the curve(AUC)of ROC and F1-measure were used to evaluate the classification effect of the ResNet-152 model.Results:For T1WI,the accuracy of the model was 88.40%,the PPV was 89.71%,the sensitivity was 82.34%,the specificity was 92.93%,the NPV was 87.55%,and the F1-measure was 91.29%.The AUC of ROC curve was 0.886.For T2WI,the accuracy of the model was 93.55%,the PPV was 95.27%,the sensitivity was 89.42%,the speci-ficity was 96.52%,the NPV was 92.09%,and the F1-measure was 92.25%.The AUC of ROC curve was 0.930.Conclusion:The ResNet-152 network performs well in classifying the MRI images into glioma and normal.The classification efficiency of ResNet-152 model on T2WI sequence is higher than that on T1WI sequence.