Differentiation Of Lower Grade Glioma And Encephalitis Using Multiparametric MR-based On Deep Learning

ObjectiveComputational aid for diagnosis on convolutional neural network(CNN)is promising to improve the diagnostic accuracy of radiologists.Therefore,the study explored whether apply pretrained CNN models based on multiparametric magnetic resonance(MR)images could help to classify lower grade glioma and encephalitis.MethodsAll MRI brain images from 164 patients with a final diagnosis of WHO Ⅱ and WHO Ⅲ glioma(n=56)and encephalitis(n=108)patients and divided into training and testing sets.We applied three MRI sequences(FLAIR、CE-T1 WI and T2WI)as the input data.The three CNN models(Alexnet,Res Net-50 and Inception-v3)pretrained by the Image Net dataset to classify lower grade glioma and encephalitis.The diagnostic performance of these models were evaluated by using the area under the receiver operator characteristic curve(AUC)of a 5-fold cross-validation and the accuracy,sensitivity,specificity were analyzed.And then compared their classification performnce with radiologists’ visual diagnostic performance.ResultsThe AUC values of the three pre-trained CNN models for differentiating lower grade glioma and encephalitis were all over 0.9 with excellent performance.The highest classification accuracy of 97.57% was achieved by the Inception-v3 model based on the T2 WI.In addition,Inception-v3 performed statistically significantly better than the Alexnet architecture(p<0.05).For Inception-v3 and Res Net-50 models,T2 WI offered the highest accuracy,followed by CE-T1 WI and FLAIR.The classification performance of Inception-v3 and Res Net-50 had a significant difference with radiologists’ diagnostic performance(p<0.05),but there was no significant difference between the results of the Alexnet model and those of a more experienced radiologist(p >0.05).ConclusionsThe pretrained CNN models can automatically extract deep features from multi-parameter MR images,which is helpful to accurately classify lower grade glioma and encephalitis,and further help to improve clinical diagnostic performance.