A Pre-Trained Deep Learning Neural Network Model For Preoperative Differentiation Of Primary Central Nervous System Lymphoma,Glioblastoma And Brain Metastasis

Objective: To investigate the feasibility and to evaluate the classification performance metrics of deep learning model trained on T1-weighted enhanced MRI scans of primary central nervous system lymphoma(PCNSL),glioblastoma(GBM),and solitary or isolated brain metastases(BM).The model is optimized and fine-tuned to provide reliable,cost-effective and expeditious decision-making support for preoperative differential diagnosis of malignant brain tumors in complex and challenging clinical settings.Methods: We enrolled 149 patients who underwent surgery and were diagnosed with primary central nervous system lymphoma,glioblastoma,brain metastasis based on pathological examination from January 2015 to June 2022 at our hospital.T1-weighted enhanced MRI scans of the patients were extracted from the Picture Archiving and Communication System(PACS).The images were preprocessed after manual delineation of regions of interest(ROIs)using Slicer.The patients were randomly divided into training set and testing set,and data augmentation was performed on the training set.A deep learning neural network model based on EfficientNetV2 with pre-trained and finetuned on ImageNetdatasets was constructed,and then trained on the training set by transfer learning.The model was validated on the testing set,and then the classification performance metrics was evaluated by ROC curves,accuracy,recall(sensitivity),precision,specificity,F1 score,and area under the ROC curve(AUC).Results: The overall accuracy of the model was 89.4%,and the average AUC was 0.96.The AUC of the model for different tumor types(PCNSL,GBM,BM)were 0.98,0.96 and 0.95.Conclusion: This study demonstrated the feasibility of utilizing a pre-trained deep learning neural network model trained on axial T1-weighted enhanced sequence MRI data for the differential diagnosis of primary central nervous system lymphoma,glioblastoma and brain metastasis.The experimental findings indicate that the model exhibits outstanding diagnostic performance and clinical guidance value.Furthermore,the proposed model displays good robustness,affirming the deep learning neural network’s diagnostic capabilities under challenging conditions.Moreover,the model’s benefits include cost-effectiveness,reliability,and prompt diagnosis,suggesting its immense potential for clinical application.