Using machine learning-based radiomics to differentiate between glioma and solitary brain metastasis from lung cancer and its subtypes

• DOI: 10.1007/s12672-023-00837-6
• Publication Date: 2023-12-06T12:02:31Z
• Authors: Feng-Ying Zhu; Yu-Feng Sun; Xiao-Ping Yin; Yu Zhang; Li-Hong Xing; Ze-Peng Ma; Lin-Yan Xue; Jia-Ning Wang
• Keywords: Multilayer perceptron; Perceptron; Cross-validation

Abstract Objective To establish a machine learning-based radiomics model to differentiate between glioma and solitary brain metastasis from lung cancer its subtypes, thereby achieving accurate preoperative classification. Materials methods A retrospective analysis was conducted on MRI T1WI-enhanced images of 105 patients with 172 cancer, which were confirmed pathologically. The divided into the training group validation in an 8:2 ratio for image segmentation, extraction, filtering; multiple layer perceptron (MLP), support vector (SVM), random forest (RF), logistic regression (LR) used modeling; fivefold cross-validation train model; evaluate assess predictive performance model, ROC curve calculate accuracy, sensitivity, specificity area under (AUC) model. Results accuracy AUC MLP differentiation high-grade 0.992, 1.000, respectively, while sensitivity 0.968, respectively. SVM small cell 0.966, 0.929, non-small 0.982, 0.999, 0.958, Conclusion application has certain clinical value differentiating subtypes. In HGG/SBM HGG/NSCLC SBM groups, had best diagnostic performance, HGG/SCLC group, models performance.