팬텀과 뇌종양 임상데이터를 사용하여 검증한 다중 자기공명영상에서의 라디오믹스 재현성

Abstract

The Image Biomarker Standardization Initiative has helped improve the computational reproducibility of MRI radiomics features. Nonetheless, the MRI sequences and features with high imaging reproducibility are yet to be established. To determine reproducible multiparametric MRI radiomics features across test-retest, multi-scanner, and computational reproducibility comparisons, and to evaluate their clinical value in brain tumor diagnosis. To assess reproducibility, T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), and diffusion-weighted imaging (DWI) were acquired from three 3-T MRI scanners (GE, Philips, and Siemens) using standardized phantom, and radiomics features were extracted using two computational algorithms (Pyradiomics and MITK). Reproducible radiomics features were selected when the concordance correlation coefficient (CCC) value above 0.9 across multiple sessions, scanners, and computational algorithms. Random forest classifiers were trained with reproducible features (n=117) and validated in a clinical cohort (internal n=50, external n=67) to evaluate whether features with high reproducibility improved the differentiation of glioblastoma from primary central nervous system lymphomas (PCNSL). Radiomics features from T2WI demonstrated higher repeatability (65-94%) than those from DWI (38-48%) or T1WI (2-92%). Across test-retest, multi-scanner, and computational comparisons, T2WI provided 41 reproducible features, DWI provided six, and T1WI provided two. The performance of the classification model with reproducible features was higher than that using non-reproducible features in training set (AUC, 0.916 vs. 0.877) and validation set (internal set, 0.957 vs. 0.869, external, 0.950 vs. 0.873). Radiomics features with high reproducibility across multiple sessions, scanners, and computational algorithms were identified, and they showed higher diagnostic performance than non-reproducible radiomics features in the differentiation of glioblastoma from PCNSL.