Natural Changes in Radiological and Radiomics Features on MRIs of Soft‐Tissue Sarcomas Naïve of Treatment: Correlations With Histology and Patients' Outcomes

Abstract

Because of long diagnostic intervals, soft-tissue sarcoma (STS) patients can undergo several MRIs before treatments. However, only the latest pre-treatment MRI is used in clinical practice and the natural changes in MRI presentations of STS occurring before any medical procedure remain unknown. To qualitatively and quantitatively depict the natural history of MRI presentations of STS prior to medical intervention, to investigate their prognostic value, and to compare methods to calculate the changes in radiomics features (named delta-radiomics features). Retrospective. Sixty-eight patients with locally advanced histologically proven STS and two pre-treatment contrast-enhanced (CE) MRIs (median age: 64 years, median delay between MRIs: 77 days). Two-dimensional (2D) turbo spin echo (TSE) T1-weighted-imaging (WI) and T2-WI; 2D TSE or 3D gradient echo CE-T1-WI at 1.5T. Radiomics analysis was performed on 2D TSE CE-T1-WI. Three radiologists independently reported morphological features, evaluating changes in STS dimensions, intra-tumoral necrotic and hemorrhagic signals and heterogeneity, and changes in the tumor peritumoral enhancement, edema, and tail sign. After homogenizing the MRIs to account for differences in acquisition parameters, STS were 3D-segmented on both CE-T1-WI MRIs and radiomic features (RFs) were extracted. Changes in RFs between the two MRIs were calculated according to five methods: absolute, absolute/time between MRIs, relative, relative/time between MRIs, and log ratio. Histopathological samples were reviewed to count mitosis and Ki67 immunostaining. Survival data regarding local relapse, metastatic relapse, and disease-related deaths were collected. Reproducibility analysis (using intra-class correlation coefficient and [weighted] kappa), hierarchical clusterings based on changes in RFs, survival analyses (using Cox regressions), and association with histopathology (using Student's t-test, Wilcoxon, or Chi-squared test). A P-value of <0.05 was considered to be statistically significant. There were 15 and 26 local and metastatic progressions, respectively. Average tumor size increase between scans was +39.8%. Metastatic relapse-free survival (MFS) was associated with: increases in size, tumor heterogeneity on T1-WI, T2-WI, and CE-T1-WI, necrotic signal, peritumoral enhancement, and tail sign. Local relapse-free survival (LFS) was associated with: increase in tumor heterogeneity on T1-WI, necrotic signal, hemorrhagic signal and peritumoral edema, and clusters based on the logarithmic changes in RFs (Log-RF). Increase in heterogeneity on CE-T1-WI and Log-RF clusters were independent predictors for MFS and LFS, respectively, in stepwise multivariate Cox regression (hazard ratio [HR]=2.78 and HR=+∞ respectively). Associations were found between changes in necrotic signal, heterogeneity on CE-T1-WI and peritumoral enhancement, and histological markers of proliferation. Changes in MRI presentation of STS before any treatment are frequent, associated with histopathology, and could help in patients' prognostication, in addition to baseline MRI feature. 3 TECHNICAL EFFICACY: Stage 3.