Tumor grading of soft tissue sarcomas using MRI-based radiomics

Abstract

Background: Treatment decisions for multimodal therapy in soft tissue sarcoma (STS) patients greatly depend on the differentiation between low-grade and high-grade tumors. We developed MRI-based radiomics grading models for the differentiation between low-grade (G1) and high-grade (G2/G3) STS.

Methods: The study was registered at ClinicalTrials.gov (number NCT03798795). Contrast-enhanced T1-weighted fat saturated (T1FSGd), fat-saturated T2-weighted (T2FS) MRI sequences, and tumor grading following the French Federation of Cancer Centers Sarcoma Group obtained from pre-therapeutic biopsies were gathered from two independent retrospective patient cohorts. Volumes of interest were manually segmented. After preprocessing, 1394 radiomics features were extracted from each sequence. Features unstable in 21 independent multiple-segmentations were excluded. Least absolute shrinkage and selection operator models were developed using nested cross-validation on a training patient cohort (122 patients). The influence of ComBatHarmonization was assessed for correction of batch effects.

Findings: Three radiomic models based on T2FS, T1FSGd and a combined model achieved predictive performances with an area under the receiver operator characteristic curve (AUC) of 0.78, 0.69, and 0.76 on the independent validation set (103 patients), respectively. The T2FS-based model showed the best reproducibility. The radiomics model involving T1FSGd-based features achieved significant patient stratification. Combining the T2FS radiomic model into a nomogram with clinical staging improved prognostic performance and the clinical net benefit above clinical staging alone.

Interpretation: MRI-based radiomics tumor grading models effectively classify low-grade and high-grade soft tissue sarcomas. FUND: The authors received support by the medical faculty of the Technical University of Munich and the German Cancer Consortium.

Keywords: Biomarker; MRI; Radiomics; Risk stratification; Soft tissue sarcoma; Tumor grading.

Copyright © 2019 The Authors. Published by Elsevier B.V. All rights reserved.

Figures

Fig. 1

The radiomics workflow. Abbreviations: DCA: decision curve analysis, ICC: intra class coefficient, KM: Kaplan Meier survival curve, ROC: receiver operator characteristic curve, TUM: Technical University of Munich, UW: University of Washington.

Fig. 2

Predictive performance of radiomics tumor grading models. Receiver operator characteristic curves (ROC) and the respective area under the curve (AUC) values depicting the performance of the prediction models Clinical, Clinical-Volume-combined, Tumor-Volume, Radiomics-T1FSGd, Radiomics-T2FS, and the Radiomics-combined on the validation cohort. The shaded blue areas depict the 95% confidence interval which is shown in parentheses. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

Patient risk stratification. Kaplan Meier survival curves for patients' overall survival displaying patient stratification by the Clinical-Volume-combined model (a,f), tumor grading (low-grade vs. high-grade) (b,g), the Radiomics-T1FSGd model (c,h), the Radiomics-T2FS model (d,i) and the Radiomics-combined model (e,j) on the training (a,e) and validation (f,j) patient cohort.

Fig. 4

A clinical radiomics nomogram. A multivariate nomogram combining the Radiomics-T2FS prediction model with the AJCC clinical staging system is illustrated (a). The receiver operator characteristic (ROC) curve and the representative area under the curve (AUC), as well as the calibration curve, each at two  years, are shown (b,c). The Kaplan Meier survival curve for patients' overall survival displaying patient stratification by the proposed nomogram is depicted (d). Finally, decision durve analysis was performed comparing the net benefit by the Radiomics-T2FS nomogram with the AJCC clinical stage and tumor grading alone. The net benefit is calculated by subtracting the proportion of false-positive patients from the proportion of true-positive patients, weighted by the relative harm of a false-negative and false-positive result [28]. The threshold probability was calculated for death after five years. For reference, the two strategies “treat all” and “treat none” are displayed. A decision model shows a clinical benefit if the respective curve shows larger net benefit values than both reference strategies.