Background and purpose: We used radiomic analysis to establish a radiomic signature based on anatomical magnetic resonance imaging (MRI) sequences and explore its effectiveness as a novel prognostic biomarker for skull base chordoma (SBC).
Materials and methods: In this retrospective study, radiomic analysis was performed using preoperative axial T1 FLAIR, T2-weighted, and enhanced T1 FLAIR from a single hospital. The primary clinical endpoint was progression-free survival. A total of 1860 3-D radiomic features were extracted from manually segmented region of interest. Pearson correlation coefficient was used for feature dimensional reduction and a ridge regression-based Cox proportional hazards model was used to determine a radiomic signature. Afterwards, radiomic signature and nine other potential prognostic factors, including age, gender, histological subtype, dural invasion, blood supply, adjuvant radiotherapy, extent of resection, preoperative KPS, and postoperative KPS were analyzed to build a radiomic nomogram and a clinical model. Finally, we compared the nomogram with each prognostic factor/model by DeLong's test.
Results: A total of 148 SBC patients were enrolled, including 64 with disease progression. The median follow-up time was 52 months (range 4-122 months). The Harrell's concordance index of the radiomic signature was 0.745 (95% CI, 0.709-0.781) for the validation cohort, and its discrimination accuracy in predicting progression risk at 5 years in the same cohort was 82.4% (95% CI, 72.6-89.7%).
Conclusions: The radiomics is a low-cost, non-invasive method to predict SBC prognosis preoperatively. Radiomic signature is a potential prognostic biomarker that may allow the individualized evaluation of patients with SBC.
Keywords: Biomarkers; Magnetic resonance imaging; Prognosis; Progression-free survival; Radiomics; Skull base chordoma.
Copyright © 2019 Elsevier B.V. All rights reserved.