Radiomics modelling in rectal cancer to predict disease-free survival: evaluation of different approaches
Open Access
- 23 August 2021
- journal article
- research article
- Published by Oxford University Press (OUP) in British Journal of Surgery
- Vol. 108 (10), 1243-1250
- https://doi.org/10.1093/bjs/znab191
Abstract
Radiomics may be useful in rectal cancer management. The aim of this study was to assess and compare different radiomics approaches over qualitative evaluation to predict disease-free survival (DFS) in patients with locally advanced rectal cancer treated with neoadjuvant therapy. Patients from a phase II, multicentre, randomized study (GRECCAR4; NCT01333709) were included retrospectively as a training set. An independent cohort of patients comprised the independent test set. For both time points and both sets, radiomic features were extracted from two-dimensional manual segmentation (MS), three-dimensional (3D) MS, and from bounding boxes. Radiomics predictive models of DFS were built using a hyperparameters-tuned random forests classifier. Additionally, radiomics models were compared with qualitative parameters, including sphincter invasion, extramural vascular invasion as determined by MRI (mrEMVI) at baseline, and tumour regression grade evaluated by MRI (mrTRG) after chemoradiotherapy (CRT). In the training cohort of 98 patients, all three models showed good performance with mean(s.d.) area under the curve (AUC) values ranging from 0.77(0.09) to 0.89(0.09) for prediction of DFS. The 3D radiomics model outperformed qualitative analysis based on mrEMVI and sphincter invasion at baseline (P = 0.038 and P = 0.027 respectively), and mrTRG after CRT (P = 0.017). In the independent test cohort of 48 patients, at baseline and after CRT the AUC ranged from 0.67(0.09) to 0.76(0.06). All three models showed no difference compared with qualitative analysis in the independent set. Radiomics models can predict DFS in patients with locally advanced rectal cancer.Funding Information
- French National Cancer Institute (INCa-DGOS_5506, PHRC-K 2012–112)
- Site de Recherche Intégrée sur le Cancer Montpellier Cancer (INCa_Inserm_DGOS_12553)
This publication has 38 references indexed in Scilit:
- Differentiating brain metastases from different pathological types of lung cancers using texture analysis of T1 postcontrast MRMagnetic Resonance in Medicine, 2015
- Machine Learning methods for Quantitative Radiomic BiomarkersScientific Reports, 2015
- Tailored rectal cancer treatment - a time for implementing contemporary prognostic factors?Colorectal Disease, 2013
- The Use of MR Imaging in Treatment Planning for Patients with Rectal Carcinoma: Have You Checked the “DISTANCE”?Radiology, 2013
- SMOTE for high-dimensional class-imbalanced dataBMC Bioinformatics, 2013
- Assessment of Primary Colorectal Cancer Heterogeneity by Using Whole-Tumor Texture Analysis: Contrast-enhanced CT Texture as a Biomarker of 5-year SurvivalRadiology, 2013
- MR Volumetric Measurement of Low Rectal Cancer Helps Predict Tumor Response and Outcome after Combined Chemotherapy and Radiation TherapyRadiology, 2012
- Diffusion-weighted MR Volumetry for Assessing the Response of Rectal Cancer to Combined Radiation Therapy with ChemotherapyRadiology, 2012
- Assessing the Performance of Prediction ModelsEpidemiology, 2010
- Colorectal Cancer: Texture Analysis of Portal Phase Hepatic CT Images as a Potential Marker of SurvivalRadiology, 2009