Anzeige
Erschienen in:
06.04.2022 | Imaging Informatics and Artificial Intelligence
Development and validation of MRI-based radiomics signatures models for prediction of disease-free survival and overall survival in patients with esophageal squamous cell carcinoma
Erschienen in: European Radiology | Ausgabe 9/2022
Einloggen, um Zugang zu erhaltenAbstract
Objectives
To develop and validate an optimal model based on the 1-mm-isotropic-3D contrast-enhanced StarVIBE MRI sequence combined with clinical risk factors for predicting survival in patients with esophageal squamous cell carcinoma (ESCC).
Methods
Patients with ESCC at our institution from 2015 to 2017 participated in this retrospective study based on prospectively acquired data, and were randomly assigned to training and validation groups at a ratio of 7:3. Random survival forest (RSF) and variable hunting methods were used to screen for radiomics features and LASSO-Cox regression analysis was used to build three models, including clinical only, radiomics only and combined clinical and radiomics models, which were evaluated by concordance index (CI) and calibration curve. Nomograms and decision curve analysis (DCA) were used to display intuitive prediction information.
Results
Seven radiomics features were selected from 434 patients, combined with clinical features that were statistically significant to construct the predictive models of disease-free survival (DFS) and overall survival (OS). The combined model showed the highest performance in both training and validation groups for predicting DFS ([CI], 0.714, 0.729) and OS ([CI], 0.730, 0.712). DCA showed that the net benefit of the combined model and of the clinical model is significantly greater than that of the radiomics model alone at different threshold probabilities.
Conclusions
We demonstrated that a combined predictive model based on MR Rad-S and clinical risk factors had better predictive efficacy than the radiomics models alone for patients with ESCC.
Key Points
• Magnetic resonance–based radiomics features combined with clinical risk factors can predict survival in patients with ESCC.
• The radiomics nomogram can be used clinically to predict patient recurrence, DFS, and OS.
• Magnetic resonance imaging is highly reproducible in visualizing lesions and contouring the whole tumor.