Computed tomography based radiomic signature as predictive of survival and local control after stereotactic body radiation therapy in pancreatic carcinoma

Abstract

Purpose: To appraise the ability of a radiomics signature to predict clinical outcome after stereotactic body radiation therapy (SBRT) for pancreas carcinoma.

Methods: A cohort of 100 patients was included in this retrospective, single institution analysis. Radiomics texture features were extracted from computed tomography (CT) images obtained for the clinical target volume. The cohort of patients was randomly divided into two separate groups for the training (60 patients) and validation (40 patients). Cox regression models were built to predict overall survival and local control. The significant predictors at univariate analysis were included in a multivariate model. The quality of the models was appraised by means of area under the curve and concordance index.

Results: A clinical-radiomic signature associated with Overall Survival (OS) was found significant in both training and validation sets (p = 0.01 and 0.05 and concordance index 0.73 and 0.75 respectively). Similarly, a signature was found for Local Control (LC) with p = 0.007 and 0.004 and concordance index 0.69 and 0.75. In the low risk group, the median OS and LC in the validation group were 14.4 and 28.6 months while in the high-risk group were 9.0 and 17.5 months respectively.

Conclusion: A CT based radiomic signature was identified which correlate with OS and LC after SBRT and allowed to identify low and high-risk groups of patients.

Conflict of interest statement

I have read the journal's policy and the authors of this manuscript have the following competing interests: L. Cozzi acts as Scientific Advisor to Varian Medical Systems and is Clinical Research Scientist at Humanitas Cancer Center. All other co-authors declare that they have no conflict interests. A. Chiti received speaker honoraria from General Electric and Sirtex Medical System; acted as scientific advisor for Blue Earth Diagnostics and Advanced Accelerator Applications; benefited from an unconditional grant from Sanofi to Humanitas University. All honoraria and grants are outside the scope of the submitted work. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Figures

Fig 1. Overall survival (OS) and local control (LC) curves stratified according to the best threshold for the clinical and radiomic features found to be significant at univariate analysis.

Data are shown for the training dataset. In the figures the blue lines correspond to the stratum above the threshold.

Fig 2

a) and b): Overall survival (OS) and Local Control (LC) curves for the training (solid red line) and validation (blue dashed line) cohorts of patients without any stratification; c) and d): Overall survival (OS) curves for the multivariate models A and B respectively. The sub-panels represent the ROC curves built out of the models. e) and f): Local Control (LC) curves as for the above. In the survival curves, solid lines correspond to the low risk group of patients; blue lines to the validation (red for the training) set. In the ROC curves, solid line is for the training, dashed for the validation.

Fig 3. Calibration plots at 6,12 and 18 months for the Cox models B for overall survival and local control.