External validation of a combined PET and MRI radiomics model for prediction of recurrence in cervical cancer patients treated with chemoradiotherapy

Abstract

The aim of this study was to validate previously developed radiomics models relying on just two radiomics features from 18 F-fluorodeoxyglucose positron emission tomography (PET) and magnetic resonance imaging (MRI) images for prediction of disease free survival (DFS) and locoregional control (LRC) in locally advanced cervical cancer (LACC). Methods Patients with LACC receiving chemoradiotherapy were enrolled in two French and one Canadian center. Pretreatment imaging was performed for each patient. Multicentric harmonization of the two radiomics features was performed with the ComBat method. The models for DFS (using the feature from apparent diffusion coefficient (ADC) MRI) and LRC (adding one PET feature to the DFS model) were tuned using one of the French cohorts (n = 112) and applied to the other French (n = 50) and the Canadian (n = 28) external validation cohorts. Results The DFS model reached an accuracy of 90% (95% CI [79-98%]) (sensitivity 92-93%, specificity 87-89%) in both the French and the Canadian cohorts. The LRC model reached an accuracy of 98% (95% CI [90-99%]) (sensitivity 86%, specificity 100%) in the French cohort and 96% (95% CI [80-99%]) (sensitivity 83%, specificity 100%) in the Canadian cohort. Accuracy was significantly lower without ComBat harmonization (82-85% and 71-86% for DFS and LRC, respectively). The best prediction using standard clinical variables was 56-60% only. Conclusions The previously developed PET/MRI radiomics predictive models were successfully validated in two independent external cohorts. A proposed flowchart for improved management of patients based on these models should now be confirmed in future larger prospective studies.

Sensitivity of radiomic features to inter-observer variability and image pre-processing in Apparent Diffusion Coefficient (ADC) maps of cervix cancer patients

“…Radiomics has been extensively applied to CT and Positron Emission Tomography (PET), with a growing number of studies addressing its role in MRI and specifically in DWI and ADC maps [7,8]. Two recent publications [9,10] as an independent predictor of disease-free survival and locoregional control in cervical cancer. These results highlight the promising role of MRI for outcome prediction in oncology and the urgent need to develop a robust methodology for feature extraction in DWI.…”

Reply to E. Laffon et al.

“…8,9 A recent study of cervical cancer supported the benefits of harmonization in that respect. 10 Nevertheless, despite all these potential measurement errors, robust significant data were reported in the LyMa-PET study. The results remain the same for the SUVpeak and to a lesser extent for the SUVmean.…”

Radiomics: Data Are Also Images

“…The ComBat method (45) has been shown to work well for small samples and to outperform similar techniques (46). ComBat was shown to allow PET radiomics predictive models to achieve higher performance in the external validation step (47,48). This approach has several advantages: it is easy and fast, and it allows all of the information to be exploited because all of the features are retained.…”

“…Obviously, the larger the cohort, the better, as evaluating the final model with only 20 patients can provide limited evidence of its usefulness. Alternatively, if different datasets are available (e.g., in a multicenter setting), then it may be appropriate to train and validate with 1 cohort and test the resulting model with other cohorts (47,80). However, this approach requires harmonization of the features because of differences in their distributions across centers.…”