Radiomics features and functional radiosensitivity enhances prediction of acute pulmonary toxicity.
Vincent Bourbonne,
France
PO-1245
Abstract
Radiomics features and functional radiosensitivity enhances prediction of acute pulmonary toxicity.
Authors: Vincent Bourbonne1,2, François Lucia1,2, Vincent Jaouen2,3, Julien Bert2, Martin Rehn1, Mathieu Hatt2, Olivier Pradier1,2, Dimitris Visvikis2, Ulrike Schick1,2
1University Hospital, Radiation Oncology, Brest, France; 2Université de Bretagne Occidentale, LaTIM, IMR 1101 INSERM, Brest, France; 3Insitut Mines, Telecom Atlantique, Brest, France
Show Affiliations
Hide Affiliations
Purpose or Objective
(Chemo)-radiotherapy is the standard treatment for patients with locally advanced lung cancer (LALC) not accessible to surgery. Despite strict application of dose constraints, acute toxicities such as acute pulmonary toxicity (APT) remain frequent, and may impact treatment’s compliance and patients’ quality of life. Two separate predictive models were previously developed: the Rad-model combined 6 radiomics features and achieved high results but consists in a posteriori evaluation of the APT risk. The Pmap-model was developed combining several clinical and dosimetric features after defining a specific zone in the posterior right lung and further validated in a prospective cohort. In the present study, we aim to combine all features to develop and validate a new model for prediction of the APT risk in two cohorts treated by volumetric-arctherapy (VMAT).
Material and Methods
For the training cohort, we retrospectively included all patients treated in our institution by VMAT for a LALC between 2015 and 2018. APT was scored according to the CTCAE v4.0 scale. All clinical (age, tobacco history, BPCO and VEMS) and dosimetric features (V5LungH, V10LungH, DMeanLungH and DMeanLungs, V30Lungs and V40Heart as well as the mean dose to the pre-defined Pmap zone (DMeanPmap)) were combined using a neural network approach and then applied to an observational prospective cohort for validation. To compensate for imbalanced data, the SMOTE package was used before model building. The model was evaluated using the Area under the curve (AUC) and the balanced accuracy (Bacc: mean of the sensitivity and specificity).
Results
165 and 42 patients were included in the training and validation cohorts, with respective APT rates of 22.4% and 19.1%. In the validation cohort, while the Rad-Model and the Pmap-Model achieved AUCs of 0.83 and 0.81 respectively, the combined model (Comb-Model) achieved an AUC of . With their respective thresholds defined on the training cohort, each model resulted in a Bacc of 0.82, 0.82 and 0.90 for the Rad, Pmap and Comb-models, respectively.
Conclusion
Addition of radiomics and DMeanPmap to usual clinical and dosimetric features enhances the accuracy of the newly developed prediction model. This model could be useful for a better assessment of the APT risk and open new treatments possibilities such as new treatment combinations.