NTCP modelling with dosiomics features for postoperative complications in oesophageal cancer
PO-2087
Abstract
NTCP modelling with dosiomics features for postoperative complications in oesophageal cancer
Authors: Ruben Duwaerts1,4, Gilles Defraene1, Pieter Populaire1,2, Edmond Sterpin1,5, Karin Haustermans1,3
1KU Leuven, Laboratory of experimental radiotherapy, Leuven, Belgium; 2UZ Leuven, Department of radiation oncology, Leuven, Belgium; 3KU Leuven, Department of radiation oncology, Leuven, Belgium; 4AZ Turnhout, Department of radiation oncology, Turnhout, Belgium; 5Université catholique de Louvain, Institut de Recherche Expérimentale et Clinique, Center of Molecular Imaging, Radiotherapy and Oncology (MIRO), Brussels, Belgium
Show Affiliations
Hide Affiliations
Purpose or Objective
Approximately 30% of patients treated for locally advanced oesophageal cancer develop pulmonary and/or cardiac complications after neoadjuvant radiochemotherapy and surgical resection. This study aims to expose risk factors for these complications and to investigate the potential of features extracted from 3D dose distributions (dosiomics) in normal tissue complication probability (NTCP) modelling.
Material and Methods
Patients treated with trimodality therapy in UZ Leuven from 2011 to 2021 were included. Model building (122 patients, 2011-2018, 3DCRT or IMRT) and validation (47 patients, 2018-2021, IMRT) sets were defined based on treatment year. The prescription dose was 45 Gy in fractions of 1.8 Gy.
NTCP models were constructed and validated for pulmonary and cardiac postoperative complication endpoints. Clinical factors (age, BMI, tumour histology etc.) and DVH features (relative VxGy and mean dose for left lung, right lung, lungs combined, heart and left ventricle) were combined with dosiomics features calculated from the 3D dose distributions in the lungs and the heart. These dosiomics features include first- and second-order statistics and matrix-based texture features derived from the organ dose maps. Dosiomics extraction was implemented using the PyRadiomics library.
A two-step multivariate logistic regression analysis was performed. In the first step, a repeated 5-fold cross-validation process was used for preselection of the features with the highest predictive value. In the second step, the final models were composed by forward selection (likelihood ratio test) of features.
Results
The pulmonary complication NTCP model obtains AUC=0.86 (95%CI 0.72-0.97) at validation. It includes an increased risk for squamous cell carcinoma as compared to adenocarcinoma histology (p=0.006). The second predictor models the observed association between increased relative V15Gy on the left ventricle (p=0.008) and reduced complication risk. This observation needs further investigation.
For cardiac complications AUC=0.66 (95%CI 0.45-0.83) is obtained at validation. Age (p=0.001) is confirmed as a key risk factor. An increased probability for cardiac events for more distal tumour sites (highest risk at the gastro-oesophageal junction, p=0.007) and for open surgery as compared to minimally invasive approaches (p=0.13) is also included. The fourth predictor is the normalized gray level non-uniformity of the dose in the lungs (p=0.015), a dosiomics feature designating an increased risk when large lung areas are homogeneously irradiated.
Conclusion
The selection of a dosiomics feature instead of a DVH feature in the cardiac complication model illustrates the added value of advanced 3D dose characteristics in NTCP modelling. Interaction between lung and heart doses and pulmonary and cardiac complications are apparent in both models.