Privacy-preserving distributed learning of radiomics to predict overall survival and HPV status in head and neck cancer

Marta Bogowicz*, Arthur Jochems, Timo M. Deist, Stephanie Tanadini-Lang, Shao Hui Huang, Biu Chan, John N. Waldron, Scott Bratman, Brian O’Sullivan, Oliver Riesterer, Gabriela Studer, Jan Unkelbach, Samir Barakat, Ruud H. Brakenhoff, Irene Nauta, Silvia E. Gazzani, Giuseppina Calareso, Kathrin Scheckenbach, Frank Hoebers, Frederik W.R. WesselingSimon Keek, Sebastian Sanduleanu, Ralph T.H. Leijenaar, Marije R. Vergeer, C. René Leemans, Chris H.J. Terhaard, Michiel W.M. van den Brekel, Olga Hamming-Vrieze, Martijn A. van der Heijden, Hesham M. Elhalawani, Clifton D. Fuller, Matthias Guckenberger, Philippe Lambin

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review


A major challenge in radiomics is assembling data from multiple centers. Sharing data between hospitals is restricted by legal and ethical regulations. Distributed learning is a technique, enabling training models on multicenter data without data leaving the hospitals (“privacy-preserving” distributed learning). This study tested feasibility of distributed learning of radiomics data for prediction of two year overall survival and HPV status in head and neck cancer (HNC) patients. Pretreatment CT images were collected from 1174 HNC patients in 6 different cohorts. 981 radiomic features were extracted using Z-Rad software implementation. Hierarchical clustering was performed to preselect features. Classification was done using logistic regression. In the validation dataset, the receiver operating characteristics (ROC) were compared between the models trained in the centralized and distributed manner. No difference in ROC was observed with respect to feature selection. The logistic regression coefficients were identical between the methods (absolute difference <10−7). In comparison of the full workflow (feature selection and classification), no significant difference in ROC was found between centralized and distributed models for both studied endpoints (DeLong p > 0.05). In conclusion, both feature selection and classification are feasible in a distributed manner using radiomics data, which opens new possibility for training more reliable radiomics models.

Original languageEnglish
Article number4542
JournalScientific Reports
Issue number1
Publication statusPublished - 1 Dec 2020

Cite this