Background and purpose: Access to healthcare data is indispensable for scientific progress and innovation. Sharing healthcare data is time-consuming and notoriously difficult due to privacy and regulatory concerns. The Personal Health Train (PHT) provides a privacy-by-design infrastructure connecting FAIR (Findable, Accessible, Interoperable, Reusable) data sources and allows distributed data analysis and machine learning. Patient data never leaves a healthcare institute. Materials and methods: Lung cancer patient-specific databases (tumor staging and post-treatment survival information) of oncology departments were translated according to a FAIR data model and stored locally in a graph database. Software was installed locally to enable deployment of distributed machine learning algorithms via a central server. Algorithms (MATLAB, code and documentation publicly available) are patient privacy-preserving as only summary statistics and regression coefficients are exchanged with the central server. A logistic regression model to predict post-treatment two-year survival was trained and evaluated by receiver operating characteristic curves (ROC), root mean square prediction error (RMSE) and calibration plots. Results: In 4 months, we connected databases with 23 203 patient cases across 8 healthcare institutes in 5 countries (Amsterdam, Cardiff, Maastricht, Manchester, Nijmegen, Rome, Rotterdam, Shanghai) using the PHT. Summary statistics were computed across databases. A distributed logistic regression model predicting post-treatment two-year survival was trained on 14 810 patients treated between 1978 and 2011 and validated on 8 393 patients treated between 2012 and 2015. Conclusion:The PHT infrastructure demonstrably overcomes patient privacy barriers to healthcare data sharing and enables fast data analyses across multiple institutes from different countries with different regulatory regimens. This infrastructure promotes global evidence-based medicine while prioritizing patient privacy.

Objective To investigate the effects of Image Biomarker Standardisation Initiative (IBSI) compliance, harmonisation of calculation settings and platform version on the statistical reliability of radiomic features and their corresponding ability to predict clinical outcome. Methods The statistical reliability of radiomic features was assessed retrospectively in three clinical datasets (patient numbers: 108 head and neck cancer, 37 small-cell lung cancer, 47 non-small-cell lung cancer). Features were calculated using four platforms (PyRadiomics, LIFEx, CERR and IBEX). PyRadiomics, LIFEx and CERR are IBSI-compliant, whereas IBEX is not. The effects of IBSI compliance, user-defined calculation settings and platform version were assessed by calculating intraclass correlation coefficients and confidence intervals. The influence of platform choice on the relationship between radiomic biomarkers and survival was evaluated using univariable cox regression in the largest dataset. Results The reliability of radiomic features calculated by the different software platforms was only excellent (ICC > 0.9) for 4/17 radiomic features when comparing all four platforms. Reliability improved to ICC > 0.9 for 15/17 radiomic features when analysis was restricted to the three IBSI-compliant platforms. Failure to harmonise calculation settings resulted in poor reliability, even across the IBSI-compliant platforms. Software platform version also had a marked effect on feature reliability in CERR and LIFEx. Features identified as having significant relationship to survival varied between platforms, as did the direction of hazard ratios. Conclusion IBSI compliance, user-defined calculation settings and choice of platform version all influence the statistical reliability and corresponding performance of prognostic models in radiomics. Key Points • Reliability of radiomic features varies between feature calculation platforms and with choice of software version. • Image Biomarker Standardisation Initiative (IBSI) compliance improves reliability of radiomic features across platforms, but only when calculation settings are harmonised. • IBSI compliance, user-defined calculation settings and choice of platform version collectively affect the prognostic value of features.