Arrow Left Health research

Use case: meerCAT: A Survival Prediction Model for NSCLC Patients Through Distributed Learning Across 3 Countries

The aim of this project is to develop a predictive model of survival at 2 years based on a large volume of historical patient data that serves as a proof of concept to demonstrate the distributed learning approach.

Medical relevance 

Tools for survival prediction for non-small cell lung cancer (NSCLC) patients treated with chemoradiation or radiation therapy have limited quality. Better survival prediction will help choosing treatment for lung cancer patients

Description

In an international consortium consisting of cancer centers in the USA, UK and Netherlands a distributed learning infrastructure was deployed. A novel prediction model based on Bayesian Networks was developed by machine learning and validated based on the data of these cancer centers. The project was conducted in 2016 and was the first time distributed learning was used with UK and USA cancer centers, by means of a commercial PHT implementation (from Varian Medical System “Varian Learning Portal”),.

Main results

A new Bayesian Network based prediction model for survival in lung cancer outperforming previous models.

Lessons learned

Bayesian networks can be learned in distributed manner. A distributed learning infrastructure is also acceptable and deployable in the USA

Follow up

Ann Arbor Michigan continues to drive this project and is focusing now on developing an ontology to support future projects

Project details

Project leader

Arthur Jochems
Funders Dutch Technology Foundation STW (duCAT)
Collaboration partners MAASTRO Clinic / University of Michigan, Ann Arbor, Michigan / The Christie NHS Foundation Trust, Manchester, UK / Varian

 

Contact person

References

https://doi.org/10.1016/j.ijrobp.2017.04.021
Developing and Validating a Survival Prediction Model for NSCLC Patients Through Distributed Learning Across 3 Countries
International Journal of Radiation Oncology, October 1, 2017 Volume 99, Issue 2, Pages 344–352

Share this page…