Internship: Distributed Survival Analyses for Cancer Research

23-11-2018
Distributed Survival Analyses for Cancer Research ​ An Internship or Graduation Project at the Netherlands Comprehensive Cancer Organisation

Background

IKNL, the Netherlands Comprehensive Cancer Organisation, is dedicated to continuously improve the care for Dutch cancer patients. To this aim, IKNL maintains the Netherlands Cancer Registry (NCR), a population-wide registry containing diagnosis and treatment data of nearly all cancer cases in the Netherlands since 1989. With over 2.5 million patients included in this database, the NCR may be of sufficient volume to reason about common cancer types. However,  about 20% of the patients today are diagnosed with a rare cancer, where data about disease, treatment and outcome may be too limited to conduct epidemiological analyses. In such cases, it is beneficial to combine NCR data with inter-institutional and international registries.  However, due to laws and privacy concerns, it may be difficult or even impossible to merge such data sets.

Project Description

IKNL developed an initial prototype system called distributedlearning.ai, which implements a distributed Cox Proportional Hazard survival analysis.[1] In this paradigm, privacy-sensitive data are kept at the institutions and only aggregated data are shared. Our current implementation, using second order derivatives, iterates rapidly, but at a high cost in time and memory at each iteration.

 The proposed project(s) would focus on the expansion of the distributedlearning.ai platform by designing, implementing and testing alternative distributed data analysis methods. On the one hand, we are interested to research, implement, and evaluate distributed Cox models using alternative, more efficient methods. On the other hand, we are interested to expand our toolset of commonly used methods in epidemiology (e.g., logistic regression, Kaplan-Meier analysis).

 

We are looking for

Students with a background in statistics and/or numerical computing, with an interest in healthcare. You have experience with R or Python. You are a problem solver and like to both thoroughly analyze and experiment. You dare to inspire people to think differently. You like discussing your ideas with future end-users: researchers in cancer epidemiology.

We are offering

The opportunity to contribute to science that may improve cancer care In a fun and stimulating working environment in our office in Eindhoven. You will have access to all stakeholders and data involved to make your project a success, good supervision, nice colleagues, and reasonable coffee. We value freedom, own ideas, and initiative.

Contact and more information

Gijs Geleijnse

g.geleijnse@iknl.nl

+31 6 418 618 53

 

 



[1] Lu et al., https://www.ncbi.nlm.nih.gov/pubmed/26159465

volg ons: