Cancer researchers use Open Source to share algorithms

Detection and analysis of cancer improved by collaboration through an Open Source project

Published on: 17/12/2020
News

Researchers around the public German Cancer Research Center (DKFZ) have used the collaborative power of Open Source to build a decentralised tool analysing medical imagery while respecting strict data protection laws.

The tool, the Joint Imaging Platform (JIP) was developed under the leadership of image processing expert Klaus Maier-Hein and radiologist Heinz-Peter Schlemmer. They teamed up with a network of radiology and nuclear medicine departments within the German Consortium for Translational Cancer Research (DKTK), a network of 20 public academic institutions at 8 locations, primarily with the aim of conducting joint clinical trials.

The consortium cooperates on the analysis of indispensable analysis, for early detection, diagnosis, therapy planning, monitoring of individual therapy success and follow-up care of cancer patients. Improving this detection would make diagnosis and analysis of cancer easier and the researchers are conducting joint clinical trials in order to improve this.

Yet, these important clinical trials faced the issue of strict data protection laws applicable to medical data. Further use for research purposes is also closely regulated. While anonymisation in theory could alleviate concerns of eliminating the identifiability of datasets, in practice this turned out to be technically difficult without rendering the data unuseful.

So the researchers decided to “bring the algorithms and processing tools to the data, not the other way around”, as project leader Klaus Maier-Hein said. In the first phase of the development of the JIP, a uniform processing infrastructure is created at all participating sites. From there, the image analysis algorithms can be executed, trained and developed in a standardised manner. This makes the easy exchange and comparison of methods and results across sites possible.

JIP is made possible through the use of the Open Source toolkit for state of the art platform provisioning in the field of medical data analysis Kaapana. Kaapana is specifically built for AI-based workflows and federated learning scenarios in radiological and radiotherapeutic imaging. As the software is federated, the actual analysis is run locally in the sites and only the distributed method development itself is shared.

The project could also be used to improve diagnostics in other areas, such as histopathology and COVID-19. The plan is to make JIP available as Open Source in the future. This would make it possible for other scientists in different areas to easily build on the project results and expand its use.