Developing more effective methods for treating cancer patients by analysing as much patient data from different sources as possible, but without having to share sensitive private data. This is now possible, thanks to the use of cryptographic techniques and artificial intelligence by the Netherlands Comprehensive Cancer Organisation (IKNL) and TNO.
In order to treat cancer patients of the future as well as possible, researchers must have access to relevant patient data that they are then able to combine. Examples that come to mind include information about the patient’s other illnesses, hereditary conditions, and what medication they are using. It is only then that researchers can determine what approach offers the best chances, the differences that exist between groups of patients, and how the illness can be detected at an earlier stage, or even prevented. Such data is often available, but not accessible in one place.
Data cannot simply be shared
‘The big challenge with analysing medical data is that it is very personal and cannot simply be shared,’ says TNO senior consultant Daniël Worm, the project manager of the research project. ‘This is why we have been looking at how we can gain information from data, but without making concessions to privacy. Our cryptographic researchers and data scientists, working with IKNL data scientists and clinical computer scientists, who already have a great deal of data at their disposal, have succeeded in doing so.’
Researchers need more and more data for each patient
IKNL is using the Netherlands Cancer Registry for monitoring the care of patients at national level. Gijs Geleijnse is a senior clinical data scientist at IKNL and chairman of the Personal Health Train (PHT) technical working group. He explains, ‘It is our mission to reduce the impact of the illness by examining patient data. The knowledge we gain means we can help choose more effective treatments and better understand what the consequences are.’
Researchers need more and more data
IKNL always supplies as little data as possible for scientific research, thereby minimising the likelihood of any specific individual being identified. However, treatment is becoming more and more personalised. Researchers therefore need more and more data for each patient. And because this data is typically found in multiple sources, there is actually a greater chance of it being traceable.
IKNL and TNO occupy a leading position in the Dutch PHT coalition. ‘It’s a sort of train that you run past hospital data, for example, so that the data can be analysed,’ Daniël continues. ‘Various technologies play a role in this. But one major challenge at present is being able to conduct privacy-friendly analyses of data concerning the same group of patients, which is spread across different organisations.’
The participants decide themselves who may see the results
Data remains encrypted using MPC
The technology known as secure multi-party computation (MPC), which has been around in universities for some time, has rapidly become more widespread in recent years and is now helping create a clearer picture. With this toolkit of cryptographic techniques, researchers can carry out analyses of data from different parties without having to share it. The data remains encrypted and the participants decide who may see the results. This means MPC is an essential technology that allows PHT to carry out analyses of data about the same group of patients in different organisations.
Whereas traditional cryptography is used for encrypting, exporting, and decrypting data, MPC makes it possible to direct artificial intelligence (AI) algorithms at data from multiple parties in a privacy-friendly way. This remarkable characteristic makes AI a more effective tool for oncology research.
‘The results from this project with TNO means more research will be possible’
More scientists at work
‘Investing in privacy technology means scientists can analyse large and sensitive datasets more securely,’ says Gijs. ‘Thanks to the results from this project with TNO, more research by more scientists will be possible.’
Oncology research is an extraordinarily important application. Will the newly developed technology offer new opportunities in the future, such as in the battle against coronavirus? Daniël believes it will, ‘AI is already being used in the fight against coronavirus. If you need to learn more about how all kinds of factors affect coronavirus, then you would want to link up the data, but without undermining privacy. So ultimately, these techniques will have great potential when it comes to better understanding coronavirus or other viruses, as well as adding value to the healthcare system in general.’
Care to know more, or looking to work in partnership?
TNO and IKNL intend to continue to work in partnership on privacy-friendly data technology and AI in the next few years, and to look for other partners to join them. ‘I invite readers wishing to know more about these cutting-edge technologies or who would like to partner us in developing open source software to please get in touch,’ Daniël concludes. ‘Together, we can then apply this technique even more widely.’
The project is being funded in part by the Appl.AI programme.