Interim assessment of medical informatics project DIFUTUREPeople suffering from rare diseases often face a major difficulty: Their illnesses occur so infrequently that only a handful of physicians are very familiar with them. This in turn makes the right diagnosis, not to mention successful treatment, very hard. Now researchers in the consortium "Data Integration for Future Medicine" (DIFUTURE) want to improve this situation using large volumes of data and Artificial Intelligence.
DIFUTURE is a consortium of several universities and their respective university hospitals working together on solutions for the precision medicine of the future. After six years, the next phase of this Germany-wide project is now beginning, under the leadership of the Technical University of Munich (TUM).
Prof. Martin Boeker and his colleagues are collecting and consolidating as much medicinal data from healthcare and medical research as possible. Given the overall German population of 82 million people, several thousand individuals nevertheless suffer from even highly rare diseases. Here Artificial Intelligence and in particular machine learning can be used to possibly derive indications of which treatments may have a chance of success, based on the patients’ health data und demographic information.
More data for treatment of illnessesProf. Boeker is a medical informatics specialist at TUM with a background in both medicine and information sciences; he is also the head of the DIFUTURE project, including seven consortium partners. "Our joint goal is to provide more and better data for the investigation and indirectly for the treatment of diseases. Better data make better research possible and can also be used in new Artificial Intelligence methods: Data can save lives."
Based on this data, intelligent software is to give physicians advice on the most promising therapies. In order to do this, the specialized algorithms capable of learning will sift through the largest possible data inventories and analyze for example what has worked well for other patients in similar medical situations. They incorporate simple categories such as gender and age as well as for example information on previous illnesses, therapeutic failures and genetic components. "This can reveal new and previously unnoticed correlations, for example in order to propose individualized treatments for patients."
The greatest challenges are data quality and the compilation of data to build databases which can be analyzed - developing generally accepted standards for this process is one of the main assignments of DIFUTURE. "In the meantime we’ve established the corresponding organization and technology with the associated contracts and processes and have made them legally sound. That’s a great success for our project," Boeker says.
Software will learn to understand medical reportsHis own research concentrates primarily on automatically filtering out the most important information from individually formulated medical reports and other medical texts, some of which can be very long. This information is then made available for deeper analysis. "Physicians will still prefer to document their findings and information on patients in text form, instead of simply checking a box next to a predefined category," Boeker observes. "But this makes access for data researchers - and thus the utilization of the information for science and improving treatment - very cumbersome. Our team, which also includes a group of young researchers focusing on automated language processing, is working on a future solution using intelligent text recognition algorithms."
Data protection as a central elementOne essential aspect for the scientists is data protection, which is regulated at a European level by the General Data Protection Regulation (DSGVO). In order to ensure the privacy of the data, various methods such as anonymization and distributed computing are used in which the data don’t leave the respective hospital. Other analyses require the prior consent of the affected party. Boeker reports a great openness on the part of patients: "They know that they can help other ill people with their data - and that they themselves will in turn benefit from a similarly generous attitude on the part of other people."
Precision medicine of the futureThe higher-level framework of the DIFUTURE consortium is the Medical Informatics Initiative which has been funded by the German Federal Ministry of Education and Research (BMBF) with approximately 180 million euros over the past six years. In 2023 the second funding phase began, during which a similar funding amount will be made available to the numerous researchers involved, with the collaboration of the BMBF’s University Medicine Network ("Netzwerk Universitätsmedizin").
One major objective of future research will be for example advances in the fight against cancer. The Artificial Intelligence application will use data on the individual manifestations of molecular markers in patients’ cancer cells to derive highly specific treatment methods. "The methodology is currently being established and will go well beyond the standard diagnostics of the past," says TUM professor Boeker. "This is the basis of our support for the precision medicine of the future."