CISPA expert for trustworthy AI helps with analysis of patient data

 He supports the other project partners with his expertise in trustworthy artificial intelligence, which is used in the analysis of patient data. The EU is funding the research project headed by Professor Dr. Heiner Wedemeyer, Director of the Department of Gastroenterology, Hepatology, and Endocrinology at Hannover Medical School (MHH), with a total of 6.75 million euros.

Hepatitis D is by far the most severe form of chronic viral hepatitis, often leading to liver failure, liver cancer, and death. It is caused by co-infection of the hepatitis B virus (HBV) and the hepatitis D virus (HDV). As many as 20 million people worldwide are infected with HDV. However, the disease still poses several mysteries to the medical community. For example, it is not yet known why up to 50 percent of those affected are spontaneously able to control the multiplication of hepatitis D viruses in the body. It is also unclear why some patients reach an advanced stage of liver disease or only some respond to antiviral treatment. The international research project "D-Solve" aims to change this. In addition to experts from MHH and CISPA, researchers from the Center for Individualized Infection Medicine (CiiM) - an institution of the MHH and the Helmholtz Centre for Infection Research (HZI) in Braunschweig - are also involved.

In the multicenter study, the researchers want to examine a large group of hepatitis D patients to understand better which personal characteristics determine the outcome of the infection. Because HDV is a rare disease, there is not much medical data on HDV-infected individuals at any given hospital or appropriate biobanks of tissue or blood samples from HDV patients. There is also a lack of a reliable animal model on which to study responses to the virus scientifically. "With the multicenter study, we can access data and biospecimens from more than 500 patients with HDV and search for biomarkers for the immune response to the viruses," explains Professor Wedemeyer. "Through this, we aim to develop a novel individualized approach to treatment against HDV that clearly defines which patients need to be treated quickly, how long the treatment should last, and what happens to those who don't respond well to antiviral drugs."

Patient data is key in developing personalized therapies. However, such information is highly sensitive, explains CISPA faculty member Dr. Yang Zhang. "To ensure that patient data in the project is kept secure and confidential, we will manage it in a decentralized manner. We will use so-called federated learning to process them collectively," explains CISPA researcher Dr. Yang Zhang. In federated learning, a special machine learning technique, the data is stored exclusively locally, not uploaded to a server, and merged as in centralized approaches. The project partners jointly train a machine learning model with the patient data they have collected, but without exchanging the data. This way, significantly more data security, and privacy protection are achievable.

At the end of the project, patient monitoring strategies and antiviral treatment approaches will be developed to reduce the burden of disease and improve patients' quality of life. The knowledge gained can then serve as a blueprint for personalized infectious medicine and help to bring other viral diseases under control.

(translated by Oliver Schedler)