German Liver Foundation: On the trail of hepatitis C with artificial intelligence
05.04.2019 13:18
According to estimates by the World Health Organization (WHO), at least 70 million people worldwide are suffering from chronic hepatitis C. The viral infection can destroy the liver - usually unnoticed. In recent years, great progress has been made in the field of hepatitis C therapy. However, the disease can only be treated if it is recognized. In Germany, the number of people infected with the hepatitis C virus is estimated at 250,000, many of whom are unaware of their illness. The DETECT project of the German Liver Foundation uses neural networks to pursue a new approach to determining predictors of the presence of chronic hepatitis C. The results of the project have now been published.
In the DETECT project (predictor determination for undetected HEpatitis C virus infections), health data from more than 1.8 million insured people from the period from 2009 to 2014 were evaluated, which were provided by the private health insurance companies “Debeka” and “HUK-COBURG”. were made available in anonymized form. The responsible ethics committee assessed this data evaluation positively. The scientific project manager was Priv. Doc. Dr. Markus Reiser, chief physician of the Medical Clinic II of the Paracelsus Clinic Marl at the Vest Clinic. The data management was carried out by Jürgen Hirsch, Managing Director of Qyte Software & Solutions GmbH. The project was financially supported by AbbVie Deutschland GmbH & Co.KG, a partner of the German Liver Foundation.
The health insurance data was analyzed using artificial intelligence, i.e. a neural network, for signs of possible hepatitis C. In the first step, the data sets of the 2,544 known hepatitis C diseases were extracted and examined for hidden clues to the infection. This includes diagnoses and treatment of non-specific side effects that are more frequently observed in hepatitis C, but alone are not sufficient to express reasonable suspicion. Such suspicions could include fatigue, joint pain, thyroid disease, depression or diabetes. An artificial neural network is similar to a brain that absorbs information, evaluates it and sorts it on a three-dimensional map. With 1.8 million insured people and a large number of outpatient and inpatient diagnoses, medication prescriptions, sick days and other variables over a period of six years, the amount of data was extremely large and an artificial “high-performance brain” was required for this analysis.
“I find the research approach very exciting,” says project leader Priv. Doc. Dr. Markus Reiser, who has been working intensively with hepatitis C for more than 20 years. “Many people affected do not know that they are infected with the potentially deadly virus. The identification of predictors that indicate hepatitis C is an important contribution to curing these people and also to the planned elimination of hepatitis C in Germany.”
The central questions of the research project were: Can the known hepatitis C diagnoses be retrieved from the total data pool of the 1.8 million insured people with the help of artificial intelligence? And how many insured people show striking similarities in their health data compared to those suffering from hepatitis C? Particular attention was paid to the HCV infections diagnosed in the last year of observation, as these data could not have been influenced by the diagnosis.
The extensive analyzes that lasted weeks showed that all known hepatitis C virus infections were found using the neural network: They appeared at the defined location (in a so-called cluster) on the data map; The infections that were only diagnosed in 2014 were also detected by the neural network. In addition, 2,217 insured people were assigned to the cluster due to the similarity of their “socio-medical” data and therefore have a potentially higher probability of having as yet unrecognized hepatitis C. Due to a requirement from the ethics committee, all data had to be anonymized in order to protect the right not to know. This meant that contact and the offer of an HCV test were excluded, meaning that it could not be checked whether those potentially affected actually had hepatitis C. The study results have now been published in the Journal of Translational Medicine.
“With this study, the German Liver Foundation was able to demonstrate a new approach to identifying hepatitis C sufferers,” emphasizes Prof. Dr. Michael P. Manns, Chairman of the Foundation, explained the importance of the project. “I would like to thank the health insurance companies and our partner AbbVie, without whose support this project would not have been possible, as well as the project manager Priv. Doc. Dr. Markus Reiser and the data analyst Jürgen Hirsch for their extremely committed work.”