Christian Dorfer

Rapid and accurate diagnosis of disease is fundamental for the effectiveness of a treatment. In most cases, the diagnostics approach is based on anamnesis, the medical doctor interviewing the patient about the history of the illness and the symptoms, paired with laboratory tests. While the number of available tests has increased over the years, choosing the relevant tests and interpreting them can be a challenging and time consuming task.

Christian Dorfer and his team are building a copilot for medical doctors that guides them in the diagnosis finding process. To do so, they use data from over 120’000 patient records and apply machine learning. With a given set of symptoms, age, gender and laboratory results of a patient, their software predicts the most likely diagnoses and suggests further tests to verify these diagnosis predictions. The goal is to speed up the diagnosis finding process and make it more accurate for the benefit of the patient. A more accurate diagnosis could prevent unnecessary treatments which in turn lowers healthcare costs.

A major effort during the Entrepreneur Fellowship is the labeling of laboratory results, symptoms and diagnoses. For this work, the team collaborates with medical students and doctors and the University Hospital Zurich. The access to a high data volume is crucial to reach a prediction quality needed in the daily clinical routine.

Affiliation: Stefan Balabanov

Start date: 10/2023