Young HSBI mathematician investigates rare diseases

Some diseases are so rare that only one in two million people are affected. It often takes years before such a “rare disease” can be diagnosed and treated. How the patients’ diagnostic journey can be shortened with mathematical help, Chiara Freitag explored in her bachelor’s thesis at Hochschule Bielefeld – University of Applied Sciences and Arts (HSBI). With Freitag’s findings, a tool is now being developed at the renowned Fraunhofer Institute for Techno- and Economic Mathematics (ITWM) in collaboration with the pharmaceutical company Chiesi to assist clinicians in diagnosis. And Chiara Freitag is involved as a research assistant—and, in addition, she is pursuing her master’s degree.

But first things first: at the start, Chiara Freitag completed an internship at Fraunhofer ITWM in Kaiserslautern. There, the development of the online support tool was being prepared. The tool should, after entering a series of symptoms, calculate the probabilities for the presence of all possible rare diseases.

One of the major challenges here is incomplete data about the presented symptoms. The mathematical — more precisely, stochastic — model used must therefore be able to handle uncertainties such as incomplete information, but also integrate newly added information as it becomes available. With the right algorithm, it should then make as reliable statements as possible about likely diseases. In addition, the model and algorithm should be able to encode prior knowledge, e.g., the known probabilities of specific symptoms for certain diseases.

This is a case for the so-called Bayesian statistics, a mathematical approach that also forms the basis for many artificial intelligence applications and that meets the criteria mentioned above. The model can be continually adapted and refined as more data becomes available during a diagnosis. This increases its diagnostic power.

To achieve this, Chiara Freitag has combined her Bayesian model with Monte Carlo algorithms. “These algorithms are especially helpful for complex probability distributions, such as those found in rare diseases. They enable estimating unknown model parameters from available data,” she emphasizes. The model learns to increasingly blend the newly entered patient information with the existing model data and to estimate the probability of a possible rare disease more accurately.

To the pride of her supervisor, Prof. Dr. Jörg Horst of HSBI, Chiara Freitag’s work at Fraunhofer ITWM is indeed seen as a substantial step toward realizing the diagnostic tool that can bring faster help to the affected individuals. While pursuing her master’s degree, she will continue to work on the project.

Click here for more information.