AI has big potential for healthcare and is expected to be the main disruptive force in the next years. However, for AI to leverage its full potential for healthcare, patients, doctors, etc. have to trust decisions supported by AI. In short, we need trustworthy AI for healthcare. But what does that mean and how can you design trustworthy AI solutions for healthcare? This is what Prof. dr. Willem Waegeman (UGent), dr. Bas Jansen (Omina Technologies) and Prof. dr. Sofie Van Hoecke (Imec, UGent) will talk about. dr. Anita Prinzie (Omina Technologies) will be the moderator.

When: March 15th 2022

Session: Trustworthy AI in Healthcare, Track A, 2.45pm to 3.30pm

Where: Sint Pietersabdij, 9000 Gent, Belgium

Fee (full day): 190 Euro excl. VAT

Register here.

Often humans do not trust machine learning systems. This distrust can be due to various reasons, such as noisy data, too few data, lack of generalization beyond the training data, existence of biases in the data, and a lack of good communication between data scientists and domain experts. In this talk I will discuss these challenges using use cases from the health domain. I will also briefly present some technical solutions to overcome these challenges.  

Inflammatory Bowel Disease (IBD) is an auto-immune disease affecting ~1.3% of the US population, which requires lifelong treatment and can have a big impact on a patient’s quality of life. Therefore, we performed an in-depth exploration of previously published and dormant IBD-BIOM datasets.

A non-invasive biomarker was identified that significantly outperformed the clinical standard (CRP), using an inherently interpretable AI model, with a hazard ratio of 25.91 vs. 9.0. However, while patenting our invention it became apparent that an interpretable and/or an explainable model, is not perse understandable.

The goal of this talk is to gain more insight into the context, profile and lifestyle of people in order to identify triggers and behaviors linked to stress, depression and/or migraines. Identifying these triggers allows the people to make adaptations to their lifestyle to improve their health.

First the differences between and benefits of data-driven versus knowledge-driven AI are explained.

Next, we show how both approaches can be combined to design a health management solution that enables the shift from subjective self-reporting towards objective continuous monitoring, empowering both patients and clinicians. To do so, we designed different machine learning algorithms to assess the activities (e.g. sedentary, walking, running, lying, commuting), sleep (e.g. duration, quality) and emotional state of the user (e.g. stress) based on physiological collected data through wearables. Semantic technologies are used to do an automatic migraine classification, while semantic rule mining is used to automatically detect the triggers. Dynamic dashboards are finally designed that use semantic reasoning to visualize whatever the clinician needs without configuration to support the clinician in trustworthy decision making.