Implementation and Evaluation of Mobile AI and healthcare applications
At the Center for Digital Health Interventions (CDHI), we develop novel applications at the intersection of medicine, artificial intelligence, and ubiquitous computing. In multiple projects, we employ mobile devices, wearables, and portable sensors (breathing, sweat, audio, images, urine, …) to measure and monitor health-related biomarkers of patients and individuals. Based on the recorded data, we develop machine learning algorithms that allow us to assess and predict the health state of patients for different diseases and in multiple scenarios (e.g., prediction of asthma control scores, chronic heart failure, biological age, etc.). We are working on closed-loop projects, in which we first collect data from the patient, analyze it and afterward provide a use for the patient in form of applications (e.g., on Smartphones), that integrate and use the implemented machine learning algorithms. For this purpose, we are developing CLAID (“Closing the Loop on AI and data collection”), an open-source cross-platform framework that allows to combine of arbitrary sensors (microphone, smartwatches, breath analyzation, spirometry, …) and connect (hardware-accelerated) machine learning algorithms (running either directly on the smartphone or on a server), in order to build applications for various medical use cases in a highly generic manner. In this realm, we have multiple possible projects for Bachelor and Master Thesis, such as (but not limited to): Prediction of biological age and deterioration from blood serum- and DNA biomarker data Cross-platform Visualizations for Mobile AI and Health Applications Prediction of biological age and deterioration from blood serum- and DNA biomarker data **Please see details in the attached PDF**
Master thesis at ETH Zurich: User churn prediction with mobile app data, systematic review
As part of our ambitions to predict and prevent dropouts in Digital Health Interventions, our project aims at understanding which measurable app parameters and which (machine learning) methods have the most promising potential in predicting user churn before it occurs.
Master thesis at ETH Zurich: Dropout Prediction in Digital Health Interventions
We aim to identify factors influencing non-adherence and explore machine learning methods that can effectively predict dropouts based on longitudinal data from real-world mHealth apps.

Powered by  SiROP - the academic career network