Explainable AI framework for Obstructive Sleep Apnea Prediction

Develop an explainable AI framework to predict obstructive sleep apnea

Predicting the physiological state of the human body from time-ordered sensor measurements is an important trend in health. Sensor technology can help monitor, inform and notify not only users and caregivers, but provide healthcare providers with actual data to identify issues before they become critical or to allow for earlier intervention. Deep learning models on labeled sensor data have proven to be highly effective in automatically learning complex nonlinear mappings from labelled sensor data to accurately predict disease states. These systems are beginning to have an impact in helping improve the workload of clinicians by potentially reducing medical errors and also by enabling patients to interpret their own data. However, the lack of transparency of how deep learning models make predictions is a serious limitation for them to be trusted and understood by humans. This project will develop an explanation framework on deep learning models that predict obstructive sleep apnea (OSA) by leveraging non-linear time series analysis.


Using a deep learning model to predict sleep apnea as input generate explanations for its prediction based on non-linear time series analysis.

Learning outcome

  1. Formulation of research questions
  2. Experience with deep learning models
  3. Introduction to nonlinear time series analysis


  • Bachelors in computer science/electrical engineering/scientific discipline with a good amount of data analysis component
  • Good knowledge of Python programming
  • Interest in interdisciplinary research


Collaboration partners

  • LHL Gardemoen


E. Bradley and H. Kantz, “Nonlinear time-series analysis revisited,” Chaos , vol. 25, no. 9, p. 097610, Sep. 2015.

E. J. Topol, “High-performance medicine: the convergence of human and artificial intelligence,” Nat. Med. , vol. 25, no. 1, pp. 44–56, Jan. 2019.

P. Voosen, “The AI detectives,” Science , vol. 357, no. 6346, pp. 22–27, Jul. 2017.

Contact person