Explainable AI for Sleep Medicine

Contact person: Thomas Plagemann
Keywords: Explainable AI, time-series data, sleep-related respiratory disorders
Research group: Analytical Solutions and Reasoning (ASR)
Department of Informatics
 

Sleep-related respiratory disorders are prevalent conditions that significantly impact quality of life, overall health, and life expectancy. The diagnosis and treatment of these disorders involve monitoring physiological signals during sleep, with the analysis of sleep data traditionally conducted by trained human experts. While AI holds immense potential to assist sleep specialists, ensuring the explainability of AI models is crucial for its effective integration into clinical practice.

Addressing the challenges in this domain requires overcoming several scientific hurdles, like:

  • Balancing model performance with explainability
  • Explaining AI findings to domain experts, patients, and policymakers
  • Diverse data types like time-series data from sensors, electronic health records, and audio/video recordings
  • Limited data availability due to privacy concerns and costs of clinical studies

Building on prior and ongoing projects with sleep medicine researchers, we provide a highly interdisciplinary research environment to address critical societal issues related to AI for sleep-related respiratory disorders.

Research topics:

  • AI analysis of sleep monitoring with explanations for health personnel and patients with focus on data quality and sensor technology
  • Personalization of mechanical ventilator settings for patients with respiratory disorders based on xAI and human expert collaboration

Applications:

  • Sensor technology, data quality, and xAI
  • Human and xAI loop to configure mechanical ventilators        

External partner:

  • Oslo University Hospital (OUH)