Knowledge-driven inference of Medical Interventions

Abstract

Physiological monitoring equipment routinely collects large amounts of time series patient data. In addition to influencing the treatment of a patient, this data is often used in medical research. However, treatment data (e.g. sedation) can be difficult to collect. In this paper we describe the AMITIE (Automated Medical Intervention and Treatment Inference Engine) system which infers a medical intervention from physiological time series data. The system comprises several domain ontologies and an algorithm to detect abnormal physiological readings and infer the subsequent associated medical intervention. To evaluate this approach we have applied AMITIE in the neuro-intensive care unit domain.