Using Artificial Intelligence to Manage New-onset Atrial Fibrillation in Patients with Critical Illnesses

Lead Research Organisation: University College London
Department Name: Institute of Health Informatics

Abstract

Brief Description of the context of the research including potential impact:
New onset atrial fibrillation (NOAF) occurs in up to 46% of critically ill adults on Intensive Care Units (ICUs) and may be associated with significant morbidity and mortality. AI and machine learning techniques facilitate the automated identification of NOAF episodes and of their duration, and thus the mining of large ICU electronic medical records to help us understand the temporal association between clinical state and AF initiation or recurrence.

Aims and objectives:
To create an AI system for identifying endotypes of NOAF in ICUs so that proactive and personalised interventions can be realised - accurately predict when a patient will develop atrial fibrillation in the ICU and the likely causes, which would suggest to clinicians how to prevent it from occurring. This may be achieved using real-time prediction on multimodal continuous monitoring data.

Research methodology:
Not all cases of NOAF in the ICU are the same, it is likely that the success of any intervention will depend upon the precise endotype of AF being treated. However, such endotypes (and the manner in which they should be managed) are currently unknown. As a result, management strategies for NOAF on ICU are untargeted and diverse. Machine learning techniques can facilitate the automated identification of NOAF episodes and of their duration, and thus the mining of large ICU electronic medical records and unsupervised clustering can be used to help us understand the temporal association between clinical state and AF initiation or recurrence. In this way, specific 'different endotypes' can be identified. Further AI/ML approaches (using clinical data and/or waveform analyses) can also be created to predict NOAF events and, by highlighting the key informing parameters, suggest therapeutic or mitigative strategies, the efficacy of which might then be tested.

EPSRC's strategies and research areas:
Artificial intelligence technologies including supervised and unsupervised learning, as well as clinical text mining for optimising treatment and management of NOAF.

Any companies or collaborators involved:
'Better Endotyping of AF Management Subgroups' (BEAMS) Consortium

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/S021612/1 01/04/2019 30/09/2027
2420472 Studentship EP/S021612/1 28/09/2020 30/09/2024 Tina Lan Yao