Alsuleman, M; Duncan, P; Thompson, A (2025) Screening of Atrial Fibrillation using Wearable PPG Devices – a Trustworthy and Safe AI Life Cycle Case Study. NPL Report. MS 61

Preview

Text (NPL Report MS 61) MS61.pdf - Published Version Download (1MB) | Preview

Abstract

NPL recently developed a Trustworthy and Safe AI Life Cycle (TSALC) which links the risks associated with AI systems to the software development process and AI trustworthiness metrics. In this report we apply the principles of the TSALC to a case study: the detection of atrial fibrillation (AF) using a wearable AI-based medical device equipped with a photoplethysmography (PPG) signal sensor. A hypothetical scenario is proposed in which the device is used for clinician-initiated screening of patients suspected of having AF, in order to inform prioritisation of diagnosis resources (for example ECG monitors). We describe a methodology for carrying out risk analysis and mitigation, inspired in part by NIST’s AI Risk Management Framework (NIST AI RMF). Assimilating input from application experts drawn from academia, industry and the NHS, we identify specific risks and mitigations. Based upon these risks and mitigations, we outline the most relevant aspects of AI trustworthiness and the associated software engineering challenges. We also describe metrics which can be used to quantify some of these aspects of AI trustworthiness. As well as representing a useful case study in its own right, we expect that the methodology described here will be applicable to numerous other AI systems.

Item Type:	Report/Guide (NPL Report)
NPL Report No.:	MS 61
Subjects:	Mathematics and Scientific Computing > Software Engineering
Divisions:	Data Science
Identification number/DOI:	10.47120/npl.MS61
Last Modified:	26 Jun 2025 13:42
URI:	https://eprintspublications.npl.co.uk/id/eprint/10185