AI-guided point-of-care ultrasound to diagnose deep vein thrombosis in primary care

Problem

The National Institute for Health and Care Excellence reports that deep vein thrombosis (DVT) has an annual incidence of 1–2 per 1000. Currently, diagnostic scans for DVT are primarily performed by specialists in hospitals. Elderly patients, individuals with multiple chronic health conditions, and those experiencing significant mobility limitations are at heightened risk for developing DVT and find it difficult to get to hospital for investigation. The emergence of point-of-care ultrasound (POCUS), coupled with handheld ultrasound probes and artificial intelligence (AI)-applications to guide non-experts in ultrasound, could empower any healthcare professional to perform DVT scans. The implementation of AI-guided POCUS has the potential to expand local diagnostic capabilities beyond secondary care, reaching underserved patient groups. The present project aims to evaluate the accuracy and acceptability of AI-guided DVT diagnosis in primary care

Approach

A diagnostic test accuracy aims to estimate the sensitivity and specificity of AI-guided scans among 500 individuals suspected of DVT. This involves AI-guided scans performed by Healthcare Assistants (HCAs) followed by a standard scan conducted by sonographers at a primary care DVT clinic. Participants will be invited to complete a patient satisfaction survey after receiving both scans, evaluating their satisfaction with the AI-guided POCUS scan. The study will use semi-structured interviews to explore the nuances of acceptance and potential resistance towards AI-guided DVT diagnosis. This phase engages a subset of participants and the HCAs, encompassing 25 participants.

Findings

This study has successfully completed the training phase for Clinical Research Network research staff and HCAs responsible for conducting the index scan. This marks the commencement of our pilot phase, with recruitment scheduled to start in February 2024. We anticipate presenting as study update by the time of the conference.

Consequences

Incorporating an AI-guided DVT diagnostic procedure could improve local diagnostic capabilities and broaden access to underserved patients through availability of point-of-care diagnostic. This could be particularly advantageous for individuals with multiple chronic health conditions or limited mobility. It could accelerate diagnostics, providing timely access to treatment and potentially reducing severe complications such as pulmonary embolism or post-thrombotic syndrome. Furthermore, this approach holds promise in reducing the costs associated with specialist scans and referrals incurred by National Health Service. It is important to note that the specific outcomes of this research are not yet known, but if successful, the study could potentially enhance DVT diagnosis in primary care.PPI played a crucial role in designing participant materials, shaping the lay summary for funding applications, and influencing interview questions. This ensures the project remains patient-centred and ethically robust.