Early Diagnosis of Systemic Sclerosis in the General Rheumatology Clinic

Problem Addressed We aim to improve early diagnosis of systemic sclerosis (SSc) – a painful, disabling disease, with high mortality, affecting 20,000 people in the UK.

It is over 20-years since nailfold capillaroscopy – viewing capillaries with a microscope – was acknowledged as key to early diagnosis, but it is still not available in most UK rheumatology clinics. Consequently, opportunities for early diagnosis are missed, leading to poor patient outcomes and avoidable NHS costs.

We aim to bring reliable capillaroscopy to the general rheumatology clinic.

Background Most patients with SSc first present to their GP with Raynaud s phenomenon (Raynaud s), providing an opportunity for early diagnosis. Raynaud s is not, however, specific to SSc: around 5% of the population suffer from benign Raynaud s.

Capillaroscopy, to detect SSc-related microvascular abnormalities, is key to identifying the 1,200 UK patients with SSc, amongst the 16,750 with Raynaud s referred annually; it is not, however, available in most rheumatology clinics because imaging and interpretation are challenging.

We have shown experienced technicians can acquire diagnostic-quality images with a hand-held digital microscope, whilst artificial intelligence (AI) software can detect SSc-related vessel abnormalities with high sensitivity and specificity (90%).

Aims and Objectives We aim to build on these results, developing and evaluating a low-cost, intelligent capillaroscopy system to support non-specialists in clinic.

The specific objectives are to: develop AI software to guide inexperienced users in acquiring diagnostic-quality images; develop AI software to automatically analyse images and generate clinical reports; implement analysis and reporting as a cloud service; evaluate the developed system in rheumatology clinics; model how implementation would affect costs and patient benefits; prepare for implementation.

Methods Each objective is addressed by one work package, with additional pilot-deployment and technology- refinement packages. Patient and Rheumatologist Reference Groups will play a key role, shaping the research through co-creation workshops. AI components will use deep learning networks, structured to produce human-interpretable outputs.

Visual and aural feedback will be used to guide image acquisition. Clinical reports will provide an abnormality score and human-interpretable explanation. The cloud service will address technical, security and information governance issues. The developed system will be deployed and assessed in seven diverse rheumatology clinics.

Each will acquire images and receive automated reports for 45 patients, ranging from benign Raynaud s to confirmed SSc.

Performance will be evaluated by comparing AI-based abnormality scores with expert clinical grading; usability will be assessed using qualitative methods.

Decision-analytic modelling and cost-consequences analysis will be used to quantify the impact of adoption on patients and the NHS. We will develop a business model with spinout Capilytics and prepare for CE/UKCA submission.

Dissemination and Impact The research will be disseminated via publications, engagement with rheumatologists and patients, and preparing for implementation.

Potential impacts include: improved efficiency in assessing patients presenting with Raynaud s; improved quality of life/survival for patients with SSc, through early intervention; early reassurance for patients for whom SSc is ruled out; cost savings to the NHS; business growth. We expect to launch a pilot reporting service within a year of project completion.