Rationale

It has long be assumed by clinicians that the shape of an aneurysm influences its risk of rupture. However, current risk prediction tools do not include detailed information from imaging data. Morphological features from vessel segmentation and haemodynamic features from computational fluid dynamic (CFD) analysis have shown differences between ruptured and unruptured aneurysms.

However, previous research is largely based on small, cross-sectional datasets which include imaging following rupture. These limitations make it difficult to apply their findings to clinical practice.

ROAR-FLOW will be the largest study of its kind, using imaging collected before rupture. By combining this with long-term follow-up, ROAR-FLOW will provide the clearest picture yet of how these features predict rupture. This research will help us move beyond current subjective methods by identifying precise changes in aneurysm size, shape, and growth that may signal higher rupture risk. This will support the development of personalised follow-up programmes that improve monitoring and decision-making for patients with unruptured aneurysms.

1. Predict rupture risk – use imaging to identify high-risk features to better select patients who require treatment. 

2. Quantify growth – develop precise and standardised measurements of growth to identify patients who need treatment or follow-up.

3. Effectiveness of follow-up imaging – by compare rupture rates between growing and stable aneurysms we will develop personalised follow-up programmes.

We have developed a secure, automated pipeline to transfer and anonymise imaging data from patients included in the ROAR study. Scans are sent to University Hospital Southampton via the Sectra IEP and stored in a research database (XNAT). Using a validated workflow, we will automate the analysis of aneurysm shape and blood flow using artificial intelligence (AI). Key shape and flow features will be extracted for rupture risk modelling, and explainable AI approaches will be used to group aneurysms into clusters to improve prediction. To overcome the limitations of subjective growth reporting, we will apply this automated workflow to analyse follow-up scans. Using AI approaches we will objectively assess changes in the aneurysms' size and shape as well as identify distinct growth patterns over time.