Background Current blood pressure (BP) measurement techniques reduce the complex Arterial Blood Pressure (ABP) waveform to simple systolic and diastolic values, ignoring the rich physiological information embedded in the continuous signal. The BP-WAVE research programme aims to use Artificial Intelligence (AI), specifically Biosignal Transformers, to uncover hidden waveform features predictive of cardiovascular events. To validate these AI models for clinical use, “ground truth” data linking these digital waveform features to actual structural vascular health is required.
Aims and Hypothesis The primary aim of this summer scholarship is to generate a preliminary “deep cardiovascular phenotyping” dataset for 10 participants to support the wider BP-WAVE study. We hypothesize that novel ABP waveform features extracted via transformer-based deep learning will demonstrate a significant correlation with established structural markers of vascular health, including carotid intima-media thickness and coronary calcium scores.
Objectives: The specific objectives are to:
Recruit 10 patients with established cardiovascular risk factors from the Hypertension Clinic lists.
Acquire high-fidelity, non-invasive ABP waveforms using a Finometer device and coordinate comprehensive vascular imaging (Carotid Ultrasound, Echocardiography, and CT Coronary Angiography).
Perform preliminary analysis using a pre-established Biosignal Transformer pipeline (pre-trained on MIMIC-III) to extract feature embeddings and explore their statistical correlation with the vascular imaging metrics.
Expected Outcomes This work is expected to establish a curated, multi-modal pilot dataset linking AI-derived waveform metrics to physical vascular pathology. This will provide essential validation data for the BP-WAVE programme, potentially advancing non-invasive cardiovascular risk prediction.