# Foundation models enable wearable signal screening for cardiovascular disease among people living with HIV

## TL;DR

AI and wearable sensors can help detect heart disease in people with HIV in low-resource settings, offering a cheaper alternative to expensive equipment.

## Contribution

Using pre-trained AI models with wearable sensors for heart disease screening in HIV-positive individuals in resource-limited areas.

## Key findings

- PaPaGei embeddings achieved an AUC of 0.769 for cardiovascular disease detection in a pilot cohort.
- Patients on dolutegravir-based regimens clustered in low-risk regions, while those with high cholesterol variability occupied high-risk areas.
- The method requires only local labels for training and avoids computationally intensive model fine-tuning.

## Abstract

Cardiovascular disease screening faces significant challenges in resource-limited settings, where infrastructure and computational constraints preclude advanced assessment. These constraints are particularly acute for people living with human immunodeficiency virus (HIV), who experience elevated cardiovascular risk yet often receive care in clinics without specialist diagnostic capacity. Pretrained physiological foundation models offer potential for low-cost screening using wearable sensors, though their applicability in resource-constrained settings remains unclear.

We evaluate pretrained physiological embeddings from foundation models for cardiovascular disease detection using photoplethysmography signals from 80 people living with HIV in Ho Chi Minh City, Vietnam. Of 80 participants, 13 (16%) had cardiologist-confirmed cardiovascular disease. We compare strictly zero-shot deployment (NormWear without local training) with frozen PaPaGei embeddings plus locally trained classifier, alongside traditional approaches.

Here we show that the PaPaGei-embedding approach achieves area under the receiver operating characteristic curve 0.769 (95% confidence interval: 0.70, 0.84) and average precision 0.489 (0.37, 0.61) in this pilot cohort, numerically higher than zero-shot NormWear (0.610; 0.226), principal component analysis features (0.651; 0.208), and supervised clinical models (0.744; 0.433). This approach requires local labels for classifier training but avoids computationally intensive foundation model fine-tuning. However, given the small positive class size (13 cases), these findings require validation in larger cohorts. PaPaGei embeddings capture clinically coherent structure: patients on dolutegravir-based regimens cluster in low-risk regions, while those with high cholesterol variability occupy high-risk areas.

These preliminary findings provide a potential methodological framework for deploying foundation models in resource-constrained settings, though adequately powered, multi-centre validation is essential before clinical implementation.

People living with human immunodeficiency virus (HIV) face higher risks of heart disease, yet many clinics in areas most vulnerable lack expensive equipment for heart screening. We tested whether artificial intelligence combined with low-cost wearable pulse sensors could help identify heart problems in 80 people living with HIV in Vietnam. Thirteen participants had confirmed heart disease based on specialist assessment. We found that using pre-trained artificial intelligence models with these simple sensors successfully identified most heart problems without requiring powerful computers or extensive training data. The mathematical predictive model predicted high risk of heart disease among participants with known risk factors, such as changes in cholesterol. While these early findings need testing in larger groups, this approach could help clinics with limited resources screen patients for heart disease using affordable technology instead of expensive specialist equipment.

Mesinovic et al. use AI foundation models to implement heart disease screening in people living with HIV in Vietnam, relying on low-cost wearable sensors. This approach achieves promising discriminative ability and requires only lightweight local training, potentially enabling affordable screening in resource-limited clinics.