Key result
Deep learning models predict arterial blood pressure from PPG, but heterogeneous validation limits clinical translation.
Why the study?
Estimating the full arterial blood pressure waveform from photoplethysmography provides more clinical information than cuff-based measurements, motivating a review of recent methodological progress and gaps limiting clinical translation.
This narrative review highlights rapid progress in deep learning methods (particularly U-Net, transformer, and autoencoder architectures) for estimating full arterial blood pressure waveforms from photoplethysmography, while emphasizing the urgent need for standardized benchmark datasets, transparent preprocessing, and validation criteria to enable clinical translation.
Requires standardized validation before clinical use; leaves open reproducible benchmarks for PPG waveform translation.
Accurate, continuous blood pressure (BP) monitoring is important for preventing, diagnosing, and managing cardiovascular disease. Estimating the full arterial BP waveform can provide more clinical information than cuff-based measurements that report only systolic and diastolic BP. This review summarizes recent progress in predicting arterial BP waveforms from photoplethysmography (PPG) and identifies the main methodological gaps that limit clinical translation. We reviewed research published since 2021 on PPG-based BP waveform estimation. We organized the literature around core pipeline elements, including open-source datasets, preprocessing and quality control, model architectures, and validation practices. We also surveyed commercially available technologies that measure BP waveform. Data cleaning approaches can be grouped into three main categories: (1) template-based pulsation comparison, (2) feature extraction paired with classical machine learning, and (3) neural network–based quality assessment. For waveform prediction, recent work is dominated by deep learning, particularly U-Net–based architectures, autoencoder families, and attention-based models. Progress in PPG-based BP waveform prediction is rapid, but comparability and reproducibility remain major barriers. Benchmark datasets with consistent acquisition protocols, transparent preprocessing, and standardized validation criteria are needed to support fair comparison.
No takes yet. Share an insight, caveat, or question.
Olszewski et al. (2026) conducted a review in Continuous arterial blood pressure waveform monitoring. Photoplethysmography-based deep learning models (U-Net, autoencoders, and attention-based architectures) vs. Invasive arterial blood pressure catheterization was evaluated on Arterial blood pressure waveform and SBP/DBP estimation accuracy (Mean Absolute Error / Root Mean Squared Error). Photoplethysmography-based deep learning models predict arterial blood pressure waveforms with SBP mean absolute errors of 1.42 to 12.08 mmHg, though lack of standardized validation limits translation.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: