Key result
PPG-based DBP prediction fails to generalize to unseen subjects despite low internal error.
Why the study?
PPG-based arterial blood pressure models are usually tested on data from the same dataset and subjects, which may overestimate accuracy in real-world use cases.
Does a regression model based on single PPG signals accurately predict diastolic blood pressure in unseen subjects?
Does a regression model based on single PPG signals accurately predict diastolic blood pressure in unseen subjects?
Machine learning models for cuffless blood pressure estimation using PPG signals may severely overestimate their accuracy in cross-validation and fail to generalize to unseen subjects.
Answers are grounded in this paper and cite their sources.
Cuffless diastolic BP models from single PPG signals warrant caution in unseen subjects; leaves open whether generalizable features or validation methods can be developed.
Photoplethysmography (PPG) sensors integrated in wearable devices offer the potential to monitor arterial blood pressure (ABP) in patients. Such cuffless, non-invasive, and continuous solution is suitable for remote and ambulatory monitoring. A machine learning model based on PPG signal can be used to detect hypertension, estimate beat-by-beat ABP values, and even reconstruct the shape of the ABP. Overall, models presented in literature have shown good performance, but there is a gap between research and potential real-world use cases. Usually, models are trained and tested on data from the same dataset and same subjects, which may lead to overestimating their accuracy. In this paper: we compare cross-validation, where the test data are from the same dataset as training data, and external validation, where the model is tested on samples from a new dataset, on a regression model which predicts diastolic blood pressure from PPG features. The results show that, in the cross-validation, the predicted and the real values are linearly dependent, while in the external validation, the predicted values are not related to the real ones, but probably just through an average value.
No takes yet. Share an insight, caveat, or question.
Żyliński et al. (2023) studied Blood pressure monitoring (n=7). Regression model for DBP prediction from PPG features vs. Cross-validation (random split) was evaluated on Mean Absolute Error (MAE) of Diastolic Blood Pressure prediction. A regression model predicting diastolic blood pressure from PPG features showed low error in cross-validation but poor generalization on unseen subjects.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: