Benchmark Problems and Benchmark Datasets for the evaluation of Machine and Deep Learning methods on Photoplethysmography signals: the D4 report from the QUMPHY project

Urs Hackstein; Jordi Alastruey; Philip Aston; Ciaran Bench; Peter H. Charlton; Loic Coquelin; Nando Hegemann; Vaidotas Marozas

cs.LGApr 1, 2026

Benchmark Problems and Benchmark Datasets for the evaluation of Machine and Deep Learning methods on Photoplethysmography signals: the D4 report from the QUMPHY project

Authors: Urs Hackstein, Jordi Alastruey, Philip Aston, Ciaran Bench, Peter H. Charlton, Loic Coquelin, Nando Hegemann, Vaidotas Marozas,

Organizations: Technische Hochschule Mittelhessen - University of Applied Sciences, Giessen, Germany · King’s College London, London, United Kingdom · National Physical Laboratory, Teddington, United Kingdom · University of Cambridge, Cambridge, United Kingdom ·

Abstract

This report is part of the Qumphy project (22HLT01 Qumphy) that is funded by the European Union and is dedicated to the development of measures to quantify the uncertainties associated with Machine Learning algorithms applied to medical problems, in particular the analysis and processing of Photoplethysmography (PPG) signals. In this report, a list of six medical problems that are related to PPG signals and serve as Benchmark Problems is given. Suitable Benchmark datasets and their usage are described also.

Figures & tables

Figure 1: An exemplary photoplethysmogram (PPG) signal showing a pulse wave for each heartbeat. Pulse onsets, representing individual heartbeats, are shown as red circles. An inter-beat interval is labelled, corresponding to the time between consecutive heartbeats. Adapted from [ 1 ] by Peter H. Charlton - Own work, CC BY 4.0, https://commons.wikimedia.org/wiki/File:Photoplethysmogram_(PPG)_signal.png

Figure 2: The example of (a) arterial blood pressure (ABP) segment with labeled fiducial points (systolic (SBP) and diastolic (DBP) blood pressure) and (b) PPG segment.

Figure 3: The example of (a) ECG segment and (b) PPG segment with atrial fibrillation from TriggersAF dataset.

Figure 4: The example of (a) ECG segment and (b) PPG segment with premature beats and tachycardia from TriggersAF dataset.

Total: 1100 : 557M, 543F (49.4%F)
	Non Elevated		Elevated		Hypertensive
	213 (19.4%)		645 (58.6%)		241 (21.9%)
	73M / 140 F (65.7%F)		337M / 308F (47.8%F)		146M / 95F (39.4%F)
CVD	no cvd	cvd	no cvd	cvd	no cvd	cvd
	172 (76.5%)	53 (23.5%)	374 (59.0%)	260 (41.0%)	77 (32.0%)	164 (68%)
Gender	51 / 114	22 / 26	189 / 191	148 / 117	53 / 24	93 / 71

Table 1: Population characteristics for the BP classification and regression cohort - HW: Healthy Weight, OW: Overweight, OB: Obese. Age, weight and height are expressed as mean

\pm

standard deviation.

Total: 546
	$<$ 30	30-39	40-49	50-59	60-69	70+
Age group	57	205	146	118	16	4
	(10.4%)	(37.5%)	(26.7%)	(21.7%)	(3.0%)	(0.7%)
Height (m)	1.70 $\pm$ 0.10	1.71 $\pm$ 0.10	1.71 $\pm$ 0.10	1.70 $\pm$ 0.09	1.69 $\pm$ 0.09	1.72 $\pm$ 0.10
Weight (Kg)	70.41 $\pm$ 15.21	82.49 $\pm$ 21.44	81.59 $\pm$ 19.65	79.42 $\pm$ 18.65	70.32 $\pm$ 9.28	75.40 $\pm$ 16.39
SBP	114.16 $\pm$ 10.55	117.34 $\pm$ 10.53	118.38 $\pm$ 10.11	118.90 $\pm$ 10.49	125.49 $\pm$ 11.66	131.66 $\pm$ 5.68

Table 2: Population characteristics for the age regression cohort - HW: Healthy Weight, OW: Overweight, OB: Obese. Age, weight, height, SBP and DBP are expressed as mean

\pm

standard deviation. SBP and SBP values referred here are the ”baseline_sbp” and ”baseline_dbp” values used to divide the population into BP classes.

Figure 5: The changes of signals during sleep apnea episodes: (a) respiratory airflow, (b) finger PPG signal, (c) arterial blood oxygen saturation, SpO2, and heart rate, HR.

Figure 6: Idealised respiratory modulations of the PPG (left) and ECG (right). From top: no modulation, baseline wander (BW), amplitude modulation (AM), and frequency modulation (FM). Adapted from [ 65 ] and [ 66 ] by Peter H. Charlton - Own work, CC BY 3.0, https://commons.wikimedia.org/w/index.php?curid=104501363

Photoplethysmography

Photoplethysmography (PPG) uses changes in light absorption to measure blood-volume pulses for wearable and camera-based health monitoring.

See all 63 papers

Good Practice Guide for quantifying uncertainties for machine learning models applied to photoplethysmography signals
Jul 22, 2026P. Harris, C. Bench, M. Rinkevičius +6Uncertainty Quantification Photoplethysmography
PTB Physikalisch-Technische Bundesanstalt · NPL National Physical Laboratory · CMI Cesky Metrologicky Institut +5
PPG-Bench: A Unified Benchmark for PPG Foundation Models
Oct 9, 2026Arvasu Kulkarni, Mahir Jain, Aditya Ray Mishra +2Photoplethysmography Time-Series Foundation Models
Mannas AI
Deriving Health Metrics from the Photoplethysmogram: Benchmarks and Insights from MIMIC-III-Ext-PPG
Mar 23, 2026Mohammad Moulaeifard, Philip J. Aston, Peter H. Charlton +1Clinical Prediction Photoplethysmography
AI4Health Department, Oldenburg University, Oldenburg, Germany · Department of Data Science and AI, National Physical Laboratory, Teddington, United Kingdom · School of Mathematics and Physics, University of Surrey, Guildford, United Kingdom +1

Jul 22, 2026J/K move · Enter open · S save

Good Practice Guide for quantifying uncertainties for machine learning models applied to photoplethysmography signals

P. Harris, C. Bench, M. Rinkevičius, V. Marozas+5

PTB Physikalisch-Technische Bundesanstalt · NPL National Physical Laboratory · CMI Cesky Metrologicky Institut+5

Uncertainty Quantification Photoplethysmography

Open paper AskPDF arXiv

This Good Practice Guide presents work done in the QUMPHY project (Uncertainty quantification for machine learning models applied to photoplethysmography signals) that considered both machine learning and uncertainty quantification for problems which used photoplethysmography (PPG) signals from wearable devices as input. It provides high-level guidance on what types of machine learning model might be used and how different models compare when applied to both regression and classification tasks. It provides guidance on the implementation of different methods for uncertainty quantification, covering both model-dependent and model-independent techniques, and on the validation of the results provided by those methods. It also describes six benchmark problems together with pointers to different benchmark datasets for each problem. Software is described that can assist practitioners in implementing the methods described herein and there is a brief consideration of ethical issues. It concludes with a summary and recommendations.

Abstract

Photoplethysmography

Explore similar work

Good Practice Guide for quantifying uncertainties for machine learning models applied to photoplethysmography signals

PPG-Bench: A Unified Benchmark for PPG Foundation Models

Deriving Health Metrics from the Photoplethysmogram: Benchmarks and Insights from MIMIC-III-Ext-PPG