Evaluation of HRV–PRV agreement with autoencoder-based artifact mitigation in resting and dynamic conditions

Funding Industrial PhD Program (grant no. DWD/7/0063/2023) awarded to Wroclaw University of Science and Technology and implemented by Bartosz Olszewski as part of his doctoral training. SONATA 18 UMO-2022/47/D/ST7/00229 National Science Centre, Poland SONATA-BIS UMO-2013/10/E/ST7/00117 National Science Centre, Poland Acknowledgments The authors declare the following potential competing interests. The work described in this paper is related to the European patent application EP25201365.1 and has been implemented in Cardiolens FFR-CT Pro®. Bartosz Olszewski is affiliated with Hemolens Diagnostics Sp. z o.o., which develops Cardiolens FFR-CT Pro®. We would like to thank Krystian Gruszczynski, Msc Eng, Piotr Urbanski, MD and Tomasz Szczepanski, MD PhD for their assistance in laboratory measurements. General information: This database contains five files: a) FILE 1 (“database_1_long_recording_baseline.csv”) Fifteen subjects (median age: 21 (19-24) years; 8 men, 7 women) were included in the analysis of the impact of encoder-based cleaning procedure on PRV metrics (derived from continuous noninvasive arterial blood pressure, CNBP). All volunteers were measured at the Neuroengineering Laboratory BrainLab at Wroclaw University of Science and Technology (WUST) between October 2023 and January 2024; Bioethical Committee agreement: KB–179/2023/N. b) FILE 2 (“database_2_baseline.csv”) Fifty-three subjects (median age: 24 (22-25) years; 23 men, 30 women) were included in the analysis of the impact of encoder-based cleaning procedure on PRV metrics (derived from continuous noninvasive arterial blood pressure, CNBP). All volunteers were measured at the Neuroengineering Laboratory BrainLab at Wroclaw University of Science and Technology (WUST) between October 2014 and June 2015; Bioethical Committee agreement: KB–170/2014. Measurements were performed during the baseline condition within 5 minutes. c) FILE (“database_2_squad_and_stand.csv”) Fifty-three subjects (median age: 24 (22-25) years; 23 men, 30 women) were included in the analysis of the impact of encoder-based cleaning procedure on PRV metrics (derived from continuous noninvasive arterial blood pressure, CNBP). All volunteers were measured at the Neuroengineering Laboratory BrainLab at Wroclaw University of Science and Technology (WUST) between October 2014 and June 2015; Bioethical Committee agreement: KB–170/2014. Measurements were performed during postural changes (squat–stand, one minute per position) over a 5-minute period. d) FILE 4 (“database_2_squad.csv”) Fifty-three subjects (median age: 24 (22-25) years; 23 men, 30 women) were included in the analysis of the impact of encoder-based cleaning procedure on PRV metrics (derived from continuous noninvasive arterial blood pressure, CNBP). All volunteers were measured at the Neuroengineering Laboratory BraiLab at Wroclaw University of Science and Technology (WUST) between October 2014 and June 2015; Bioethical Committee agreement: KB–170/2014. Measurements were performed during postural changes (all squat epochs). 3) FILE 5 (“database_2_stand.csv”) Fifty-three subjects (median age: 24 (22-25) years; 23 men, 30 women) were included in the analysis of the impact of encoder-based cleaning procedure on PRV metrics (derived from continuous noninvasive arterial blood pressure, CNBP). All volunteers were measured at the Neuroengineering Laboratory BrainLab at Wroclaw University of Science and Technology (WUST) between October 2014 and June 2015; Bioethical Committee agreement: KB–170/2014. Measurements were performed during postural changes (all stand epochs). Detailed description of files: Metrics derived from the electrocardiogram (ECG) were labelled as "HRV" Metrics derived from the continuous noninvasive arterial blood pressure (CNBP) were labelled as "PRV" Metrics derived from the continuous noninvasive arterial blood pressure (CNBP) after encoder-based cleaning procedure were labelled as "PRV_CLEAN" Columns include: ID: anonymized participant label Demographics: age, sex Device: device used for ECG recordings Autonomic Nervous System Metrics: Standard time-domain, frequency-domain, and nonlinear metrics, including: Time-Domain Metrics: meanNN – Mean interval between normal R-peaks SDNN – Standard deviation of R-R intervals RMSSD – Square root of the mean squared successive differences between adjacent R-R intervals pNN20, pNN50 – Proportion of R-R intervals greater than 20 ms or 50 ms, respectively Frequency-Domain Metrics: LFn, HFn – Normalized power spectral density in the low-frequency (LF, 0.04–0.15 Hz) and high-frequency (HF, 0.15–0.40 Hz) ranges, obtained by dividing the respective power spectra by total power (TP, 0.04–0.40 Hz) Nonlinear Metrics: SD1 – Poincaré plot standard deviation perpendicular to the line of identity SD2 – Poincaré plot standard deviation along the line of identity