. 2023 Dec 15;6:231. doi: 10.1038/s41746-023-00973-x

Table 2.

A summary table of clinical studies based on VCPM algorithm (The deadline of search is May 22, 2023).

First Author	Country	AI	Year	Subject (n)	Distance (cm)	Data length	Method and clinical condition	Vital-sign: ME ± SD (HR:bpm; RR:brpm)
Liu¹⁴⁷	China	✓	2023	405	50–150	Each 10 min	Unsupervised learning; ICU patients; 67.8 Y/O. (mean); RGB; Chest ROI; 1440 × 1080/1080 × 720 px, 30fps	RR: 2.8 ± 3.0
Barde¹⁵⁶	USA	×	2023	60	28–44	—	Atrial fibrillation patients; 67 ± 10 Y/O.; ≥50, 100, or 200 lux; Smartphone (Samsung S10 or S3), 30 fps	HR:0.05 ± 1.4
Batbayar¹⁴²	Japan and Mongolia	×	2022	301	40	—	154 COVID-19 patients and 147 healthy individuals; Thermal (9fps, 80 × 60 px) and RGB camera (30 fps, 424 × 240 px); Lie in bed	HR:-0.2 ± 2.3; RR:0.0 ± 1.6
Svoboda¹⁵⁷	Germany	—	2022	42	150	—	3D camera system containing RGB and NIR camera (950 nm); Newborn infants; 39.6 ± 1.3 weeks; 330.4 ± 13.9 lux	HR-3D:8.6 ± 4.4; 7D2HR-2D:3.0 ± 3.4
Allado¹⁵⁸	France	—	2022	963	70–100	Each 60 s	Patients; 56.6 ± 16; Uncompressed video; BMI concordance; Fitzpatrick skin color scale concordance	HR:-0.1 ± 4.1
Allado¹⁵⁹	France	—	2022	924	70–100	Each 60 s	Ditto	RR:-0.7 ± 3.5
Sahoo¹²⁷	India	✓	2022	19	—	19 * 10 mins	An end-to-end DL model integrates a non-learning based approach; NICU; 25 fps, 1920 × 1080 px; NIR and RGB	HR: 0.8 ± 3.5
Zeng¹⁶⁰	China	×	2022	10	focal length	—	Living-skin detection; Critically-ill premature infants; 160 × 90, 20 fps; RAW format video	HR:MAE = 5.68; RR:MAE = 10.82
Jorge¹⁸	UK	✓	2022	15	—	233.5 h	Post-operative ICU patients; 62.2 ± 12.4 Y/O.; Uncompressed 8-bit data (850 nm); 1024 × 768 px, 100 fps	HR:MAE = 2.5; RR:MAE = 2.4
Wang¹⁶¹	China	×	2022	10	—	each 10 min	ROI optimization using living-skin and respiratory maps; ICU Patients; CCTV camera in ceiling; 1920 × 1080 px, 25 fps	HR:1.7^△ ± 6.0; RR:1.6^△ ± 2.9
Ottaviani¹⁶²	Italy	×	2022	12	>16	—	Extract 3D displacements of chest depth images; Preterm infants RR monitoring; RGBD, 90 fps, 1280 × 720 px	RR:0.02 ± 0.51
Hajj¹⁶³	Canada	×	2022	4	—	—	NICU; Neonatal RR estimation with RGBD sensor (Intel SR300); Using automated ROI segmentation algorithm	RR:MAE =5.0
Pediaditis¹⁴⁹	Greece	✓	2022	5	180	25–35 h	Using Eulerian magnification and 3D CNNs; Sleeping; 67 ± 10.8 Y/O; RGB and IR camera; 1080 × 720 px, 25 fps	RR:MAE = 2.29
Nagy¹⁶⁴	Hungary	×	2022	10	80-150	240 h	ROI detection based on UNet++, but the principle method are not based on AI; NICU, GA:33.2 weeks (mean); Raw format, 500 × 500 px, 20 fps	RR: MAE = 1.09
Varma¹⁶⁵	India	—	2022	158	122–152	Each 30 min	Adult patients’ HR and RR measurement in hospital; 18–80 Y/O; IDS UI3060 and CCTV camera	HR:−1.27 ± 5.6; 7D3RR: −0.3 ± 3.06
Huang¹⁷	China	✓	2021	257	50–100	9.6 h	In-hospitalized neonates (0–7 days); DL based method to extract PPG waveform and HR; 640 × 480 px, 30 fps	HR:−0.21 ± 5.32
Chen¹⁶⁶	China	×	2021	9	26–36	4.21 h	Hospitalized newborns with motion artifacts; RGB camera; 640 × 480 px, 30fps; Mean GA: 36.6 weeks	HR-R:MAE = 3.4; HR-M:MAE = 4.3
Kyrollos¹⁶⁷	Canada	×	2021	5	—	—	Video Magnification based method; NICU; RGBD camera	RR: MAE=3.5
Khanam¹⁶⁸	Australia	×	2021	6	100–200	Each 10 min	Automatic ROI detection based on neural networks, but the principle method are not based on AI; NICU; Nikon D610 and D5300; 1920 × 1080 px, 30 fps	HR:0.44 ± 2.2; RR:0.71 ± 2.65
Laurie¹⁶⁹	Australia	×	2021	7	—	—	patients with acute mental; 12 bit GigE camera; 648 × 488 px	RR:0.11 ± 0.86
Lyra¹⁴⁸	Germany	✓	2021	26	—	—	Investigated skin temperature trend measurement and respiration-related chest movements; Using YOLO detector; ICU patient; IRT camera, 4fps; 382 × 288 px; Optical Flow	RR:−0.18 ± 2.8
Yu¹⁷⁰	Germany	×	2020	20	180	25*20 min	Patient HR and HRV monitoring with NIR and RGB camera	HR: 0.0 ± 0.7
Villarroel¹⁴³	UK	✓	2020	40	—	304.1 h	Patients undergoing haemodialysis treatment; 64.7 ± 15.3 Y/O; Uncompressed video format; 15 fps, 2448 × 2048 px	HR:0.0 ± 2.4; RR:0.2 ± 1.66
Malafaya¹⁷¹	Portugal	×	2020	3	—	15 min	Spectral and peak analysis; newborns in NICU; two cameras(Kinect V2 & Tesseract OCR engine); lateral	HR:RMSE = 6.93
Chen¹⁷²	China	×	2020	5	25–36	5*3 min	EVM algorithm; infants in NICU(13-25 days); Fluke Tix580; right ahead; 640 × 480 px, 30fps	HR:7.4^± 12.5
Imano¹⁷³	Denmark	×	2020	39	110	—	Estimate tidal volume (TV) and RR from chest depth image; Elderly people, 65-75 Y/O; still; RGBD, 30fps; right ahead	Tidal volume: :0.14 ± 0.03
Paul¹⁷⁴	Germany	×	2020	19	70	240 min	Neonates; RFGB, IRT and NIR sensor at side of bed; short-time Fourier transform; 1920 × 1200/640 × 480 px, 30/25 fps	HR: ME=3.0
Negishi¹⁷⁵	Japan	×	2020	28	100	—	Influenza patients(45 Y/O); RGB-thermal image sensors; right ahead; 640 × 480 px(visible); 320 × 240 px(thermal)	RMSE:HR:5.93 ± 5.85; RR&BT:1.73 ± 1.68
Villarroel⁶⁸	UK	✓	2019	30	—	426.6 h	Preterm infants in NICU; GA:30.7 weeks; ECG sensor; above; 1620 × 1236 px	HR:0.3 ± 4.39; 7D3RR:−1.1 ± 5.66
Chaichulee¹⁴⁵	UK	✓	2019	15	—	226.4 h	Preterm infants; above; 1620 × 1236 px; skin segmentation	ROI accuracy:94.5%
Rehouma¹⁴¹	Canada	×	2019	2	85 & 160	—	Children in PICU, 0-18 Y/O; still; two depth cameras; 45^∘; 512 × 424 px	RR:0.8 ± 0.41(patient1); 0.7 ± 0.56(patient2)
Slapnicar¹⁷⁶	Slovenia	×	2019	22	—	—	POS algorithm; adult patient; still; right ahead	HR:MAE = 7.92
Antognoli¹⁷⁷	Italy	×	2019	40	50	—	Neonates; GA:35 ± 2 weeks; still; a uniform light source; above; 1280 × 720 px	RMSE: 6.8(HR), 2.1(RR)
Rasche¹⁷⁸	Germany	×	2019	70	60–100	—	OPP; adult patients(70.3 ± 11.4 Y/O); fluorescent illumination and ambient light; right ahead; 420 × 720 px	—
Antognoli¹⁷⁹	Italy	×	2018	7	50	—	Neonates; GA:32 ± 4 weeks; still; above; 1280 × 720 px, 30fps	HR:12.2,RR:7.6(RMSE)
Trumpp¹⁸⁰	Germany	×	2018	41	50–100	—	Intraoperative patients(65.2 ± 12 Y/O); RGB and NIR cameras; surgical light; over head; 320 × 420 px	HR median: 95.6%(G),76.2%(NIR)
Chaichulee¹⁴⁶	UK	✓	2018	20	—	110 h	Using MatConvNet to detect the head, torso and diaper; NICU; GA:≤37 weeks; still; 1620 × 1236 px, 20 fps	HR:2.4 ± 4.1
Cobos¹⁸¹	Spain	×	2018	—	50	—	Neonates; GA:25-40 weeks; natural/artificial light and skylights; above; 1920 × 1080/1280 × 720 px	HR:-1.5 ± 4.21, RR:0.6 ± 4.97
Blanik¹⁸²	Germany	×	2016	10	50–100	Each 10 min	Neonates, 8-41 days; IR camera, infrared light emitting diode (LED) array (850 nm); above; 1920 × 1080 px, 25 fps	HR:−2 ± 9.69
Sikdar¹⁸³	India	×	2015	14	—	—	Comparing the result of using RGB, RG, GB and RB channel; Infants(3–15 months) under Hammersmith examinations;	HR: –
Villarroel¹⁴⁴	UK	×	2014	2	150	39.8 h	Neonates; GA:29 weeks; above; JAI AT-200CL digital 3CCD progressive scan, 1620 × 1236 px, 20 fps	HR:MAE = 2.83
Aarts¹⁸⁴	Netherlands	×	2013	2	100	—	Neonates; 3 days–4 weeks; still; Blue Sensor NEOX, 150 lux.; 300 × 300 px, 30/15fps	HR:0.3 ± 2.68
Scalise¹⁸⁵	Italy	×	2012	7	20	28 min	Neonates; GA:33 ± 2.5 weeks; large band light; right ahead; 640 × 480 px	HR:−0.9 ± 4.54

The databases are collected by the authors and are private.

BMI body mass index, GA Gestational ages, Y/O years old, RGBD RGB Depth camera, IR infrared, NIR Near infrared, CCTV closed circuit television, EVM Euler video magnification, IRT Infrared thermal imaging, fps Frames per second.