High-quality Image (NIR and RGB) Dataset Synchronized With Contact Vital Sings Recordings and Clinical Data of Stratified Healthy Population. Algorithms and AI Models to Obtain a Set of Vital Signs Image-based.

5.2.1. SCIENTIFIC JUSTIFICATION OF THE PROJECT Vital signs monitoring is essential in modern clinical patient care both in hospital and at home. Some examples are heart rate (HR), respiratory rate (RR), oxygen saturation (SpO2) or blood pressure (BP), which provide crucial information on the state of a human biological system (2). The recording of these signals is usually performed with traditional devices that require contact with the patient through elements such as surface electrodes, pulse oximeter, or piezoelectric transducers. These devices are unsuitable for long-term vital signs monitoring as they can cause discomfort, irritation and a cumulative risk of fungal and bacterial infections. Moreover, they can cause skin infections, wounds and harmful reactions on patients with fragile skin, such as premature babies, older people or burned patients (3). On the other hand, many of the materials used to monitor patients are only intended for a single use (such as electrodes and leads), and then must be disposed, which entails important costs (2).

The idea of recording of these signals without the need for direct skin contact is rising as an alternative able to eliminate potential medical complications and to solve usability issues related to the need for contact devices. New technologies aim to monitor multiple vital signs with a single device, at any time and in any place (4), which would allow the prevention of certain pathologies while improving patients' health. In addition, the costs associated with the acquisition of single-use sensors, as well as patients' hospitalisation, would be reduced.

Recently, there has been a notable development in the techniques for obtaining vital signs from non-contact technologies. These focus on the observation of physical and physiological variations of the human body due to the activity of the respiratory and cardiovascular systems. Notably, these variations include skin colour, temperature, impedance changes, head motion, arterial pulse motion, and thoracic and abdominal motion (2,5).

One of the most used technologies in the development of portable devices to monitor physiological signals is video camera imaging. This passive contactless method exploits two main principles (2). First, the colour of blood varies due to the exchange of gases through cardiorespiratory action, altering the optical properties of the skin. This can be described as photoplethysmography (PPG) (5). These variations in skin colour can be appreciated in sequences of images obtained with a video camera and be exploited to obtain a signal of remote PPG (rPPG), from which several vital signs can be measured. The second principle is based on the measurement of the ballistocardiogram (BCG), which depends on subtle cyclic head motions generated as a result of pressure changes from the influx of blood at each pulse (2,5).

Since video cameras do not need any kind of interaction with the patient, they represent a promising approach for contactless monitoring. Their main advantages are their robustness, reliability, safety, cost-effectiveness, and suitability for long distance and long-term monitoring.

Having a database that includes the main important factors to calculate vital signs from image-camera will allow both the development of specific and more accurate algorithms and the training of AI-based models to obtain vital signs using image-based technology.

In this context, it should be added that the Institute of Biomechanics of Valencia has worked on a large number of projects, both national and European, having extensive experience in the signal processing and analysis of physiological signals:

• SuaaVE (https://www.suaave.eu/)
• Senserror (https://www.ibv.org/proyecto/senserror-modelizacin-y-evaluacin-del-error-humano-mediante-procesos-cognitivos/)
• SOLFIS (https://www.ibv.org/proyecto/solfis-soluciones-aplicadas-basadas-en-medidas-fisiologicas/)
• imc2 (https://www.ibv.org/proyecto/emc2incorporacin-de-los-modelos-emocionales-y-cognitivos-en-el-sector-del-transp/) 5.2.2. PROJECT OBJECTIVES LONG-TERM GOAL The objective of this study is to collect an image dataset of healthy and pathological population with a characterization of the potential factors that may influence the methodologies used to obtain vital signs. To this end, an observational study of qualitative and quantitative data collection through cameras, contact recording technologies (gold standards) and questionnaires has been proposed.

Having a database that includes the main important factors to calculate vital signs from image-camera will allow both the development of specific and more accurate algorithms and the training of AI-based models to obtain vital signs using image-based technology.

Based on these data, the specific objectives of this study are as follows:

• To determine the differences according to the factors analysed.
• Develop/fine-tune algorithms/AI models to obtain quality rPPG/rBCG signals.
• Develop/fine-tune algorithms/AI models to obtain the main vital signs as HR, RR, SpO2 and BP from image-based-data.

5.2.3. METHODOLOGY TYPE OF STUDY Observational study to collect an image dataset of healthy population with a characterization of the potential factors, both intrinsic (age, BMI, sex, Fitzpatrick phototype, risk factors) and extrinsic to the subject (distance and illuminance) that may influence the methodologies used to obtain vital signs.

STUDY SAMPLE Sample size: At least 60 subjects that cover all the specific ranges defined for each of the intrinsic subject factors: age, BMI, sex and Fitzpatrick phototype. The number of subjects defined for each combination was based on the percentage it represents in the Spanish population.

The sample it has been strata based on gender, Fitspatrick Phototype, BMI and age with the following ranges:

• Sex: female (F) and male (M).
• Fitzpatrick Phptotype: I-II (F1), III-IV (F2) and V-VI (F3).
• BMI: <21 (B1), 21<= BMI<=29 (B2) and >29 (B3).
• Age: <50 (A1) and >=50 (A2).

Based on the defined ranges it has been obtained 36 subgroups with the following characteristics:

• G1: F/F1/B1/A1
• G2: F/F1/B1/A2
• G3: F/F1/B2/A1
• G4: F/F1/B2/A2
• G5: F/F1/B3/A1
• G6: F/F1/B3/A2
• G7: F/F2/B1/A1
• G8: F/F2/B1/A2
• G9: F/F2/B2/A1
• G10: F/F2/B2/A2
• G11: F/F2/B3/A1
• G12: F/F2/B3/A2
• G13: F/F3/B1/A1
• G14: F/F3/B1/A2
• G15: F/F3/B2/A1
• G16: F/F3/B2/A2
• G17: F/F3/B3/A1
• G18: F/F3/B3/A2
• G19: M/F1/B1/A1
• G20: M/F1/B1/A2
• G21: M/F1/B2/A1
• G22: M/F1/B2/A2
• G23: M/F1/B3/A1
• G24: M/F1/B3/A2
• G25: M/F2/B1/A1
• G26: M/F2/B1/A2
• G27: M/F2/B2/A1
• G28: M/F2/B2/A2
• G29: M/F2/B3/A1
• G30: M/F2/B3/A2
• G31: M/F3/B1/A1
• G32: M/F3/B1/A2
• G33: M/F3/B2/A1
• G34: M/F3/B2/A2
• G35: M/F3/B3/A1
• G36: M/F3/B3/A2 MEASUREMENT TOOLS

The measurement equipment used to register has been:

• RR register: biosignalsplux tool-kit with Respiration (PZT) sensor data. To acquire the raw data, OpenSignals revolution software have been used.
• HR register: biosignalsplux tool-kit with electrocardiography (ECG) sensor data. To acquire the raw data, OpenSignals revolution software have been used.
• SpO2 register: biosignalsplux tool-kit with blood volume pulse (BVP) finger clip sensor data. To acquire the raw data, OpenSignals revolution software have been used.
• BP register: Withings BPM Connect WPM05 digital sphygmomanometer.
• Sync register: biosignalsplux synchronization (SYNC) accessory data. To acquire the raw data, OpenSignals revolution software have been used.
• Images register: FLIR GS3-U3-41C6C and FLIR GS3-U3-41C6NIR cameras. All the biosignalsplux sensors have been synchronized with cameras thought the SYNC accessory.

On the other hand, a questionnaire designed by the IBV on clinical factors was completed. The questionnaire is described in detail in section 9.

All the recordings have been carried out by IBV qualified staff to perform the tests and have been performed in the IBV facilities, specifically in the Living Lab laboratory with climate and lighting control to adapt the test conditions.

MEASUREMENT PROTOCOL Potential participants will initially be contacted by the corresponding key entity or agent (collaborators specified in the corresponding section 3. Sponsor/Collaborators), including the IBV itself through the databases of previous trials. The IBV will contact those users interested in participating, who meet the inclusion criteria, and who give their consent to be contacted.

Two sessions were carried out at different times of the day for each of the subjects. In the first one (9 a.m. to 12 a.m.), 4 measurements were taken modifying the distance between the subject and the equipment (1 m and 2 m) and the entrance of natural light (presence or absence of natural light). In the second session (4 p.m. to 5:30 p.m.), only one measurement was taken at 1 m with natural light, aiming to record pressure at its maximum peak of the day. The subject remained seated and instrumented throughout the sessions. The forehead, face and lower neck were exposed to the cameras. Prior to each measurement, the ambient light condition was recorded using a luxmeter. Subjects were required to remain still and quiet, breathing normally. They also had to raise their hands to the level of their heads with the right palm facing forward. At the beginning and at the end of each recording, systolic and diastolic blood pressure were measured with the Withings BPM Connect WPM05 digital sphygmomanometer.

Each recording lasted 80s: during the first 20s the scene was illuminated with a pulsed light and in the following 60s a constant light was used.

VARIABLES ANALYZED The variables analysed correspond to the data collected in each sensor data and questionnaires. All the variables are described in section 9: Outcome Measures.

STATISTICAL ANALYSIS

A statistical analysis will be carried out that will serve to respond to the objectives of the study. The statistical analysis will treat the data provided by the parameters obtained with the reference measurement techique compared with the parameters obtained with the algorithms/AI models developed. To do this, the following statistics will be used:

• Pearson's Correaltion coefficient: linear correlation between two sets of data. It is the ratio between the covariance of two variables and the product of their standard deviations.
• Cronbach's alpha: reliability coefficient and measure of internal consistency os tests and measures.
• Accuracy: value of how close given set of measurements are to their true value. Depending on the type of algorithms/AI models implemented, the statistics may be different.

DATA MANAGEMENT AND PROCESSING A user code will be assigned to each participant, so that the questionnaires can be linked while maintaining the anonymity of the participants for the IBV. This code will be the one provided to identify yourself in the online questionnaire, or the one that will be added to the questionnaire that is completed by telephone, if applicable.

All information will be anonymized for processing and analysis, and may be used under the terms and conditions dictated by the current legal framework.

Once the project is finished, the data obtained will be stored, always in an anonymized form and under the legal custody of the IBV. The subsequent use and retention period of the data are governed as contemplated in the informed consent documents, with respect to current legislation.

The acceptance of the written document of Informed Consent is made through checkboxes at the beginning of the online survey itself. In the case of patients who are surveyed by telephone, the document corresponding to Written Informed Consent will be sent before the interview, and subsequently they will be asked for consent verbally, after ensuring that they have reviewed and understood the conditions of their participation in the study.