The vital signs are critical in assessing the severity and prognosis of infections, such as Covid-19. The devices used today for measuring the vital signs have to be in physical contact with the patients. There is an apparent risk of transferring infections from one patient to the next (or to healthcare professionals). This project aims to evaluate a new camera-based system for contactless measurement of vital signs as well as an artificial intelligence (AI) predicting hospitalization or death within 30 days. This particular study will evaluate the new system's ability without interfering with standard care of the patient.
Background and aim: The vital signs are critical in assessing the severity and prognosis of infections, i.e., Covid-19, influenza, sepsis and pneumonia. Quick and accurate triage is critical when facing a pandemic with an overwhelming number of cases (confirmed and suspected). This study aims a) to evaluate a new method for rapid camera-based non-contact measurement of five vital signs; body temperature, heart rate, blood oxygen saturation, respiratory rate, and blood pressure, and b) if an AI can predict hospitalization or death within 30 days. Methods: A method-comparison study design is used comparing each vital sign measured with the new method to the corresponding standard reference method. Furthermore, a cohort design is used to follow up any hospitalization or death within 30 days. The investigated new system consists of a high-speed digital video camera, a digital radiometric infrared camera, LED lights and a computer for data recording. This system faces the subject at a distance of approximately one meter, capturing a 30 second recording of the subject's face. First, all vital signs will be measured using one set of reference devices. Secondly the investigated device will record a 30 second video of the patient's face. Thirdly, and last, all vital signs will be measured using the same set of reference devices. A copy of the vital sign readings (using the standard reference methods) will be handed over by an investigator to the clinical professionals responsible for the subsequent medical care for each subject. Afterwards, the collected 30-second recordings will be run through specific software algorithms to extract the vital signs. The results from the new camera-based contactless measurement of vital signs and the outcome of the AIs prediction of risk for hospitalization or death will not be presented in the care situation of the patient. Expected Findings: It is expected that the proposed study will show that the new method can estimate body temperature, heart rate, respiratory rate, blood oxygen level, and blood pressure with an acceptable agreement compared with the reference method and also estimate hospitalization or death within 30 days. Implications of the expected findings: Being able to measure vital signs quicker than before by using a new contactless method would greatly facilitate triage of large number of patients. Also being able to predict hospitalization or increased risk for death would further improve the triage of patients.
The investigated new system consists of a high-speed digital video camera, a digital radiometric infrared camera, LED lights and a computer for data recording. This system faces the subject at a distance of approximately one meter, capturing a 30 second recording of the subject's face.
Östra Sjukhuset
Gothenburg, Sweden
Agreement between the new camera based method and reference standard to estimate body temperature
Body temperature will be measured with the new camera based method as well as with a conventional ear thermometer. Both measurements will estimate the body temperature in degrees Celsius. The agreement between body temperature estimated with the new method and the reference method will be made using the statistical methods Bland-Altman plots and limits of agreement as the outcome.
Time frame: Two minutes between measurements
Agreement between the new camera based method and reference standard to estimate heart rate
Heart rate will be measured with the new camera based method as well as with a conventional apparatus for measuring pulse rate. Both measurements will estimate the heart rate in beats per minute. The agreement between body temperature estimated with the new method and the reference method will be made using the statistical methods Bland-Altman plots and limits of agreement as the outcome.
Time frame: Two minutes between measurements
Agreement between the new camera based method and reference standard to estimate blood oxygen saturation
Blood oxygen saturation will be measured with the new camera based method as well as with a conventional apparatus for measuring blood oxygen saturation. Both measurements will estimate the blood oxygen saturation in percent (ranging from 0-100%). The agreement between blood oxygen saturation estimated with the new method and the reference method will be made using the statistical methods Bland-Altman plots and limits of agreement as the outcome.
Time frame: Two minutes between measurements
Agreement between the new camera based method and reference standard to estimate systolic blood pressure
Systolic blood pressure will be measured with the new camera based method as well as with a conventional apparatus for measuring systolic blood pressure. Both measurements will estimate the systolic blood pressure in mm Hg. The agreement between systolic blood pressure estimated with the new method and the reference method will be made using the statistical methods Bland-Altman plots and limits of agreement as the outcome.
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.
Study Type
OBSERVATIONAL
Enrollment
214
Time frame: Two minutes between measurements
Agreement between the new camera based method and reference standard to estimate diastolic blood pressure
Diastolic blood pressure will be measured with the new camera based method as well as with a conventional apparatus for measuring diastolic blood pressure. Both measurements will estimate the diastolic blood pressure in mm Hg. The agreement between diastolic blood pressure estimated with the new method and the reference method will be made using the statistical methods Bland-Altman plots and limits of agreement as the outcome.
Time frame: Two minutes between measurements
Agreement between the new camera based method and reference standard to estimate respiratory rate
Respiratory rate will be measured with the new camera based method as well as manually using a stethoscope. Both measurements will estimate the respiratory rate in breath per minute. The agreement between respiratory rate estimated with the new method and the reference method will be made using the statistical methods Bland-Altman plots and limits of agreement as the outcome.
Time frame: Two minutes between measurements
Prediction of hospital admission using vital signs estimated using reference standard methods
An artificial intelligence (AI) algorithm will use 75% of patient observations of vital signs for training and the remaining 25% will be used to test the AIs predictive capabilities to predict hospital admission within 30 days. For each patient the AI will produce a probability (0-100%) for hospital admission within 30 days. These probabilities will undergo a receiver operating (ROC) analysis where area under curve (AUC) with 95% confidence interval will be the reported outcome measure.
Time frame: Hospital admission for any cause up until 30 days after inclusion
Prediction of death using vital signs estimated using reference standard methods
An artificial intelligence (AI) algorithm will use 75% of patient observations of vital signs for training and the remaining 25% will be used to test the AIs predictive capabilities to predict death within 30 days. For each patient the AI will produce a probability (0-100%) for death within 30 days. These probabilities will undergo a receiver operating (ROC) analysis where area under curve (AUC) with 95% confidence interval will be the reported outcome measure.
Time frame: Death for any cause up until 30 days after inclusion
Prediction of hospital admission using vital signs estimated using the new camera based method
An artificial intelligence (AI) algorithm will use 75% of patient observations of vital signs for training and the remaining 25% will be used to test the AIs predictive capabilities to predict hospital admission within 30 days. For each patient the AI will produce a probability (0-100%) for hospital admission within 30 days. These probabilities will undergo a receiver operating (ROC) analysis where area under curve (AUC) with 95% confidence interval will be the reported outcome measure.
Time frame: Hospital admission for any cause up until 30 days after inclusion
Prediction of death using vital signs estimated using the new camera based method
An artificial intelligence (AI) algorithm will use 75% of patient observations of vital signs for training and the remaining 25% will be used to test the AIs predictive capabilities to predict death within 30 days. For each patient the AI will produce a probability (0-100%) for hospitalization within 30 days. These probabilities will undergo a receiver operating (ROC) analysis where area under curve (AUC) with 95% confidence interval will be the reported outcome measure.
Time frame: Death for any cause up until 30 days after inclusion
Prediction of hospital admission using raw camera data
An artificial intelligence (AI) algorithm will use 75% of patient observations of raw camera data for training and raw data from the remaining 25% of patients will be used to test the AIs predictive capabilities to predict hospital admission within 30 days. For each patient the AI will produce a probability (0-100%) for hospital admission within 30 days. These probabilities will undergo a receiver operating (ROC) analysis where area under curve (AUC) with 95% confidence interval will be the reported outcome measure.
Time frame: Hospital admission for any cause up until 30 days after inclusion
Prediction of death using raw camera data
An artificial intelligence (AI) algorithm will use 75% of patient observations of raw camera data for training and raw data from the remaining 25% of patients will be used to test the AIs predictive capabilities to predict death within 30 days. For each patient the AI will produce a probability (0-100%) for death within 30 days. These probabilities will undergo a receiver operating (ROC) analysis where area under curve (AUC) with 95% confidence interval will be the reported outcome measure.
Time frame: Death for any cause up until 30 days after inclusion