Results of remote measurement of certain physiological parameters based on artificial intelligence

Abstract

Background: According to the WHO, 1.28 billion people worldwide suffer from hypertension, with the majority living in developing countries. In Mongolia, the prevalence of hypertension remains high, and the STEPS-2019 survey found the disease is common even among younger age groups. Manual/Automated blood pressure measurement methods require specific conditions and carry risks of measurement error. In recent years, the availability of AI-based mobile applications to measure blood pressure and vital signs has increased; however, there is a lack of research in Mongolia evaluating their accuracy in accordance with the ANSI/AAMI/ISO 81060-2:2022 standard, highlighting the necessity for this study. Aim: To compare the results of blood pressure, heart rate, and peripheral oxygen saturation SpO₂ determined by an AIbased application with the results of traditional clinical device measurements. Materials and Methods: We included 120 participants who consented to the study. Measurements were taken 3 times each using both the clinical device and the AI-based application, collecting a total of 360 measurement values. Data were analyzed using Pearson’s correlation coefÏcient, Paired t-Test, Bland-Altman plots, and Intraclass Correlation CoefÏcient (ICC). Results: The mean difference between the clinical device and the AI-based smartphone measurements was -2.05 mmHg (p<0.001) for systolic pressure, -2.36 mmHg (p<0.001) for diastolic pressure, -0.30 bpm (p=0.088) for heart rate, and 0.32% (p<0.05) for SpO₂ Pearson’s correlation coefÏcients were r=0.772 (p<0.001) for systolic pressure, r=0.892 (p<0.001) for diastolic pressure, r=0.978 (p<0.001) for heart rate, and r=0.391 (p<0.001) for SpO₂ Conclusion: While AI-based application measurements are capable of determining blood pressure, heart rate, and peripheral oxygen saturation, they are not accurate enough for use in clinical practice.