Assessment of Blood Pressure Using Only a Smartphone and Machine Learning Techniques: A Systematic Review

Abstract

Regular monitoring of blood pressure (BP) allows for early detection of hypertension and symptoms related to cardiovascular disease. Measuring BP with a cuff requires equipment that is not always readily available and it may be impractical for some patients. Smartphones are an integral part of the lives of most people; thus, detecting and monitoring hypertension with a smartphone is likely to increase the ability to monitor BP due to the convenience of use for many patients. Smartphones lend themselves to assessing cardiovascular health because their built-in sensors and cameras provide a means of detecting arterial pulsations. To this end, several image processing and machine learning (ML) techniques for predicting BP using a smartphone have been developed. Several ML models that utilize smartphones are discussed in this literature review. Of the 53 papers identified, seven publications were evaluated. The performance of the ML models was assessed based on their accuracy for classification, the mean error measure, and the standard deviation of error for regression. It was found that artificial neural networks and support vector machines were often used. Because a variety of influencing factors determines the performance of an ML model, no clear preference could be determined. The number of input features ranged from five to 233, with the most commonly used being demographic data and the features extracted from photoplethysmogram signals. Each study had a different number of participants, ranging from 17 to 5,992. Comparisons of the cuff-based measures were mostly used to validate the results. Some of these ML models are already used to detect hypertension and BP but, to satisfy possible regulatory demands, improved reliability is needed under a wider range of conditions, including controlled and uncontrolled environments. A discussion of the advantages of various ML techniques and the selected features is offered at the end of this systematic review.

Optimal signal quality index for remote photoplethysmogram sensing

“…Recent research in rPPG has explored the evaluation of red, green, and blue channels for heart rate detection 3 , the development of less complex methods for improved heart rate measurement via rPPG 4 , the evaluation of biases in rPPG methods 5 , and investigations into the effectiveness of various rPPG methods in different settings 6,7 . Additionally, there have been studies on the use of machine learning for blood pressure detection using rPPG 8,9 . However, the accuracy of rPPG is often compromised by artifacts, primarily due to motion and external light interference, which significantly impact signal quality 10 .…”

Recent Advances in Materials, Devices and Algorithms Toward Wearable Continuous Blood Pressure Monitoring

“…Many researchers have developed a series of ML-assisted systems with various models for BP estimation 231 in which feature selection is a crucial part that greatly influences the quality of these assessment models (Figure 12D). The commonly adopted features include pulse waveform features, e.g.…”

Effectiveness of Remote PPG Construction Methods: A Preliminary Analysis

“…In the current body of literature, the rPPG signal is frequently compared solely to the extracted health-related information, such as heart rate or blood pressure, rather than to the ground truth PPG signal [ 5 , 6 ]. The error metrics used in this case are often the mean absolute error (MAE) or Pearson’s correlation coefficient ( r ) between the estimated and ground truth health-related information.…”