Non-invasive continuous blood pressure measurement using machine learning

Blood Pressure (BP) is a vital biosignal of the human body and a biomarker for many diseases. High BP significantly increases the risk for strokes, heart attacks, ischaemic heart disease, dementia and chronic kidney disease [1]. With over 25% of adults suffering from hypertension (i.e., Systolic Blood Pressure (SBP) ě 140 mmHg or Diastolic Blood Pressure (DBP) ě 90 mmHg), it is a major contributor to cardiovascular diseases which are the leading cause of premature death worldwide [1]. Hypertension is also called "the silent killer" since it comes without symptoms and slowly damages organs and the cardiovascular system. According to the World Health Organisation, 46% of adults with elevated BP are unaware of their condition and only one in five affected adults have their hypertension under control [2]. Moreover, in the clinical setting, changes in BP can indicate a hypovolemic shock, hypoxemia, infections or other conditions that require immediate treatment. Therefore BP is closely monitored especially during surgeries and in the Intensive Care Unit (ICU). During these high risk situations, an invasive measurement is employed which is considered to be the gold-standard. For the invasive measurement, a catheter with a pressure sensor is introduced into the patient’s artery such that BP can be monitored continuously and at various locations along the arterial tree. Despite the high accuracy and continuity of these measurement systems, they are limited to critical clinical applications since they entail a high risk for thrombosis and infections. Hence, for common BP monitoring and ambulatory BP measurement, auscultatory and oscillometric cuff-based devices are employed instead, without complications but exhibit inferior accuracy. Moreover, cuff-based systems are discontinuous with measurement durations up to a minute and cause discomfort in many patients. Other less established methods are vascular unloading, tactile sensors and volume clamps, each with their own set of disadvantages.