Learning to Predict Blood Pressure with Deep Bidirectional LSTM Network. PAPER
Around 3 in 10 deaths globally are caused by cardiovascular diseases (CVD) - diseases of the
heart and blood vessels that can cause heart attacks and stroke. 2 As the leading risk factor
of CVD (Lim et al., 2013), high blood pressure (BP) has been commonly used as the critical
criteria for diagnosing and preventing CVD. High BP, which is also known as hypertension,
normally develops without obvious symptoms at early stage, making it a “silent killer”.
Therefore, accurate and continuous BP monitoring during people’s daily live is extremely
imperative for CVD prevention and diagnosis. In addition, blood pressure variability (BPV)
reflects how a cardiovascular system regularize itself and response to external stimulus, and
is another critical cardiovascular indicator that can only be obtained through continuous
and long-duration BP monitoring.
Current BP measurement devices, e.g., Omron products, are cuff-based and therefore
bulky, discomfort to wear, and only suitable for snapshot measurements. These disadvantages
restrict the use of the cuff-based devices for continuous and frequent BP measurement,
which are essential for nighttime monitoring and precise diagnosis of different CVD symptoms.
Recent advancements in sensing technologies provide a wearable sensor network
solution that can achieve cuffless and continuous BP monitoring (Chan et al., 2007; Zheng
et al., 2014). These new emerged sensing technologies can detect several human physiological
signs through contacting corresponding sensors to the human body. For example,
electrocardiography (ECG) sensor can detect tiny skin impedance variations that arise from
the hearts electrophysiologic pattern during each heart beat; photoplethysmogram (PPG)
sensor can probe blood volume variation inside arteries, and etc. While all these physiological
sensing signals contain enormous information of the functionality and health status of
our cardiovascular system, the data is difficult to mine effectively due to noisy observation,
missing value and varying length. Extensive research efforts have been made to develop
effective models to predict or estimate BP from the sensor output; examples include the
well established physiological modeling method - Pulse transit time model, and the recently
proposed machine learning approach - regression model such as support vector machine,
decision tree and etc
10 facts on the state of global health WHO
from:
A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010
