Long Term Short Memory approach for Hypertension (LOTSOM-H)

It is well recognized that ambulatory blood pressure (BP) monitoring by means of wearable sensors has the potential to enable new levels of health-related vigilance and medical care in a number of novel settings, including, for example, controlling chronic hypertension and monitoring in-patients during convalescence.

However, a significant challenge to realizing true non-invasive blood pressure (NIBP) measurement remains the problem of accounting for the unknown tension in the underlying arterial wall: If one simply measures pressure external to an artery (for instance, on the overlying skin), one is measuring the balance of intra-arterial pressure and the rapidly varying arterial wall tension.

Ideal NIBP methods solve the problem of estimating intra-arterial wall pressures independently of the arterial wall tension. Yet, there is no optimal solution to truly wearable NIBP measurement. The ideal wearable device would be lightweight, easy-to-apply, non-invasive, small, unobtrusive, and as close to imperceptible as a regular wrist-watch.

The fundamental assumption in Machine Learning is that analytical solutions can be built by studying past data models. Machine Learning supports that kind of data analysis that learns from previous data models, trends, patterns, and builds automated, algorithmic systems based on that study.

As Machine Learning relies solely on pre-built algorithms for making data-driven analysis and predictions, it claims to replace data analytics and prediction tasks carried out by humans. In Machine Learning, the algorithms have the capability to study and learn from past data, and then simulate the human decision-making process by using predictive analysis and decision trees.

Long short-term memory (LSTM) is a recurrent neural network (RNN) Machine Learning Algorithm architecture that remembers values over arbitrary intervals. Stored values are not modified as learning proceeds. RNNs allow forward and backward connections between neurons.

An LSTM is well-suited to classify, process and predict time series given time lags of unknown size and duration between important events. Relative insensitivity to gap length gives an advantage to LSTM over alternative RNN, hidden Markov models and other sequence learning methods in numerous applications.

LOTSOM-H looks to use a uniquely configured Machine Learning Algorithm to identify trends between optical sensor samples and thus develop a map of arterial performance which can thus allow a user to calculate a value for trending Blood Pressure. It is hoped that these works will enable the resolution of a continuous Blood Pressure as a metric that can be acquired by consumer health wearable devices.