Developing knowledge on the electronics for the wearable prototypes

Work carried out for the development of the project, has been focused on acquisitions of physiological optimal data for what concerns time and frequency domains to extract informative features to be used with machine learning techniques to obtain Blood Pressure (BP). The work on retrieval of cuff-less BP is divided in two consecutive parts. 

The first one was divided into two paths in parallel, one for training and testing algorithms for estimation of BP using MIMIC III database signals ,and the other by using Shimer modules. The first phase has afforded the team to evaluate the capability of algorithm to predict SBP and DBP by using PTT (pulse transit time). Multiple information extracted from the PPG and ECG signals have been explored to measure the pressure through this signal as accurately as possible. Phase Training and testing (MIMIC III database signals): best results through the use of random tree algorithm, observing the scatter plots. 

Estimation of the implemented models is concentrated along the central straight line, indicating an almost linear correlation between the target and the estimated values. However, these techniques need to be improved with accuracy to give machine learning approaches precision to allow practical usage in clinical or sportif application.  

During this phase also a preliminary experiment was conducted to compare the Antenna and interfaces for IBC. Through this experiment, it is found that in the free space Fat IBC antenna received more signals than the Bluetooth antenna. This operation will bring the definition of an experiment on the Fat channel using Phantoms.  

Pilot tests 

As previously mentioned, two different pilot tests of the 3rd generation smart patches have been conducted: one with a combination of running and walking followed by strength training and stretching, and another with a mountain bike training. Two different positions of sensors were tested representing, since the beginning, a revolution for sports measurements. The first patch was attached on the chest, on the side of the heart.  

While the second patch was stationed on the internal side of the forearm. 

The initial pilot tests revealed the following: the software Loligolog for the iOS device that was part of the package containing smart patch was intuitive and not difficult to operate. Realtime data were being able to be monitored on the device’s screen and data were being able to quickly check on diagrams. Other progresses on those tests are expected by the end of the project.