Some of the scientific contributions of the project, representing progress with respect to the state of the art, have been:
1-A novel algorithm for automatic Identification of fundamental and additional sounds from cardiac sounds recordings without an auxiliary signal reference; together with a methods to differentiate cardiac events corresponding to normal cycles from those which are due to abnormal activity of the heart. The proposed algorithm achieved a sensitivity of 91.79% and 89.23% for the identification of normal fundamental, S1 and S2 sounds, and a true positive rate of 81.48% for abnormal additional sounds, outperforming other algorithms in literature.
2-A novel algorithm to extract heart rate from the signal obtained with a customized sensor located on the suprasternal notch, which on real-life testing (i.e. not controlled conditions) proved to achieve an accuracy of over 94% in identification of S1 and S2 sounds, and heart rate. The performance was superior to any other algorithm in literature.
3-A novel algorithm to automatically determine heart rate variability (HRV) by processing the acoustic data, recorded by placing a small, wearable sensor on the suprasternal notch (at the neck) of an adult subject, primarily intended to record breathing sounds. The instantaneous heart rate (IHR) comparisons yielded an accuracy of 95.78% and 92.35% for S1 and S2 sounds respectively. The experimental results showed, for the first time, that the proposed algorithm can provide an accurate HRV analysis for the cardiac signals recorded at the neck, using a very small sensor.
4-A method and algorithm to extract heart rate from neck PPG signals. Mean absolute error (MAE), standard deviation error (SDAE) and root-mean-square error (RMSE) resulted in 1.22 1.54 and 1.98 beats per minute (BPM), respectively. This was the first time HR automatically extracted from a sensor of this kind was presented in literature.
5-A experimental method to characterize artifacts from PPG signals sensed at the neck. This was the first time this type of work was reported in literature.
6-A novel method to extract the jugular venous pressure (JVP) from the neck using a PPG reflection based approach. Until now JVP could only be obtained using invasive approaches. We have now proven for the first time ever that it is possible to extract JVP with a small, easy to use wearable device approach. This could potentially open up new avenues for easier diagnosis of certain cardiovascular diseases.
7- A novel algorithms which demonstrates for the first time the ability to identify seizures using acoustic internal body signals acquired on the neck. Tested on 667 hours of acoustic data acquired from 15 patients with at least one seizure, the algorithm achieved a detection sensitivity of 88.1% (95% CI: 79%-97%) from a total of 36 seizures, out of which 24 had no motor manifestations.
8- A novel algorithm that is able to automatically discriminate from EEG signals whether a subject has got epilepsy or not; achieving 100% sensitivity and 98.7% specificity in classifying 267 recordings from 105 subjects.
9-Proof-of-concept design of a wearable able to extract biomarkers from PPG signals on the neck.
10- Methods of automatic sleep position detection using one wearable channel of accelerometry data sensed on the neck, with at least 98% average accuracy in all three models; memory space down to 1.765 KB and prediction time around 0.8ms.
11-Two novel classification algorithms for artifacts and apnea detection from PPG.
12-A methodology and system architecture to supply power over mid-distance, in the context of bio-sensors. Both the whole idea of the system as well as architecture are the first time presented in literature.
The research from now until the end of the project will continue focusing on extracting as much multi-modal physiological information as possible from one single sensing location in the body, and creating integrated circuits topologies that would allow the creation of custom unobtrusive wearables for single or multi-modal physiological signal acquisition, requiring very little or zero input from patients.
