Skip to main content
European Commission logo print header

Telehealth-ready AI-powered multi-parametric system for surveillance of COVID-19 and cardio-pulmonary chronic patients

Periodic Reporting for period 1 - PyXy.AI (Telehealth-ready AI-powered multi-parametric system for surveillance of COVID-19 and cardio-pulmonary chronic patients)

Berichtszeitraum: 2020-12-01 bis 2021-11-30

Patients with underlying chronic conditions are among the most vulnerable in the COVID-19 pandemic and in new potential pandemics, having a higher risk of developing life-threatening pneumonia when infected with COVID-19 and acute complications of their chronic condition due to other viral and bacterial infections. Evidence has shown that patients who have reported no pre-existing conditions had a case fatality rate of 0.9% while those with a pre-existing illness have had a 5-10 times higher chance of a negative outcome. Respiratory and cardiovascular diseases are amongst the chronic conditions that represent higher vulnerability to negative outcomes.
This is particularly relevant for nursing homes and home care (in some cases with large travel distances), responsible for a large share of such risk-group patients. They are the frailest with complex disease history, often as many as 5-6 diagnoses, and thus need closer medical surveillance. Nursing homes in the OECD countries take care of up to 7 million residents, nearly 70% have dementia-related disorders.
The vulnerability and devastating effects of the COVID-19 pandemic in nursing homes have already been documented in the scientific and medical literature. During a pandemic, an outbreak at a nursing home has shown to be difficult to contain, due to the need for close contact with personnel and lack of facilities for cohort sick individuals. There is an increased need for health personnel at all levels of care but staff shortage is expected due to infected personnel and quarantines, who will need to be replaced by less skilled personnel, leading to limited infection prevention and control practices. Early evidence shows that this cohort has the largest percentage of deaths from COVID-19 – consistently reported at over 40% across several European countries and in the US. This evidence additionally reports excess deaths as comparing to the same weeks the previous year, suggesting that COVID-19 has had an indirect negative impact on overall higher number of deaths, potentially as a result of lockdown measures.
The primary objective of the project is to mature, optimize and demonstrate an AI-powered cloud-based system for patient’s lung health assessment and surveillance, which integrates an intelligent sensor device – PyXy – and machine learning to increase the efficiency and the quality of the care delivery to risk group patients, especially under a pandemic outbreak. The methods should have clinically valuable precision in early detection of individual’s alarming health change by monitoring cardio-respiratory parameters using combined clinical indicators. The system should fit with routine workflows at the health institution as well as fit with established infrastructure of e-health / IT systems.
Product maturation – the PyXy design was optimized, adding further functionalities to it and improving its ergonomics. Industrial, mechanical and electronic designs were finalized, and the development of the device’s SW has started. The new device incorporates sound capturing (ranging from infrasound to audible sound) with one-lead ECG and SpO2 measurement capabilities.
Cloud App, mobile and Web applications – A cloud app that includes dashboards for the care providers (also hospitals, clinics) was built and demonstrated to clinical partners within the consortium. Operating API for cloud app has been developed as well. This enables to integrate with existing IT systems used by the healthcare sector, be it electronic patient journals or e-health platform providers. When such external systems for test-integration are identified further, new parameters for exchange of data with 3rd party solutions might be introduced.
Clinical affairs – ethics approvals for data collection from COVID-19 and chronic heart and lungs patients were obtained in Germany, Spain, Israel and Norway. Data collection has begun in Germany, Spain and Israel, and is expected to commence in Norway in February. So far, heart and lung sounds were collected from over 200 patients overall. Delays in ethics approvals caused a slow start to the data collection process, but it is expected to quickly pick up in the coming months.
AI exacerbation detection algorithm – One of the main development targets for AI is to develop algorithms for indicating development in the lungs state for both COVID context (i.e. positive/negative progression in the condition's development to be able to identify onset of alarming symptoms) as well as patient-centered, resource-efficient chronic care management of lung and heart patients for indicating positive/negative development in their condition. So far, the focus has been on modifications of current Medsensio’s algorithms for data collected in this project using Sanolla’s VoqX. In addition, efforts were put in development of annotation guidelines for organizing annotation of data from WP1 which will be necessary for further ML training.
Exploitation and dissemination – a website and a LinkedIn page for the project were created. These platforms are continuously updated with project related news to bring up the awareness to the project.
By the end of the project we expect to introduce to the market a technological leap forward in remote monitoring of COVID-19 and chronic heart and lungs patients. Research has shown that in some diseases, lung sounds are in constant development. The AI algorithms to be developed in this project will analyze different types of abnormal heart and lung sounds, for detecting differences in the sounds, even of the same type, over time. The innovation factor here is that auscultation results, when analysed with objective AI algorithms, have never been saved under patient’s history in a format that allows to review them in detail and use AI to compare developments in the patient’s breathing over time across different time-entries.
Screens of mobile application for end user
Screens of web dashboard for physician use
The device in its charging station
The final design of the device