AI4pep

Home detection of contagious and transmissible respiratory infections using an artificial intelligence-based cough monitor.

During the COVID-19 pandemic, the global health system faced an unprecedented challenge: an avalanche of suspected cases of respiratory diseases, which overwhelmed hospitals and medical services. The overlap of symptoms between different respiratory infections further complicated the diagnosis, making it difficult for health professionals to quickly classify and treat patients. In this context, there is an urgent need for innovative and accessible tools that optimize disease triage.

ABOUT THE PROJECT

Our proposal introduces an artificial intelligence (AI)-based solution for cough monitoring, a technology that has the potential to revolutionize the detection of infectious respiratory diseases such as tuberculosis, COVID-19, influenza A and/or B, and RSV. The key question of our research is whether these AI tools, which capture and analyze cough patterns in real time, can improve the accuracy and speed of case identification. We strongly believe that AI tools will not only accelerate case detection but will also contribute to earlier and more accurate initiation of treatment, decreasing the spread of disease. AI's ability to analyze large volumes of data in fractions of a second may be the key to transforming how we manage future epidemics.

Situation in Peru

Tuberculosis (TB) is an infectious respiratory disease caused by the bacillus Mycobacterium tuberculosis (Mtb). In 2022, approximately 16,000 cases were estimated nationwide.
Coronavirus disease (COVID-19) is an infection caused by SARS-CoV-2, which is potentially fatal but preventable through vaccination. Cases of COVID-19 are still occurring in Peru. In 2023, around 10,000 confirmed cases were detected in the Lima Centro Health Network alone.
Influenza, caused by influenza A and/or B viruses, as well as respiratory syncytial virus (RSV), increases its number of cases in winter and is reported in Peru as part of acute respiratory infections (ARI).

Each of the respiratory infections mentioned above has similar characteristics, such as cough as a symptom and the fact that they are common circulating infections in Peru, which is why the transmission of these diseases is of great importance in order to control them in the population.

Through new technologies based on artificial intelligence (AI), coughs can be recorded, their frequency and origin analyzed, and algorithms created and trained to help identify cough patterns and enable early disease diagnosis . These algorithms are validated for identifying human coughs and differentiating them from other external sounds.

Currently, work is underway to validate the classification of cough sounds, which would help distinguish their etiology based on the specific disease or pathogen—for example, distinguishing a cough from a COVID-19 patient from a tuberculosis patient.

Objectives

To analyze the role of artificial intelligence-based cough monitoring for the care of respiratory infections in household contacts in a district of Lima, Peru.

1. EXPAND

a database of cough sounds to develop and train an artificial intelligence algorithm for the classification of 5 respiratory infectious diseases.

2. RECORD

cough sounds through longitudinal monitoring analyzing changes in the frequency and pattern of coughing in high-risk contacts for respiratory infectious diseases.

3. EXPLORE

the initial steps for the implementation of health-related artificial intelligence tools through qualitative data collected from end users and authorities of the Peruvian health system.

Duration and Population of the Study

The study lasts 4 years.

The target population for Phase I of the study will be people who receive care in primary care facilities in the district of San Juan de Lurigancho (SJL) and in the Huaycán Hospital in the district of Ate, in Lima, Peru.

The target population for Phase II of the study will mainly be the participants selected in Phase I of the study, which include household contacts diagnosed with Tuberculosis, COVID-19, Influenza A, Influenza B or RSV, from health facilities in the district of San Juan de Lurigancho in Lima, Peru.

Study procedures

Phase I, related to Objective 1 and Objective 2, will collect cough sounds from contacts of patients with respiratory diseases such as tuberculosis, COVID-19, influenza A, influenza B and/or respiratory syncytial virus (RSV), which are circulating respiratory pathogens in Lima, Peru. In addition, the diagnostic accuracy of the cough monitoring tool will be evaluated in comparison with the reference diagnostic tests for infectious respiratory diseases in the national context. Patients confirmed for the aforementioned diseases will be consulted about members residing in the same household, who would be considered as contacts, then the contacts of the index patient will be recruited. A longitudinal cough record will be made of the contacts, depending on the case of the disease detected.

Phase II, through a qualitative component related to Objective 3, will evaluate the feasibility and acceptability of using an artificial intelligence tool for cough monitoring in a longitudinal and transversal manner, through semi-structured interviews directed at end users such as patients, contacts and health providers. Additionally, the feasibility of implementing artificial intelligence tools in the context of the national health system will be analyzed, through in-depth interviews with local or national health authorities.

Institutions