Speaking up for COPD through artificial intelligence in Brazil: Analysis of voice biomarkers for better diagnosis, early identification of exacerbation and assessing the effect of pulmonary rehabilitation

• Grant number: 24/10232-0
• Support Opportunities: Research Projects - Thematic Grants
• Start date: February 01, 2025
• End date: January 31, 2028
• Field of knowledge: Health Sciences - Physiotherapy and Occupational Therapy
• Principal Investigator: Renata Gonçalves Mendes
• Grantee: Renata Gonçalves Mendes
• Principal researcher abroad: John Robert Hurst
• Institution abroad: University College London (UCL), England
• Host Institution: Centro de Ciências Biológicas e da Saúde (CCBS). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil
• Associated researchers: André Ricardo Backes ; Fábio Fernandes Neves ; Manuela Karloh ; Naiara Tais Leonardi ; Sami O Simons ; Valéria Amorim Pires Di Lorenzo
• Associated scholarship(s): 25/02011-7 - Early identification of exacerbation of COPD using voice and cough analysis through a machine learning model, BE.PQ; 25/12387-4 - Analysis of Voice Biomarkers in Chronic Obstructive Pulmonary Disease: Investigating a New Frontier for Assessing Motivational Profiles for Physical Exercise and Response to Pulmonary Rehabilitation, BP.DR

Abstract

The burden and disability associated with chronic respiratory diseases (CRDs) are considerable, particularly in the most disadvantaged societies such as Low and Middle-Income Countries. Prospectively, the chronic obstructive pulmonary disease (COPD) will continue to increase worldwide. With a population exceeding 210 million, COPD in Brazil constitutes a health challenge with the prevalence surpassing the global average of 17%. The current diagnostic test for COPD is spirometry, but this is not readily and equitably available in many primary healthcare settings leading to potential underdiagnosis and terrible consequences. Therefore, COPD remains a significant global health burden, yet innovative approaches using artificial intelligence (AI), offer promising avenues to facilitate early detection of respiratory diseases and monitor outbreaks, directly influencing management. Recent advances in speech and sound analysis have been demonstrating a fabulous potential from the identification and automated extraction of different features in speech. Evidence of speech as a digital biomarker are correlated with hospitalisation and mortality in different diseases as well as underlying the potential of speech analysis as a prognostic biomarker. This work is an equitable partnership between UCL and Universidade Federal de São Carlos. Main scientific challenge- The proposal will address the ability of AI voice biomarkers to enhance COPD management strategies in LMICs. We will do this in the following ways: AIM 1 - Setting up a dataset of voices from individuals with and without CRDs; AIM 2 - Testing the discriminative accuracy of vocal biomarkers to reliably differentiate people living with COPD from those with normal lung function in Brazil; AIM 3 - Evaluating the capability of vocal biomarkers to provide early detection of COPD exacerbations in people living with COPD in Brazil; AIM 4- Assessing the accuracy of voice as outcome parameter for pulmonary rehabilitation, including improvements in breathlessness, health status, and exercise capacity. To this proposal, 4 work packages (WPs) will address our aims: WP1 and WP2 - Mixed model design to (a) develop a voice dataset and develop a use-case of voice-based digital COPD diagnosis in Brazil; (a) develop an initial machine learning (ML)-classification model for detecting COPD using voice (b) develop a use-case of voice-based digital COPD diagnosis in Brazil. WP3 -Observational longitudinal study to validate the ML algorithm (voice-based disease deterioration) using a gold standard (EXACT score) and WP4 - Clinical trial study to assess voice as a biomarker of pulmonary rehabilitation effectiveness and exercise capacity. Therefore, in this proposal, we will embed novel diagnostic techniques, based on voice-based ML techniques, within this existing framework of case-finding to increase availability and access to diagnosis. Moreover, we will extend these techniques to pulmonary rehabilitation (PR), a cost-effective and evidence-based treatment of COPD, to explore whether voice can be used to monitor the efficacy of PR. If successful this would provide evidence for new ways of delivering and monitoring PR in different settings through digital technology. In summary, our project leverages international advancements in vocal biomarker research while addressing the specific needs of the Brazilian COPD population. By bridging scientific knowledge gaps and positioning within ongoing national and international research, we aim to pioneer the dissemination of vocal biomarkers in COPD management, ultimately improving patient outcomes and healthcare delivery in Brazil. (AU)