In Brazil, where over 17% of adults have chronic obstructive pulmonary disease (COPD), most remain undiagnosed because the standard test—spirometry—is not widely available in primary care clinics. This project will test whether an AI system that analyses a person’s voice can diagnose COPD and monitor its progression in Brazilian Portuguese, a language where the technique has never been applied. Researchers at the Federal University of São Carlos, University College London, and Maastricht University will first build a dataset of voices from people with and without chronic respiratory diseases. They will then measure how accurately the AI can distinguish diseased from healthy lungs, detect flare-ups that cause hospitalisation and lost productivity, and track improvements after pulmonary rehabilitation. If successful, the vocal biomarker could become a cheap, scalable diagnostic tool deployable in any primary care setting with a smartphone or microphone—bypassing the equipment shortages and inequities that currently leave millions of Brazilians untreated. The work is applied, not fundamental science: it adapts an existing technique to a new language and real-world clinical settings where the need is greatest.
View original technical description
CONTEXT: The burden and disability associated with chronic respiratory diseases (CRDs) is considerable, often falling on the most vulnerable in societies including those living in low- and middle-income countries such as Brazil. Specifically, the prevalence of chronic obstructive pulmonary disease (COPD) is increasing, and COPD presents a particular health challenge in Brazil where the prevalence in adults exceeds 17%. Most people with COPD in Brazil remain undiagnosed, and therefore untreated, because of limited access to the current diagnostic test called 'spirometry'. Spirometry is not widely, or equitably available in many primary healthcare settings in Brazil - including in Sao Paulo state. Innovative approaches to the diagnosis and management of COPD are therefore urgently required. THE CHALLENGE WE ADDRESS: We seek to transform the diagnosis of CRDs in primary care in Sao Paulo state, Brazil. We will do this through the use of vocal biomarkers derived by artificial-intelligence analysis of speech patterns. This technique has shown promise in English and Dutch languages, as a diagnostic and prognostic marker in CRDs, but has not been applied in (Brazilian) Portuguese, nor been deployed in real-life primary care settings where the need for easier tools to diagnose CRDs is greatest. AIMS and OBJECTIVES: Our over-arching aim is to develop and test AI-derived vocal biomarkers to support better diagnosis and management of CRDs in Brazil. To do this, we will work as an equitable partnership between the Federal University of Sao Carlos (Brazil) and University College London (UCL), with voice analysis experts at the University of Maastrict (Netherlands). We will: AIM 1: establish a dataset of voices from individuals with and without CRDs in Brazil. AIM 2: test the discriminative accuracy of AI-derived vocal biomarkers to distinguish those with CRDs from those with normal lung function. AIM 3: evaluate the utility of vocal biomarkers in COPD to detect the development of exacerbations of disease which are the major cause of ill-health and lost productivity in COPD. AIM 4: evaluate the utility of vocal biomarkers in COPD to provide objective evidence of benefit from pulmonary rehabilitation (PR) programmes, reflecting improvements in breathlessness, health status, and exercise capacity. POTENTIAL APPLICATIONS and BENEFIT: Transforming diagnosis and management of CRDs in Brazil would have wide health, social and economic benefits and provide an exemplar AI-health solution in an area of considerable unmet need
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know