My UM
My UM
Search

From Maastricht to Brazil: A new dawn for vocal biomarkers – detecting COPD in Brazil

Chronic obstructive pulmonary disease (COPD) remains one of the most under‑diagnosed respiratory diseases globally, especially in low‑resource settings where access to routine lung‑function testing is limited. NUTRIM researchers at the forefront of turning voice into a powerful digital biomarker have shared their Maastricht-based innovation, knowledge and expertise with partners in Brazil. 

In Brazil, COPD is a major public‑health challenge, with many people going undiagnosed for years. Together with researchers from the Federal University of Santa Catarina in Florianópolis and UFSCar (Universidade Federal de São Carlos), the Maastricht team developed an app to detect COPD through voice. The project aims to detect COPD using voice‑based screening via a smartphone app, bringing lung‑health diagnostics within reach of communities that need it most. They visited Brazil to see how local partners were deploying it in community clinics in deprived neighbourhoods. Patients simply speak into their phones while an algorithm “listens” for signatures of respiratory disease.

“Seeing a patient simply speak into a phone while an algorithm detects breathing patterns stays truly mind‑blowing.”

VOCAL: a new sound for global lung health screening

The VOCAL project combines machine learning, clinical expertise and mobile technology to analyse subtle changes in speech and breathing. These digital fingerprints can help identify individuals at risk of COPD, allowing early diagnosis and intervention. By integrating voice‑based screening into primary care, the project aims to make detection faster, cheaper and more equitable — both in Brazil and worldwide.

Building impact through collaboration

The visit to Brazil has already led to the creation of a COPD‑specific Brazilian voice dataset, linking speech samples with medical data such as spirometry results. “Combining research, data science, clinical expertise and mobile technology was exactly what we did during live studies at UFSCar.” This real‑world dataset now allows researchers to study how COPD affects the voice and train AI models to recognise subtle changes in speech and breathing patterns, turning them into risk scores for COPD. By integrating voice‑based screening into primary‑care pathways, the team aims to improve early detection, reduce diagnostic delays and ensure timely treatment.

The power of a single word: COPD innovation in Brazil

Lauren Reinders and Loes van Bemmel, PhD students at NUTRIM and based at MUMC+, were eager for their working visit to Brazil. “We had, of course, been working on the collaboration for quite some time,” says Lauren. “Our Brazilian colleagues had extensive experience with clinical trials, and we brought the innovative voice analysis tools from our TACTICAS and SPEAK studies. So we had a common goal.” “The aim was to help set up the study, share expertise on voice analysis and test the approach in a different context,” she adds. Loes: “It really is a two‑way exchange and a win‑win collaboration.”

In Brazil, Lauren and Loes saw COPD in a different light; whereas in the Netherlands smoking plays a major role in the development of the disease, there is greater variation in Brazil. There, in addition to smoking, occupational exposure with inadequate protective equipment such as face masks, air pollution and indoor cooking with poor ventilation also play a role. “This trip was a powerful reminder that innovation in digital health only has impact when it reaches the people and places that need it most.”

Eye‑openers in practice

The context was fundamentally different. Loes: “We came across things we’d never thought about. For example, the reference values for lung‑function equipment differ between Europe and Brazil.” ‘Here, it’s quite normal to have a trained respiratory physiologist. But doing that with a handheld device is, of course, very different from here; voice analysis is perhaps the most universal method.’

Illiteracy. Loes: “Fully informed consent isn’t always guaranteed, as some participants simply can’t understand questionnaires due to illiteracy. This excludes people unfairly — not because of cognitive issues, but because not everyone reads or writes.” Lauren and Loes recall a man who came to a health centre with his daughter but was excluded after struggling with a questionnaire. When asked to rate his symptoms from 0 to 4, he answered 5. “With the voice app, he likely would’ve been diagnosed instead of overlooked. The power of ‘A’ is accessibility and inclusive. This trip reminded us that innovation in digital health only really makes an impact when it reaches the people who need it most.”

“Voice is everywhere — if we can safely and reliably use it as a screening tool, we can reach many more people, earlier.”

Lessons back in Maastricht

Lauren and Loes: “We’ve started discussing things more openly and focusing on accessibility.” What real benefit does a participant gain? How do you keep someone motivated and make sure you’re truly helping them — not just later, but right now? Like in the SPEAK study, where you can provide immediate feedback.” “The study with Brazil isn’t finished yet — it’s a large‑scale project with several phases. We’ve now seen the full procedures for a participant in phase 1; this phase will run for about another year. It already has nearly 100 participants! And then phase 2 is coming up, in which they’ll give people an app for use at home that will allow us to monitor participants.” But the challenges are different from those in the Netherlands. You can’t assume the Wi‑Fi connection is stable — or even exists,” they note. “App development must account for these realities. Such factors can greatly impact results — a reminder that technology must adapt to the real world.” Loes: “In the world of machine learning, there’s a concept known as ‘garbage in, garbage out’ (GIGA) -  the output quality depends entirely on the input. AI tools are only as good as the data and instructions you provide.”

From Maastricht to São Carlos: vocal biomarkers in action

The research visit to Brazil has already led to concrete progress in detecting COPD. One major outcome is a new Brazilian voice dataset — recorded on smartphones and linked to clinical data like spirometry and symptom scores. It helps the team study how voices change with obstructive lung disease, validate Dutch‑based algorithms (the SpeaktoCOPD dataset) and expand to other languages. Building on these datasets, researchers are training machine‑learning models to analyse short voice samples to identify COPD or flag high‑risk individuals. Because the approach is low‑cost and non‑invasive, it could support routine screening and remote monitoring. 

Meanwhile, the MEMIC team in Maastricht (led by Koert Heinen) is working with the Brazilian team to develop a user-friendly app and a prototype for use in both clinical settings and communities. The focus is on creating a scalable, reliable tool for large-scale COPD screening — even in places where lung function tests are not yet feasible.

Turning real‑world voices into life‑saving insights. A vision for the future!

The Maastricht research team has recently been awarded a SPRINT grant as a follow‑up to a UKRI grant, enabling researcher exchanges between Brazil and Maastricht to further develop this technology. “Brazilian researchers will visit Maastricht to learn how to analyse voices, and we will travel to Brazil to teach about AI and voice in COPD.” This collaboration between NUTRIM/MUMC+, UFSCar and international partners shows how scientific innovation can have real‑world impact — bridging continents and making healthier futures possible.

Lauren and Loes: “Ideally, someone could visit their GP or health centre, say ‘A,’ and a simple phone call would reveal potential COPD symptoms. Isn’t that fantastic?” Not for self‑diagnosis — nuances matter — but for accessible screening and home monitoring to prevent flare‑ups, supported by personalized advice on lifestyle and rest. “This trip was so educational and enriching thanks to our collaboration and Brazilian colleagues,” they add. “Together, NUTRIM’s COPD research and the partnership in Brazil are advancing early diagnosis and treatment. If the voice analysis app becomes more widely available, we could reach many more people — though it will take time. Brazil offers a diverse, ideal research population. We’ve learned not everyone can read or write,” they reflect, “but almost anyone can say ‘A.’” That insight closes their journey on a hopeful note: technology that transcends borders, for global lung health.

Text: Danielle Vogt  - Photography: Lauren Reinders, Loes van Bemmel

Biography

Lauren Reinders studied medicine and is now a NUTRIM PhD student and medical researcher. Her research focuses on identifying variants of COPD, for example using the voice. The aim is to make diagnosis more accessible, and hopefully also to facilitate easier analysis for more specific treatment methods.

Loes van Bemmel has a background in artificial intelligence and data science and is now a NUTRIM PhD. Her research project, SPEAK, focuses primarily on recognising lung attacks in a person’s voice. This involves long-term monitoring of people known to have COPD, measuring their voice daily at home and observing how this relates to their symptoms.

Sami Simons is a NUTRIM researcher and pulmonologist at Maastricht UMC+. His clinical and research focus is on chronic lung diseases, particularly COPD. He has developed an app for lung patients that can detect from a person’s voice whether they are about to have a lung attack. In addition, he is responsible for the pulmonary function departments and participates in the training and clinical supervision of junior doctors.

COPD UM BRAZIL

Also read