A promising tool for early detection of heart disease

Rheumatic heart disease, caused by rheumatic fever, is a serious and often fatal disease that is prevalent across Africa. Tragically, it disproportionately affects children and young adults in rural, impoverished conditions. Researchers at Carnegie Mellon University Africa (CMU-Africa) are exploring how to use artificial intelligence (AI) for early detection of this preventable disease, primarily through the use of AI-powered stethoscopes.

Assistant Teaching Professor Carine Mukamakuza and her research team in the AI Healthcare Research Laboratory recently published a review paper in the journal Computer Methods and Programs in Biomedicine Update which explored deploying AI-powered stethoscopes as a cost-effective alternative to traditional electrocardiograms.

For the study, the researchers reviewed 37 papers published from 2015 through 2025 that applied machine learning to analyze data from electrocardiograms and phonocardiograms to support accessible, scalable screening of rheumatic heart disease. The study supports the World Health Federation’s goal to reduce mortality from rheumatic heart disease 25 percent by 2025.

“There is a high prevalence of heart disease in Africa, and one of the root causes is the lack of early detection,” said researcher Damilare Emmanual Olatunji (MS EAI ’25). “The review shows that an AI-enabled stethoscope will help us to go into low-resource communities and test patients early without needing clinical experts on site.”

The review was supported with funding from the African Engineering and Technology Network (Afretec) — a pan-African network of technology-focused universities led by CMU-Africa. It was done as part of a larger research project with co-principal investigators Mukamakuza, Vijayakumar Bhagavatula (Carnegie Mellon University), Khalil El Khodary (American University in Cairo), and Francesco Renna (University of Porto). The project aims to transform inexpensive electronic stethoscopes into reliable testing tools that quantify the severity of reduced or blocked blood flow (stenosis) at the point of care, thereby enabling improved healthcare delivery in underserved regions of Africa.

To achieve this, they adopt a bidirectional integration of physics-based simulations of blood flow, pressure, and resistance within the cardiovascular system, and data-efficient machine learning to model and interpret stenotic blood flow acoustics using population-relevant datasets.

Validated insights are embedded into lightweight, physics-informed neural networks optimized for low-resource deployment, culminating in an AI-powered diagnosis and intervention decision support tool that is robust, accessible, and clinically reliable.

“The review is significant because it is a very good starting point for other researchers. It provides a bank of knowledge around the evolution of medical machine learning-based analysis on electrocardiograms and phonocardiograms,” explained researcher Godbright Nixon (MSIT ’25).

I derive joy when a solution has been built for good. One of the reasons I joined the lab is to contribute to a body of knowledge and to have a solution that impacts lives.

Julius Zannu (MSIT ‘26), AI Healthcare Research Laboratory, CMU-Africa

“One of the gaps in deploying this type of technology found through the study is explainability between researchers and clinicians,” said researcher Samuel Chol Buol (MS EAI ’25). “We need to be able to explain how this works in a way that makes sense to health care workers.” Mukamakuza explained that “the lab works hand-in-hand with clinicians on building solutions, and our focus is on developing simple tools anyone can use.”

Other gaps identified through the research include the need for standardized data sets, reliance on private data sets, low percentage of validation, and no assessment of cost effectiveness.

The researchers are already exploring the best way to deploy AI-powered stethoscopes in low resource settings and are building other AI solutions that will be deployed alongside the phonocardiogram technology.

“It has been a great experience working with graduate student researchers on this review paper. They are passionate about applying AI to healthcare in Africa,” said Mukamakuza. “This research also gives them important experience as many apply for Ph.D. programs.”

Julius Zannu (MSIT ’26) is personally motivated to work on this study. “What drives me is to build solutions for people,” he said, “I derive joy when a solution has been built for good. One of the reasons I joined the lab is to contribute to a body of knowledge and to have a solution that impacts lives.”