The human body produces a symphony of sounds, each containing valuable information about our health. These bioacoustic signals have the potential to transform the way we detect, diagnose, and manage a wide range of health conditions, such as tuberculosis (TB) and chronic obstructive pulmonary disease (COPD).
Google researchers understand the power of sound as a health indicator and the widespread availability of smartphone microphones. This has led them to explore using AI to extract health insights from acoustic data.
Earlier this year, researchers unveiled an innovative AI model called Health Acoustic Representations (HeAR), designed to identify acoustic biomarkers for diseases like tuberculosis. This groundbreaking technology has the potential to revolutionize healthcare by simplifying disease diagnosis using sound.
The team at Google Research trained HeAR on a vast and diverse dataset of 300 million audio recordings, with a specific focus on training the cough model using approximately 100 million cough sounds.
HeAR learns to discern patterns within health-related sounds, paving the way for advanced medical audio analysis. HeAR consistently outperforms other models across a wide spectrum of tasks and excels at generalizing across different microphones, showcasing its unparalleled ability to discern valuable patterns in health-related acoustic data.
Furthermore, models trained using HeAR have showcased remarkable performance even with limited training data, addressing a critical need in the often data-scarce realm of healthcare research.
HeAR is now accessible to researchers, providing the means to expedite the development of tailored bioacoustic models with reduced data, setup, and computational requirements. The goal is to facilitate further exploration of models for specific health conditions and demographics, regardless of sparse data, cost constraints, or computational barriers.
Salcit Technologies, a pioneering respiratory healthcare company based in India, has developed a groundbreaking product called Swaasa. This innovative solution harnesses the power of AI to analyze cough sounds and assess lung health.
The company is now exploring the potential of HeAR to further enhance the capabilities of its bioacoustic AI models. Initially, Swaasa is leveraging HeAR to advance their research on early TB detection through cough sound analysis.
TB presents a significant public health challenge, with millions of cases going undiagnosed each year due to limited access to healthcare services. Improving the diagnosis of TB is crucial in the global effort to eradicate the disease. Leveraging AI technology can play a pivotal role in enhancing detection and making healthcare more accessible and affordable to people worldwide.
Swaasa has a proven track record of using machine learning to detect diseases early, effectively addressing the issues of accessibility, affordability, and scalability by offering location-independent, equipment-free respiratory health assessment. With the integration of HeAR, they envision a unique opportunity to expand TB screening across India, building on their existing research and making a meaningful impact on public health.
“Every missed case of tuberculosis is a tragedy; every late diagnosis, a heartbreak,” says Sujay Kakarmath, a product manager at Google Research working on HeAR. “Acoustic biomarkers offer the potential to rewrite this narrative. I am deeply grateful for the role HeAR can play in this transformative journey.”
The team is excited to have garnered support for this innovative approach from various organizations, including The StopTB Partnership, a United Nations-hosted organization dedicated to uniting TB experts and affected communities to eradicate TB by 2030.
“Solutions like HeAR will enable AI-powered acoustic analysis to break new ground in tuberculosis screening and detection, offering a potentially low-impact, accessible tool to those who need it most,” said Zhi Zhen Qin, digital health specialist with the Stop TB Partnership.
HeAR marks a significant leap forward in acoustic health research. The aim is to drive the advancement of diagnostic tools and monitoring solutions in TB, chest, lung, and other disease areas. Through this research, researchers aspire to enhance health outcomes for communities worldwide.
