Google’s AI Model HeAR Detects Diseases Through Sound Analysis

TLDR

  • Google has developed a bioacoustic foundation model called Health Acoustic Representations (HeAR) to help detect diseases through sound analysis.
  • HeAR was trained on 300 million pieces of audio data, including 100 million cough sounds.
  • Salcit Technologies, an India-based company, is using HeAR in their Swaasa app to analyze cough sounds for early detection of tuberculosis (TB).
  • HeAR can potentially help improve accessibility and affordability of TB screening in India.
  • The model can also be used to detect other conditions like chronic obstructive pulmonary disease (COPD) and potentially dementia.

Google Research has introduced a new artificial intelligence model that could revolutionize disease detection through sound analysis.

The Health Acoustic Representations (HeAR) model, a bioacoustic foundation model, is designed to help researchers build systems that can listen to human sounds and identify early signs of disease.

The Google Research team trained HeAR on an extensive dataset of 300 million pieces of audio data, carefully curated and de-identified. Of particular note, the cough model within HeAR was trained using approximately 100 million cough sounds.

This vast amount of data allows the model to discern patterns within health-related sounds, creating a robust foundation for medical audio analysis.

Shravya Shetty, Google Research Director of Engineering, explained that HeAR has shown superior performance compared to other models across a wide range of tasks. It has demonstrated an impressive ability to generalize across different microphones, indicating its proficiency in capturing meaningful patterns in health-related acoustic data.

One of the key advantages of HeAR is its ability to achieve high performance with less training data. This is particularly crucial in healthcare research, where data can often be scarce. The model’s efficiency could significantly accelerate the development of custom bioacoustic models, even in situations where data or computational resources are limited.

In a real-world application, Salcit Technologies, an India-based respiratory healthcare company, is already exploring the potential of HeAR. The company has integrated the AI model into their product called Swaasa®, which uses artificial intelligence to analyze cough sounds and assess lung health. Currently, Salcit Technologies is focusing on using HeAR to enhance their early detection capabilities for tuberculosis (TB) based on cough sounds.

The potential impact of this technology on TB detection is significant. TB is a treatable disease, but millions of cases go undiagnosed each year, often due to lack of access to healthcare services.

By improving early detection through accessible technology like smartphone-based cough analysis, HeAR could play a crucial role in making TB care more accessible and affordable for people around the world.

Sujay Kakarmath, a product manager at Google Research working on HeAR, emphasized the potential of acoustic biomarkers in transforming TB diagnosis.

“Every missed case of tuberculosis is a tragedy; every late diagnosis, a heartbreak,” Kakarmath stated. “Acoustic biomarkers offer the potential to rewrite this narrative.”

The StopTB Partnership, a United Nations-hosted organization aiming to end TB by 2030, has expressed support for this approach. Zhi Zhen Qin, a digital health specialist with the partnership, noted that solutions like HeAR could “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.”

Beyond TB, the HeAR model shows promise in detecting other respiratory conditions such as chronic obstructive pulmonary disease (COPD). Researchers are also exploring its potential in analyzing speech patterns for early signs of conditions like dementia.

Google has made HeAR available to researchers, aiming to accelerate the development of custom bioacoustic models. This move could potentially lead to breakthroughs in various areas of health research, from respiratory diseases to neurological conditions.

Source: https://blockonomi.com/googles-ai-model-hear-detects-diseases-through-sound-analysis/