Artificial Intelligence in the Fight Against Tuberculosis: Google's Initiative with HeAR

The fight against tuberculosis (TB), a disease that affects millions of people annually, particularly in low- and middle-income countries, is gaining new momentum through the development of innovative technologies. A prominent example of this is Google's bioacoustic model Health Acoustics Representation (HeAR), which was introduced in March 2024. HeAR utilizes advanced deep-learning algorithms to analyze extensive datasets of sounds such as coughing and sneezing to recognize TB patterns and distinguish them from other respiratory diseases. This technology is based on publicly available audio samples collected worldwide, ensuring broad applicability. Google's collaboration with Salcit Technologies, an Indian company that developed the AI model Swaasa, showed promising progress in August 2024. Swaasa enables the recording of coughs via mobile devices. This partnership aims to facilitate TB screenings in resource-poor regions and make rapid initial diagnoses more accessible through AI-driven solutions. This initiative is supported by organizations like the Stop TB Partnership. This United Nations-supported group noted growing interest in utilizing AI to classify coughs in April 2024 and has mapped existing software solutions like HeAR and Swaasa AI. Zhi Zhen Qin, a digital health expert from the Stop TB Partnership, emphasized the significance of solutions like HeAR as pioneering in new methods of TB detection and treatment. However, these AI models face challenges such as biased data and a lack of real-world testing. Despite using large public databases, these often come from controlled environments, which can lead to inaccuracies. Traditional TB diagnostic methods such as laboratory tests have disadvantages compared to AI-based solutions due to their low accuracy and long waiting times, while AI solutions offer quick and easy diagnoses. Sujay Kakarmath from Google Research sees the potential in HeAR for new discoveries in the field of acoustic health biomarkers. Given the global shortage of radiologists, HeAR could also help alleviate the pressure on scarce medical resources.