Helfie Cough AI
Research Paper Summary

Helfie's Cough.AI is backed by science. 

​

At Helfie we work closely with reputable and published research partners to design and develop the health AIs that underpin our technology.

 

Here’s  a summary of the scientific report titled "Development and clinical validation of Swaasa AI platform for screening and prioritization of pulmonary TB" from our research partner Salcit technologies. 

 

The Swaasa AI platform, developed by Helfie’s research partner Salcit Technologies, is a groundbreaking AI-driven tool designed to screen and prioritize pulmonary tuberculosis (PTB) cases through cough analysis. This platform, which Helfie cough AI is built on,  uses sophisticated deep learning and signal processing techniques to analyze cough sounds, identifying unique PTB signatures and enabling quick, accurate assessments without invasive procedures or laboratory infrastructure. Swaasa’s primary innovation lies in its use of a combined model—integrating a convolutional neural network (CNN) and a feedforward artificial neural network (FFANN)—which processes cough sound spectrograms and patient information to detect potential PTB cases with high precision.

 

The study supporting Swaasa’s clinical effectiveness involved a multi-phase validation process. In the derivation phase, Swaasa’s model was trained and optimized on cough data from over 500 subjects, both PTB-positive and PTB-negative, which allowed it to identify the distinctive cough signatures associated with TB against a background of various other respiratory conditions, such as asthma and chronic obstructive pulmonary disease (COPD). During the clinical validation phase, Swaasa achieved an impressive accuracy of 86.82%, with a sensitivity of 90.36% and specificity of 84.67%, surpassing the WHO’s criteria for community-based TB screening and proving its potential as an alternative to traditional diagnostic tools.

 

Furthermore, the pilot phase of Swaasa's deployment took place in a primary care setting, assessing its real-world applicability in detecting PTB cases. Here, the model identified PTB-positive cases with a 75% positive predictive value, further validating its suitability for field use. The non-intrusive nature of Swaasa makes it ideal for deployment in remote and resource-limited settings, where conventional TB tests, like sputum analysis or chest X-rays, may not be accessible. Operable via smartphone, Swaasa also provides instant results, making it suitable for use by community healthcare workers without needing specialized training or equipment.

 

Swaasa’s design aligns with the increasing interest in leveraging AI to enhance early detection and triaging of respiratory diseases. This platform not only contributes significantly to the timely diagnosis of PTB but also addresses the need for scalable, cost-effective screening solutions that can help reduce TB transmission. 

 

To learn more about the science and validation behind the Swaasa platform, you can read the full study here.