Unmet Need
Digital stethoscopes, which is projected to reach a market of $500 million by 2025, have the capability of recording the sound while the physician listens to the patient’s heart. These recordings are often saved in a digital health record. However, in busy clinical settings, the lung sounds can vary highly depending on the surrounding conditions. There also does not exist a method of measuring the quality of a recording.
Technology Overview
The technology is a metric of assessing the quality of a lung recording obtained with a stethoscope. It uses a combination of signal-derived features (e.g. spectral shape, temporal dynamics) and learned embeddings (via a 3 layer convolutional neural network) to assess whether the signal is valuable or has been masked by ambient sounds, making it difficult to interpret for physicians or for machine learning classification systems.
Stage of Development
A demonstration of the software has been completed through 2 expert physicians. It is currently being validated by more experts. However, the software is not yet directly usable by an end user.
Publications
Conference paper submission: Kala A, Husain A, McCollum ED, Elhilali M. An objective measure of signal quality for pediatric lung auscultations. Engineering in Medicine and Biology Conference. July 2020.
