Feasibility of Using Acoustic Tools for Improved Patient Experience with CPAP therapy

The Mask Identification Clinical Study was a single site pilot study for ongoing product development. Eighteen current PAP therapy users were provided with a new PAP therapy device with an external recording module to collect acoustic signals from their mask for a period of 7 nights. Participants used their device and current mask as normal, and completed a diary entry each night on their bedroom and PAP set-up. The study was conducted to evaluate the feasibility of using acoustic signals to accurately identify mask type in the user home environment. Participants in the study were those currently using CPAP therapy. Eighteen evaluable participants in total were targeted for this study. 6 participants of each of the following PAP mask types; ResMed P10, ResMed N20, ResMed F20. Acoustic signatures do not change significantly over multiple nights for a single patient, nor do they change significantly over time within a single night. However, although visual similarities are apparent, performance of a mask identification algorithm differ when training a machine learning model using controlled (bench) and field data. Training a model with controlled data alone is not representative of the home environment and leads to unsatisfactory classification of mask types from acoustic signatures acquired in the field.