Examine outcomes display that respiratory sounds recorded by smartphones throughout sleep at dwelling can reliably determine obstructive sleep apnea (OSA), presenting an modern and accessible various to conventional in-laboratory polysomnography (PSG). Between February 2022 and February 2023, researchers aimed to validate the efficacy of a prediction mannequin for OSA utilizing respiratory sound recorded from smartphones, particularly these with iOS and Android working programs, along with stage 2 PSG carried out at dwelling.
Girls sleeping subsequent to alarm clock | Picture Credit score: stokkete – inventory.adobe.com.jpg
The potential diagnostic examine, revealed in JAMA Otolaryngology: Head & Neck Surgical procedureconcerned contributors 19 years and older who underwent unattended stage 2 dwelling PSG. Respiration sounds have been concurrently recorded utilizing smartphones within the contributors’ own residence atmosphere. The examine included people with identified OSA and people with out a earlier analysis.
Of the 101 contributors, with a imply age of 48.3 years, and a close to equal gender distribution, the predictive mannequin’s efficiency was assessed at completely different apnea-hypopnea index (AHI) ranges.
For the iOS smartphone, the information revealed notable efficiency metrics in predicting OSA. Sensitivity values at completely different AHI ranges, together with 5, 15, and 30 per hour, have been persistently excessive at 92.6%, 90.9%, and 93.3%, respectively. Specificities registered at 84.3%, 94.4%, and 94.4%, indicating the mannequin’s precision in accurately figuring out non-OSA cases. The accuracy charges underscored the reliability of the predictive mannequin, standing at 88.6%, 93.3%, and 94.3% throughout the corresponding AHI ranges.
The Android smartphone additionally exhibited predictive capabilities for OSA. Sensitivity values at AHI ranges of 5, 15, and 30 per hour have been persistently excessive at 92.2%, 90.0%, and 92.9%, indicating the mannequin’s effectiveness in figuring out true constructive circumstances. Specificities have been measured at 84.0%, 94.4%, and 94.3%, highlighting the mannequin’s proficiency in accurately recognizing non-OSA cases. General, the accuracy charges additional emphasize the reliability of the predictive mannequin for the Android smartphone, reaching 88.1%, 93.1%, and 94.1% at AHI ranges of 5, 15, and 30 per hour, respectively.
“This diagnostic examine has produced promising outcomes relating to the efficiency of smartphone-based prediction of AHI in a house setting,” the researchers wrote. “The validation of smartphone AHI with stage 2 dwelling PSG as a reference suggests {that a} future situation through which clinicians use each in-laboratory PSG and residential smartphones to comprehensively assess the sleep traits of sufferers with OSA over a number of nights is inside attain.”
The examine authos acknowledged a number of limitations to their findings. They centered solely on complete AHI with out investigating AHI adjustments in numerous affected person positions throughout sleep, and on account of reliance on smartphone-recorded respiratory sounds, the examine lacked information on positional adjustments and oxygen saturation, limiting evaluation of their impression on AHI.
Secondly, the examine’s solely Korean participant pool and small validation inhabitants restricted the generalizability of the prediction mannequin to different racial and ethnic teams. To deal with this, future research ought to contain a bigger and extra various participant pattern, in accordance with the examine findings.
The researchers additionally acknowledged the well-documented subject of stage 2 dwelling PSG outcomes underestimating OSA and AHI ranges in contrast with stage 1 in-laboratory PSG. Regardless of excessive accuracy in relation to stage 2 dwelling PSG, the mannequin should underestimate OSA and AHI ranges.
Regardless of these limitations, the examine represents a big development in validating a deep-learning mannequin for predicting OSA utilizing smartphone-recorded respiratory sounds.
“By harnessing the comfort and accessibility of smartphones, our prediction mannequin provides the potential for a extra holistic understanding in people’ sleep patterns past the restrictions of the sleep laboratory,” the researchers concluded. “Nevertheless, you will need to acknowledge that steady refinement and validation of the mannequin can be crucial to make sure the accuracy and reliability of the outcomes.”
Reference
Han SC, Kim D, Rhee C, et al. In-home smartphone-based prediction of obstructive sleep apnea along with stage 2 dwelling polysomnography. JAMA Otolaryngol Head Neck Surg. Revealed on-line November 16, 2023. doi:10.1001/jamaoto.2023.3490
