AI-enabled stethoscope provides reliable exacerbation information in home monitoring

Key takeaways:

• The StethoMe records auscultatory sounds and measures heart rate and respiratory rates.
• The best exacerbation discriminators were wheeze and rhonchi intensity in children.

Patients with asthma can use a stethoscope augmented with artificial intelligence to monitor their exacerbations at home and provide reliable data to their physicians, according to a study published in Annals of Family Medicine.

This device is effective in monitoring exacerbations among children as well, even those aged 5 years and younger, Andrzej Emeryk, MD, PhD, DSc, professor, department of pediatric pulmonology and rheumatology, faculty of medicine, University of Lublin, and colleagues wrote.

The StethoMe (StethoMe Sp. z o.o.) home stethoscope uses artificial intelligence (AI) to detect and record pathologic auscultatory phenomena such as continuous sounds including wheezes and rhonchi and transient sounds such as coarse and fine crackles. It also measures heart rate, respiratory rate and inspiration-to-expiration duration ratio (I/E).

The device wirelessly transfers these recordings and data to a mobile app, and the AI module analyzes them. The app then displays the results. Also, the app surveys users about other health information. The AI was trained using more than 10,000 recordings of respiratory sounds, and the StethoMe is CE certified as a class IIa medical device.

The 6-month study comprised 52 children (63.5% boys) aged 0 to 5 years, 38 children (76.3% boys) aged 6 to 17 years and 59 adults (27.1% men). In addition to the StethoMe, participants received a peak flow meter and pulse oximeter.

Participants used these devices to perform examinations once a day 30 minutes or later after administration of asthma control drugs for the first 14 days. For the rest of the study period, participants performed examinations at least once per week.

Also, participants performed examinations twice a day when they experienced exacerbations or other alarming symptoms. Assigned physicians could ask for additional examinations as well.

The participants performed 6,442 complete examinations and produced 41,872 recordings, although 6.4% did not meet quality criteria and were excluded from the study. These examinations included 282 from 54 participants that identified moderate or severe exacerbations.

Seventeen physicians — two internal medicine specialists, four pulmonologists, nine pediatricians, five allergologists and four family medicine specialists — including physicians with double specializations, analyzed data for each examination.

The researchers obtained receiver operating characteristics (ROCs) with the greatest curvatures for the classifiers that used all the provided data, including the StethoMe AI analysis results, peak expiratory flow (PEF), peripheral capillary oxygen saturation (SpO₂) and survey data in addition to the ROCs that only used the data from the StethoMe, including four pathologic sound intensities, heart rate, respiratory rate and I/E.

Area under the curve (AUC) values for younger children included 93.2% (95% CI, 92.1%-94.4%) for the full data set and 93% (95% CI, 92.1%-93.9%) for the data from StethoMe alone. AUCs for older children included 92.4% (95% CI, 90.9%-93.9%) for the full data set and 92.4% (95% CI, 91.1%-93.7%) for the data from StethoMe alone.

In the adult group, survey data performance pertaining to symptoms had an AUC of 92% (95% CI, 89.4%-94.6%), and the AUC for all input data was 93.7% (95% CI, 92.1%-95.3%), which the researchers called quantitatively similar. However, the AUC for the StethoMe parameters was 81% (95% CI, 75.1%-86.8%).

The heart rate, respiratory rate, I/E, PEF and SpO₂ data were close to the diagonal of the plot as well, the researchers continued, indicating poor performance in classification tasks for all age groups, with AUCs close to or less than 70%.

Specifically, the best single-parameter discriminators of exacerbations included wheeze intensity among young children with an AUC of 84% (95% CI, 82%-85%), rhonchi intensity among older children with an AUC of 81% (95% CI, 79%-84%) and survey answers for adults with an AUC of 92% (95% CI, 89%-95%).

Overall, wheezes and rhonchi were the most efficient single value parameters, the researchers added, with AUCs of approximately 70% for adults and 80% for children. A combination of parameters provided the greatest efficacy in terms of AUC overall, the researchers continued.

Although Global Initiative for Asthma guidelines say that asthma exacerbation evaluations in children aged 5 years and younger should be based on subjective assessments of the patient’s condition, the subjective information that the caregivers in this study provided was not sufficient for confirming or excluding exacerbations, with an AUC of 72% (95% CI, 70.1%-73.9%), the researchers said.

The data provided by the StethoMe devices including the survey data, along with PEF and SpO₂, had the best performance in identifying asthma in all the groups, the researchers said. Yet the parameters provided by StethoMe alone had equally high efficiency in both groups of children, they added.

Patient reporting of disease-specific symptoms was key to effective diagnosis of exacerbation in adults, the researchers said, with the addition of the StethoMe data, PEF and SpO₂ only improving performance slightly. Possibly, the researchers said, adults can describe their health status precisely, but caregivers may have difficulty describing their children.

Based on these findings, the researchers concluded that measurements of auscultatory phenomena, heart rate, respiratory rate and I/E with a home stethoscope aided by AI enable the detection of exacerbations without PEF measurement, particularly among children aged 5 years and younger.