AI-based stethoscope reliably spots aortic stenosis

A stethoscope that uses artificial intelligence (AI) and has infrasound capabilities delivers accurate diagnosis of aortic stenosis (AS), a recent study has shown.

“AI-based electronic auscultation is a promising new tool for automatic screening and diagnosis of valvular heart disease,” said the researchers, led by Tamer Ghanayim of the Department of Cardiology, Lady Davis Carmel Medical Center, Haifa, Israel.

In this two-stage study, Ghanayim and colleagues developed an electronic stethoscope, based on AI and infrasound, for the diagnosis of AS. The AI-based stethoscope had subsonic capabilities and acoustic range of 3‒2,000 Hz.

In the first stage of the study, the researchers used VoqX to record heart sounds from 100 patients referred for echocardiography (derivation group), 50 with moderate or severe AS and 50 without valvular disease. They applied an AI-based supervised learning model to the auscultation data from the first 100 patients used for training, to construct a diagnostic algorithm, which was tested on a validation group consisting of 50 other patients: 25 with AS and 25 without.

The second stage of the study was carried out at a different medical centre. Here, the researchers tested the device on 106 additional patients referred for echocardiography, which included individuals with other valvular diseases.

Analysis of the data obtained at the aortic and pulmonic auscultation points from the derivation group revealed that the AI-based algorithm accurately identified moderate or severe AS, with a sensitivity of 86 percent and a specificity of 100 percent. [Am J Med 2022;135:1124-1133]

When applied to the validation group, the sensitivity and specificity of the device were 84 percent and 92 percent, respectively. In the additional testing group, the corresponding values were 90 percent and 84 percent. The sensitivity rates for mild, moderate, and severe AS were 55 percent, 76 percent, and 93 percent, respectively.

“Our initial findings show that an AI-based stethoscope with infrasound capabilities can accurately diagnose AS,” the researchers said.

Acoustic vs electronic stethoscopes

One of the most simple and practical diagnostic tools in medicine, the stethoscope currently has two different types that are available in the market, acoustic and electronic, according to a report by Marie-Claude Grenier and colleagues. [Am J Cardiol 1998;81:653-656; Circulation 1996;93:1250-1253]

Grenier, who works at the Institut de recherches cliniques de Montréal in Montréal, Canada, states that acoustic stethoscopes are useful because of their robustness and ergonomic design.

“However, they are not ideal because they attenuate sound transmission proportional to frequency, their frequency response shows maxima and minima at very specific frequencies due to tubular resonance effects, and differences in the transmission properties are observed between different models,” Grenier and her team reported. [J Acoust Soc Am 1992;91:2224-2228; Bull Europ Physiopathol Respir 1985;21:49-54; J Appl Physiol 1983;54:714-719; Am Heart J 1982;104:269-275]

“Because the intensity of heart sounds and murmurs is generally faint with some sounds below the threshold of hearing, amplification of the acoustic signal with a more uniform frequency response has been introduced in electronic stethoscopes to solve the main limitations of acoustic stethoscopes,” they added. [Clin Eng 1988;5:355-357; J Clin Monit 1986;2:151-154]

However, the sensitivity of electronic stethoscopes to manipulation artifacts and electronic and ambient noises offsets such advantages. In addition, they do not make allowances for the sensitivity of the human ear, which greatly varies as a function of frequency.

“Consequently, the acoustic stethoscope is still the one mostly used today, even if several sophisticated designs have been patented for the construction of electronic stethoscopes during the last 20 years,” the authors noted.