Artificial intelligence-based stethoscope for the diagnosis of aortic stenosis. (14th October 2021)

• Record Type: Journal Article
• Authors: Ghanayim, T; Bachner-Hinenzon, N; Hanuka, D; Adler, D; Adawi, S; Shiran, A
• Is Part Of: European heart journal. Volume 42(2021)Supplement 1
• Publication Date: 2021-10-14
• Subjects: Diagnostic Methods; Cardiology -- Periodicals; Heart -- Diseases -- Periodicals; 616.12005
• DOI: 10.1093/eurheartj/ehab724.1588
• Languages: English
• ISSNs: 0195-668X

Abstract

Abstract: Background: Although the stethoscope is used for auscultation in the diagnosis of valvular heart disease for more than 200 years, its diagnostic accuracy is limited and highly dependent on clinical expertise and acoustic range of the human ear. Purpose: To develop an electronic stethoscope, based on artificial intelligence (AI), for the diagnosis of aortic stenosis (AS). Methods: We developed an electronic stethoscope (VoqxTM, Sanolla) with subsonic capabilities and acoustic range of 0–2, 000 Hz. Using the VoqxTM, we recorded heart sounds from 100 patients referred for echocardiography (derivation group), 50 with moderate or severe AS (aortic valve area (AVA) ≤1.5 cm 2 ) and 50 without valvular disease, using the 5 standard auscultation points. An AI based supervised learning model was applied to the auscultation data from the first 100 patients, to construct a diagnostic algorithm that was then tested on a validation group (50 other patients, 25 with AS and 25 without AS). Results: The derivation group included 50 patients with AS (age 78±9 years, 28 males, AVA=0.87±0.19 cm 2, mean gradient 44±16 mmHg, 39 with severe AS) and 50 without AS (age 58±16 years, 39 males). The AI based algorithm using 1–4 auscultation position (all except the mitral position), correctly identified 47/50 AS patients (sensitivity 94%) and 49/50 patients without AS (specificity 98%), total accuracy 96% (Table). The algorithm was then applied to the validation group (AS: n=25, age 76±9 years, 12 males, AVA=0.88±0.21 cm 2, mean gradient 41±15 mmHg, 18 with severe AS; No AS: n=25, age = 59±18 years, 17 males). The algorithm correctly identified 21/25 AS patients (sensitivity 84%) and 24/25 patients without AS (specificity 96%), total accuracy 90% (Table 1). Conclusion: Our initial findings show that an AI based stethoscope can accurately diagnose AS. AI based electronic auscultation is a promising new tool for automatic diagnosis of valvular heart disease. Funding Acknowledgement: Type of funding sources: Private company. Main funding source(s): Sanolla Company