Background Patients with severe aortic stenosis (AS) are at high risk of mortality, regardless of symptom status. Despite this, aortic valve replacement rates remain low for patients with severe AS due to challenges in identifying clinically significant AS in time. This has prompted the need to develop and investigate novel diagnostic modalities. The objective of this study was to develop and validate novel, non-invasive diagnostic algorithm leveraging seismocardiography (SCG) data to detect severe AS.

Method A device capable of collecting a single-lead ECG and a three-dimensional SCG signal using a microelectromechanical-based accelerometer was used to collect sensor data. Phase 1 data were collected for training and validation of an algorithm for AS detection. Phase 2 data were collected as a blinded independent test set with age-matched and sex-matched patients as controls.

Results In phase 1 of the study, 115 subjects (n=56 AS patients and n=59 controls; mean age 73.8±10.4 years) were collected for training and validation of an algorithm for AS detection. Once model development was complete, the frozen model was then evaluated in a fully independent, single blinded phase 2 cohort of 99 subjects (n=50 AS patients and n=49 controls; mean age 76.8±6.4 years) for final analysis. The algorithm accurately classified 89 out of 99 patients, with four true AS cases misclassified as controls and six true control cases misclassified as AS. The sensitivity, specificity and area under the curve of the model were 92% (95% CI 84.5% to 99.5%), 87.8% (95% CI 78.6% to 96.9%), and 96% (95% CI 91.9% to 99.9%), respectively.

Conclusions This SCG-based algorithm to detect severe AS demonstrated high sensitivity and specificity when tested in a blinded, age-matched and sex-matched cohort. These findings suggest that this technology may hold potential as a low-cost diagnostic tool for the detection of AS.