Aortic Valve Stenosis Screening Using Deep Learning on Noninvasive Arterial Pressure Waveforms.

BACKGROUND: Aortic stenosis (AS) presents a significant health burden. Early identification of significant AS, defined as at least moderate AS, may facilitate timely confirmatory testing, referral, and management. OBJECTIVES: This study sought to develop and externally validate a deep learning model applied to arterial pressure waveforms obtained from a noninvasive finger-cuff system to screen for significant AS, using transthoracic echocardiography as the reference standard. METHODS: This prospective international multicenter cohort study was conducted from December 2020 through July 2025. Participants underwent concurrent transthoracic echocardiography and noninvasive arterial pressure waveform acquisition. A 1-dimensional convolutional neural network was developed using the Cleveland Clinic development cohort and evaluated in a temporally distinct, multicenter validation cohort including sites in the United States, Spain, and the Netherlands. RESULTS: Significant AS was present in 307 of 3,634 (8.4%) patients in the development cohort and 241 of 1,336 (18.0%) in the validation cohort. In the validation cohort, the model demonstrated an area under the curve of 0.893 (95% CI: 0.876-0.910), with a sensitivity of 92.5% (223/241; 95% CI: 88.5%-95.2%), negative predictive value of 97.5% (706/724; 95% CI: 96.1%-98.4%), specificity of 64.5% (706/1,095; 95% CI: 61.6%-67.3%), and positive predictive value of 36.4% (223/612; 95% CI: 32.7%-40.3%). For severe AS, the area under the curve was 0.906 (95% CI: 0.890-0.922), with a sensitivity of 94.3% (183/194; 95% CI: 90.1%-96.8%), negative predictive value of 98.5% (706/717; 95% CI: 97.3%-99.1%), specificity of 64.5% (706/1,095; 95% CI: 61.6%-67.3%), and positive predictive value of 32% (183/572; 95% CI: 28.3%-35.9%). CONCLUSIONS: Deep learning applied to arterial pressure waveforms obtained from a finger-cuff system serves as a promising screening test for significant AS.

Authors

Institutions

Publication Details

Journal
PubMed
Published
2026-11-20
DOI
https://doi.org/10.1016/j.jacadv.2026.103475
Primary Topic
Cardiac Valve Diseases and Treatments
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Aortic Valve Stenosis Screening Using Deep Learning on Noninvasive Arterial Pressure Waveforms.

Rhonda Miyasaka, Eline Kho, Jimmy Schenk, Serge C. Harb et al.
PubMed
Cardiac Valve Diseases and Treatments
article

Aortic Valve Stenosis Screening Using Deep Learning on Noninvasive Arterial Pressure Waveforms.

Rhonda Miyasaka, Eline Kho, Jimmy Schenk, Serge C. Harb, Arghavan Arafati, Richard A. Grimm, Joseph T. Kassab, Denise Petra Veelo, Alexander P. J. Vlaar, Patrick H. Collier, Samir Ramesh Kapadia, Brian P. Griffin, Jay Ramchand, Miguel Angel Gonzalez Reyes
article en

Abstract

BACKGROUND: Aortic stenosis (AS) presents a significant health burden. Early identification of significant AS, defined as at least moderate AS, may facilitate timely confirmatory testing, referral, and management. OBJECTIVES: This study sought to develop and externally validate a deep learning model applied to arterial pressure waveforms obtained from a noninvasive finger-cuff system to screen for significant AS, using transthoracic echocardiography as the reference standard. METHODS: This prospective international multicenter cohort study was conducted from December 2020 through July 2025. Participants underwent concurrent transthoracic echocardiography and noninvasive arterial pressure waveform acquisition. A 1-dimensional convolutional neural network was developed using the Cleveland Clinic development cohort and evaluated in a temporally distinct, multicenter validation cohort including sites in the United States, Spain, and the Netherlands. RESULTS: Significant AS was present in 307 of 3,634 (8.4%) patients in the development cohort and 241 of 1,336 (18.0%) in the validation cohort. In the validation cohort, the model demonstrated an area under the curve of 0.893 (95% CI: 0.876-0.910), with a sensitivity of 92.5% (223/241; 95% CI: 88.5%-95.2%), negative predictive value of 97.5% (706/724; 95% CI: 96.1%-98.4%), specificity of 64.5% (706/1,095; 95% CI: 61.6%-67.3%), and positive predictive value of 36.4% (223/612; 95% CI: 32.7%-40.3%). For severe AS, the area under the curve was 0.906 (95% CI: 0.890-0.922), with a sensitivity of 94.3% (183/194; 95% CI: 90.1%-96.8%), negative predictive value of 98.5% (706/717; 95% CI: 97.3%-99.1%), specificity of 64.5% (706/1,095; 95% CI: 61.6%-67.3%), and positive predictive value of 32% (183/572; 95% CI: 28.3%-35.9%). CONCLUSIONS: Deep learning applied to arterial pressure waveforms obtained from a finger-cuff system serves as a promising screening test for significant AS.

PubMedVol. 5(12 Pt 2)
Twitter (United States) (US), Cleveland Clinic (US), Cleveland Foundation (US), Hospital Universitario Reina Sofía (ES), Amsterdam University Medical Centers (NL), Edwards Lifesciences (United States) (US), University of Amsterdam (NL)
Cardiac Valve Diseases and Treatments
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon