Comprehensive Aortic Stenosis Characterization Using Multiview Deep Learning

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Aortic stenosis (n=45852). Intervention: EchoNet-AS automated multiview deep learning framework (combining B-mode/color Doppler video CNNs and automated Doppler peak velocity measurement) vs. Single-view deep learning models and finalized physician echocardiography reports (reference standard). Primary outcome: Discrimination of severe aortic stenosis (area under the receiver-operating characteristic curve [AUC]) (AUC 0.964 in KPNC held-out cohort (0.985-0.986 in temporal cohort, 0.985 at SHC, 0.989 at CSMC), 95% CI 0.952-0.973).

Q: What does this research mean for the field?

An integrated multiview deep learning framework combining echocardiographic video analysis and automated Doppler measurements accurately identifies severe aortic stenosis with AUCs between 0.964 and 0.989 across diverse external cohorts, outperforming single-view approaches. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Hirotaka Ieki; Yuki Sahashi; Miloš Vukadinovic; Meenal Rawlani; Christina Binder; Neal Yuan; Andrew P. Ambrosy; Alan S. Go; Wansu Chen; Mingsum S Lee; Bryan D. He; Paul Cheng; David Ouyang

doi:10.1016/j.jcmg.2026.07.013

This research is Trending!

October 1, 2026JACC. Cardiovascular imagingOpen Access

Comprehensive Aortic Stenosis Characterization Using Multiview Deep Learning

View Full Paper

Ask AI

Bookmark

Share

Why the study?

Accurate assessment of aortic stenosis requires integration of both structural and functional information, but existing AI models use only structural or functional data alone.

Does an automated multiview deep learning framework combining B-mode/Doppler echocardiography videos and automated peak aortic velocity measurements accurately classify aortic stenosis severity compared to physician interpretation and single-view models?

Key Points

To develop and validate EchoNet-AS, an automated multiview deep learning model that integrates structural video data and Doppler measurements to assess aortic stenosis severity.
EchoNet-AS combined video convolutional neural networks assessing valve motion with segmentation models automating aortic valve peak velocity measurements.
The model was trained on 210,193 images from 16,076 KPNC studies and evaluated on 1,588 held-out test studies, a temporally distinct cohort of 19,202 studies, and two external cohorts (2,415 studies from SHC and 9,038 from CSMC).
For detecting severe aortic stenosis, EchoNet-AS achieved an area under the receiver-operating characteristic curve of 0.964 (95% CI: 0.952-0.973) in the KPNC held-out cohort and 0.986 (95% CI: 0.983-0.989) in the temporally distinct cohort.
External validation demonstrated robust performance for identifying severe aortic stenosis, with an AUC of 0.985 (95% CI: 0.975-0.992) at SHC and 0.989 (95% CI: 0.986-0.992) at CSMC.
Across all cohorts, positive predictive values ranged from 0.731 to 0.848, negative predictive values from 0.909 to 0.991, recall from 0.622 to 0.911, and specificity from 0.917 to 0.996.

Study Design

Type

Cohort (n=45,852)

Multicenter

Yes

Structured PICO

Does an automated multiview deep learning framework combining B-mode/Doppler echocardiography videos and automated peak aortic velocity measurements accurately classify aortic stenosis severity compared to physician interpretation and single-view models?

P

Population

45,852 adult patients (48,324 TTE studies, 608,086 videos) undergoing transthoracic echocardiography for aortic stenosis assessment across three US medical centers (mean age 65.7-74.0 years across cohorts).

E

Exposure

EchoNet-AS, an open-source two-stage multiview deep learning framework integrating six video-based ResNet R(2+1)D convolutional neural networks (assessing PLAX, PSAX, apical 5/3-chamber B-mode and color Doppler views) with a DeepLabv3 segmentation model (EchoNet-Measurement) for automated aortic peak velocity estimation via a HistGradientBoostingClassifier ensemble.

C

Comparator

Finalized physician echocardiography reports by board-certified cardiologists (reference standard), single-view deep learning models (e.g., PLAX only), and multi-view video-only ensembles without Doppler peak velocity measurements.

O

Outcome

Area under the receiver-operating characteristic curve (AUC) for discrimination of severe aortic stenosis and clinically significant (moderate or severe) aortic stenosis.surrogate

Main Result

Effect estimate: AUC 0.964 in KPNC held-out cohort (0.985-0.986 in temporal cohort, 0.985 at SHC, 0.989 at CSMC) (95% CI 0.952-0.973)

Limitations

Ground truth for AS severity was based on clinical reports rather than a dedicated core lab analysis, which could introduce variability in the reference standard.
Peak aortic velocity was the only quantitative Doppler measurement incorporated; other measures like mean gradient or aortic valve area were not included.
Lack of prospective validation integrated into routine clinical workflow to establish effectiveness and impact on patient care.
Ground truth for AS severity was derived from finalized clinical echocardiography reports rather than a dedicated core lab analysis.
Only peak aortic velocity was incorporated as a quantitative measure, omitting other parameters such as mean gradient or aortic valve area.
Retrospective study design lacking prospective validation integrated into routine clinical workflow.

Ask about this paper

Answers are grounded in this paper and cite their sources.

Key expert perspectives

Captured external expert commentary on this paper, strongest first. Original sources are linked where available.

HI Hirotaka IekiPhysician-researcher, Stanford University

“Given that existing approaches do not fully leverage the complementary information available from multiple videos and Doppler measurements, we hypothesized that an integrated framework combining all relevant echocardiographic information could improve the accuracy and clinical reliability of AS assessment.”

Stanford UniversityNews Coverage

Implication

EchoNet-AS demonstrates that integrating multiview echocardiographic B-mode and color Doppler videos with automated Doppler peak velocity measurements provides highly accurate and generalizable classification of aortic stenosis severity across multiple institutions.

Abstract

BACKGROUND: Accurate assessment of aortic stenosis (AS) requires integration of both structural and functional information characterized by visual traits as well as quantitation of gradients. Existing artificial intelligence models use solely either structural or functional information. OBJECTIVES: The authors aimed to develop and validate an automated multiview deep learning framework to assess AS severity. METHODS: The authors developed EchoNet-AS, an open-source end-to-end integrated approach combining video-based convolutional neural networks to assess valve motion with segmentation models to automate measurement of aortic valve peak velocity and classify AS severity. RESULTS: EchoNet-AS was trained on 210,193 images from 16,076 studies from KPNC (Kaiser Permanente Northern California) and validated on 1,588 held-out test studies and a temporally distinct cohort of 19,202 studies. The final model was also externally validated on 2,415 studies from SHC (Stanford Healthcare) and 9,038 studies from CSMC (Cedars-Sinai Medical Center). Combining assessments from multiple echocardiographic videos and Doppler measurements, EchoNet-AS achieved excellent discrimination of severe AS with an area under the receiver-operating characteristic curve of 0.964 (95% CI: 0.952-0.973) in the KPNC held-out cohort and 0.986 (0.983-0.989) in the temporally distinct cohort, which was superior to models using single views or only Doppler measurements. The performance was consistently robust in distinct external cohorts with an area under the receiver-operating characteristic curve of 0.985 (0.975-0.992) at SHC and 0.989 (0.986-0.992) at CSMC. CONCLUSIONS: EchoNet-AS synthesizes information from both B-mode videos and Doppler images to accurately assess AS severity. Its strong performance generalizes robustly to external validation cohorts and shows potential as an automated clinical decision support tool.

Coverage & sources

Journal, society, and media accounts. Useful signal, not independent expert judgment.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Ieki et al. (2026) conducted a cohort in Aortic stenosis (n=45,852). EchoNet-AS automated multiview deep learning framework (combining B-mode/color Doppler video CNNs and automated Doppler peak velocity measurement) vs. Single-view deep learning models and finalized physician echocardiography reports (reference standard) was evaluated on Discrimination of severe aortic stenosis (area under the receiver-operating characteristic curve [AUC]) (AUC 0.964 in KPNC held-out cohort (0.985-0.986 in temporal cohort, 0.985 at SHC, 0.989 at CSMC), 95% CI 0.952-0.973).

synapsesocial.com/papers/6ac592e138c9af7ead3834f2 https://doi.org/10.1016/j.jcmg.2026.07.013

View Full Paper

Ask AI

Bookmark

Share

Comprehensive Aortic Stenosis Characterization Using Multiview Deep Learning

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Key expert perspectives

Implication

Abstract

Coverage & sources

Discussion

Cite This Study

Comprehensive Aortic Stenosis Characterization Using Multiview Deep Learning

Key Points

Study Design

Structured PICO

Main Result

Limitations

Ask about this paper

Key expert perspectives

Implication

Abstract

Coverage & sources

Discussion

Cite This Study

Keep reading

Keep reading