Deep learning for predicting transvalvular gradient outcomes for patients undergoing transcatheter aortic valve replacement for native aortic stenosis

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Native aortic valve disease (predominantly aortic stenosis) (n=179). Intervention: Deep learning-based predictive model vs. Clinical measurement (ground-truth post-TAVR Doppler echocardiogram). Primary outcome: Point-by-point agreement (accuracy) and average prediction error in mean and peak gradient between predicted and clinically measured post-TAVR gradient waveform.

Q: What does this research mean for the field?

Deep learning models can accurately predict post-TAVR transvalvular pressure gradient waveforms and metrics using only preprocedural Doppler echocardiography for both balloon-expandable and self-expandable valves. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Wenyuan Song; Dhruv Polsani; Taylor Sirset-Becker; Luis René Mata Quinonez; Pradeep Yadav; Vinod Thourani; Lakshmi Prasad Dasi

doi:10.1016/j.xjse.2026.100142

August 7, 2026JTCVS structural and endovascular.Open Access

Deep learning for predicting transvalvular gradient outcomes for patients undergoing transcatheter aortic valve replacement for native aortic stenosis

View Full Paper

Ask AI

Bookmark

Share

Key result

Deep learning predicts post-TAVR gradient waveforms from preprocedural echo with ~85% accuracy.

n=179

Figures

Why the study?

To develop a deep learning-based model to preprocedurally predict post-TAVR gradient waveforms using only pre-TAVR echocardiographic data for both balloon-expandable and self-expandable valves.

Does a deep learning model accurately predict post-TAVR transvalvular gradient outcomes using pre-TAVR echocardiographic data in patients undergoing TAVR for native aortic stenosis?

Study Design

Type

Cohort (n=179)

Multicenter

No

Structured PICO

Does a deep learning model accurately predict post-TAVR transvalvular gradient outcomes using pre-TAVR echocardiographic data in patients undergoing TAVR for native aortic stenosis?

P

Population

179 patients with native aortic valve disease undergoing TAVR who had paired pre- and postprocedural Doppler echocardiograms available for deep learning model training and validation.

E

Exposure

Deep learning-based predictive model using pre-TAVR echocardiographic information (continuous-wave Doppler scan) and deployment scenario (THV type and size) to predict post-TAVR gradient waveform.

C

Comparator

Clinically measured post-TAVR transvalvular pressure gradient obtained from Doppler echocardiography (ground-truth).

O

Outcome

Point-by-point agreement between predicted and clinically measured post-TAVR gradient waveform, and average prediction error in mean and peak gradient.surrogate

Limitations

Developed on a small, single-institutional cohort, exposing the model to the risk of overfitting.
Underrepresentation of the smallest THV sizes (e.g., no SE THV smaller than 26 mm).
Patients at the extremes of body size (BSA <1.3 or >2.5 m2) were largely absent.
Exclusion of patients with specific procedural details such as residual paravalvular leak.
Gradients were derived from routine clinical echocardiograms rather than core-laboratory studies, and paired invasive catheter gradients were not available.
Cohort was predominated by patients with AS only, with few mixed AS/AR or pure AR patients included.
Small, single-institutional cohort risking overfitting
Underrepresentation of smallest THV sizes and extremes of body size
Exclusion of patients with residual paravalvular leak
Gradients derived from routine clinical echocardiograms rather than core-laboratory studies or invasive catheter gradients
Predominantly patients with AS only, with few having both AS and AR

Implication

A deep learning model can accurately predict post-TAVR transvalvular pressure gradients using only pre-procedural echocardiographic data, potentially aiding in pre-procedural planning and valve selection.

Abstract

Objectives: To develop a deep learning-based predictive model for preprocedurally predicting post-transcatheter aortic valve replacement (TAVR) gradient waveform using pre-TAVR echocardiographic information only for candidates for both balloon-expandable and self-expandable THV in a TAVR procedure. Methods: A total of 69 patients (mean age 81 ± 9.04 years, 58% female) receiving Edwards SAPIEN 3 and 77 patients (mean age 85 ± 8.56 years, 43% female) receiving Medtronic Evolut were included for pressure gradient collection. Two deep machine learning models were trained on the cohorts, respectively, each using the paired pre/postprocedural gradient waveform. Results: The SAPIEN model demonstrated an average accuracy of 84% on point-by-point agreement between predicted and clinically measured post-TAVR gradient waveform and an average prediction error of 2.1 and 5 mm Hg in mean and peak gradient, specifically. This group of metrics were reported as 87%, 1.4 mm Hg, and 3.1 mm Hg for the Evolut model. Bland-Altman analyses on both cohorts showed >90% agreement between clinical measurement and prediction for both mean and peak gradients. The SAPIEN model was additionally validated on a prospective cohort (n = 33) with average prediction error of 2.2 and 4.8 mm Hg for mean and peak gradient. Conclusions: A deep machine learning rationale was introduced to infer the full post-TAVR pressure gradient pattern directly from the preprocedural one obtainable through Doppler echocardiogram. It helps guide decision-making for the prevention of various post-TAVR complications. Further studies are necessary to investigate the gradient change of other valve types under specific deployment scenarios in a lifetime timespan.

Discussion

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

Cite This Study

Song et al. (2026) conducted a cohort in Native aortic valve disease (predominantly aortic stenosis) (n=179). Deep learning-based predictive model vs. Clinical measurement (ground-truth post-TAVR Doppler echocardiogram) was evaluated on Point-by-point agreement (accuracy) and average prediction error in mean and peak gradient between predicted and clinically measured post-TAVR gradient waveform. A deep learning model predicted post-TAVR gradient waveforms from preprocedural echocardiograms with 84% and 87% point-by-point accuracy for SAPIEN and Evolut valves, respectively.

synapsesocial.com/papers/6aa538085175d8ad94395d1a https://doi.org/10.1016/j.xjse.2026.100142

View Full Paper

Ask AI

Bookmark

Share

Deep learning for predicting transvalvular gradient outcomes for patients undergoing transcatheter aortic valve replacement for native aortic stenosis

Figures

Study Design

Structured PICO

Limitations

Implication

Abstract

Discussion

Cite This Study

Deep learning for predicting transvalvular gradient outcomes for patients undergoing transcatheter aortic valve replacement for native aortic stenosis

Figures

Study Design

Structured PICO

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading