Skip to content
Open access

Artificial intelligence–based calcium scoring using 3D transesophageal echocardiography in aortic stenosis: a pilot study

Jul 2026 · Journal of Cardiovascular Imaging · Vol 34 · 0 citations · 36 references
Medicine

TL;DR

AI-based calcium quantification using 3D TEE is feasible and correlates with CT-derived scores, and this radiation-free approach may provide a promising alternative for assessing aortic valve calcification.

Abstract

Aortic valve calcium scoring by computed tomography (CT) is an established method for assessing aortic stenosis severity but is limited by radiation exposure and availability. Artificial intelligence (AI)-based calcium detection using transthoracic echocardiography has shown promise but depends on acoustic window quality. Transesophageal echocardiography (TEE), particularly 3D TEE, may overcome these limitations by providing improved visualization without radiation. The objective of this study is to evaluate the feasibility of AI-based quantification of aortic valve calcium using 3D TEE. In this prospective pilot study, 23 patients (median age, 76 years; 56.5% male) with moderate or severe aortic stenosis underwent 3D TEE and CT. Multiplanar reconstruction generated 1.5-mm diastolic short-axis slices. A computer vision–based model identified calcium-related speckles. An automated TEE calcium score was derived from the sum of calcium pixels across 11 frames per patient, which was compared with the CT Agatston score. The TEE calcium score showed a significant positive correlation with CT Agatston scores (r = 0.65, P < 0.001). Receiver operating characteristic analysis yielded an area under the curve of 0.87 (95% confidence interval, 0.69–1.00) for identifying severe calcification (cutoff, 68,813 pixels; sensitivity, 89.5%; specificity, 75.0%). AI-based calcium quantification using 3D TEE is feasible and correlates with CT-derived scores. This radiation-free approach may provide a promising alternative for assessing aortic valve calcification.

Read PDF

Similar papers

Aug 2026

Vendor-independent Aortic Valve Calcium Scoring in Patients with Discordant Echocardiographic Criteria of Severe Aortic Stenosis.

Purpose To apply a vendor-neutral Agatston score (vnAS) for aortic valve calcification across multiple CT systems and to aortic stenosis severity classification. Materials and Methods Anthropomorphic phantoms with varying degrees of calcification were scanned using seven different CT systems and a reference electron-beam CT system to determine the vnAS. The vnAS was calculated using a regression model, whose performance was evaluated by both the coefficient of determination and analysis of variance. The vnAS was subsequently validated on a retrospective patient sample from eight international centers. Thresholds of severe aortic valve stenosis were defined as 1274 and 2065 AU for women and men, respectively. Results The Agatston scores from all CT systems were strongly correlated with that of the reference system (R2 > 0.932). On the basis of their echocardiographic findings, 183 of 831 (22.0%) patients (mean age ± SD, 77 years ± 10; 351 of 831 [42.2%] women) had discordant echocardiographic aortic valve stenosis findings. Among women with discordant results, the vnAS was higher than the traditional aortic valve calcium score (median, 1693 AU [IQR, 908-2513 AU] vs 1181 AU [IQR, 726-1964 AU], respectively; P < .001). Similar findings were observed for men (median, 2658 AU [IQR, 1894-3822 AU] vs 2056 AU [IQR, 1408-3056 AU], respectively; P < .001). Using the vnAS, 8.5% of women and 17.8% of men with discordant echocardiographic findings were reclassified into the severe aortic stenosis group. Conclusion A previously developed calibration tool, the vnAS, was validated and reclassified patients with discordant echocardiographic findings into the severe aortic stenosis group. Keywords: CT, Echocardiography, Percutaneous, Cardiac, Valves, Calcifications/Calculi, Technology Assessment, Vendor-neutral Agatston Score, Aortic Stenosis, Aortic Valve Calcification, Phantoms Supplemental material is available for this article. © RSNA, 2026 See also commentary by Jaltotage and Leipsic in this issue.

M. Dobrolinska, N. R. van der Werf, M. Greuter et al. · 0 citations
Aug 2026

Diagnostic Value of Calcium Score Derived From Virtual Noncontrast, Non-ECG-gated Dual-energy CT as an Indicator of Severe Aortic Stenosis.

STUDY TYPE Retrospective monocentric study. PURPOSE To estimate the probability of severe aortic stenosis(SAS) by calculating the aortic valve calcium score (AVCS) using virtual noncontrast (VNC) reconstructions derived from non-ECG-gated dual-energy computed tomography (DECT) scans. MATERIALS AND METHODS This study retrospectively included 59 patients who underwent DECT scans for noncardiac indications and presented with visible aortic valve calcifications. AVCS was calculated on VNC images using semiautomatic software. All patients had an echocardiogram performed within 12 months before or after the DECT scan. We then analyzed the correlation between AVCS values and the degree of aortic stenosis. RESULTS Both male and female patients with SAS had a significantly higher mean AVCS (919.2 AU, SD=1262.4) compared with those with nonsevere aortic stenosis (NSAS) (127.9 AU, SD=278.6). AVCS demonstrated good diagnostic performance in distinguishing between severe and nonsevere AS in both female and male (AUC=0.91 and 0.95, respectively). Estimated thresholds values showed high sensitivity in both genders, with good specificity in men, and moderate in women. When analyzed irrespective of gender, AVCS demonstrated good ability to distinguish severe AS from mild and no stenosis (AUC=0.926), and moderate ability to distinguish moderate AS from mild and no stenosis (AUC=0.781). However, it lacked statistical significance in discriminating severe and moderate AS (AUC=0.714, P=0.1325). CONCLUSIONS Aortic Valve Calcium Score obtained on VNC images from DECT examinations is a promising method for noninvasively differentiation between severe and nonsevere aortic stenosis, which could be potentially useful in patients who have undergone examinations for noncardiac indications.

Jakub Byczkowski, Gregorio Chierchia, G. Muscogiuri et al. · 0 citations
Open access Aug 2026

CT-based multi-threshold calcium scoring for aortic valve assessment: comparable predictive performance across HU thresholds in a pre-TAVI cohort.

To determine whether higher Hounsfield unit (HU) thresholds for computed tomography-derived aortic valve calcium (CT-AVC) scoring provide additional hemodynamic or discriminatory value beyond conventional 130-HU Agatston scoring in classical high-flow/high-gradient severe aortic stenosis (AS). This single-center retrospective cohort included 63 consecutive pre-TAVI patients with trileaflet, classical high-flow/high-gradient severe AS. CT-AVC was quantified on non-contrast ECG-gated CT at 130, 200, 500, 800, and 1000 HU within a manually defined leaflet/annulus region of interest, excluding left ventricular outflow tract and mitral annular calcification. Associations with peak gradient, mean gradient, and aortic valve area were assessed using correlation and multivariable linear regression.Exploratory ROC analysis assessed discrimination of very severe hemodynamic burden, defined as a mean gradient ≥ 60 mmHg, within the established severe-AS cohort, with ROC areas compared using DeLong testing. CT-AVC at all HU thresholds was independently associated with higher peak gradient (β, 0.010 at 130 HU to 0.144 at 1000 HU; all P ≤ 0.005) and higher mean gradient (β, 0.0059 at 130 HU to 0.073 at 1000 HU; all P ≤ 0.024). Associations with aortic valve area were not statistically significant. Within-cohort discrimination of very severe hemodynamic burden was modest and comparable across thresholds (AUC 0.65-0.67), with no statistically significant difference by DeLong testing (P = 0.68). In classical high-flow/high-gradient severe AS, CT-AVC demonstrated consistent associations with transvalvular gradients across HU thresholds. Higher HU thresholds did not outperform conventional 130-HU scoring and showed only modest, comparable performance for within-severity hemodynamic stratification. These thresholds should be interpreted as complementary densitometric analyses rather than alternative diagnostic cut-offs.

Michael Welt, M. Alnees, Yazan Hamdan et al. · 0 citations
Aug 2026

Updates on Imaging Modalities for the Diagnosis of Aortic Stenosis.

Aortic stenosis (AS) is the most common degenerative valvular disease in elderly patients and is linked to high morbidity and mortality. Accurate diagnosis and risk stratification are critical for effective management. Transthoracic echocardiography is the standard diagnostic tool, but its reliance on flow-dependent parameters can lead to inconsistent grading, especially in low-flow, low-gradient, or normal-flow, low-gradient AS. Advanced echocardiographic methods, such as 3D imaging, stress echocardiography, and Doppler indices, such as the mean gradient-to-effective orifice area ratio, improve the evaluation of AS severity and assist in clinical decision-making. Computed tomography provides a flow-independent evaluation of AS. It uses noncontrast calcium scoring with sex-specific thresholds, along with contrast-enhanced angiography, for detailed anatomical assessment. These modalities are essential for procedural planning, particularly for transcatheter aortic valve replacement. Cardiac magnetic resonance (CMR) provides additional prognostic information. It quantifies myocardial remodeling and fibrosis, which are associated with outcomes and recovery potential. Emerging technologies are expanding diagnostic capabilities in AS. Examples include 18F-sodium fluoride positron emission tomography for detecting microcalcification, artificial intelligence-based ECG and echocardiography for early diagnosis, and 4D flow CMR. Integration of echocardiography, computed tomography, CMR, and emerging positron emission tomography and artificial intelligence-based approaches can help address diagnostic uncertainty. This integration helps refine AS subtype classification and inform individualized intervention strategies.

H. Itani, M. Moumneh, A. Zayed et al. · 0 citations