What is the clinical evidence from this study?

Study design: Cohort. Population: Aortic valve calcification (n=400). Intervention: nnU-Net deep learning segmentation vs. Fixed Hounsfield unit thresholds (450 HU and 850 HU). Primary outcome: Correlation of CT-derived calcium mass-equivalent with ex vivo gravimetric calcium weight (Pearson r 0.967, 95% CI 0.940-0.981, p=<0.001).

What does this research mean for the field?

Deep learning segmentation provides the most accurate quantification of aortic valve calcification on contrast-enhanced CTA, while an 850 HU threshold is the preferred fixed-threshold method when validated against true physical calcium weight. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

This study aims to validate fixed Hounsfield unit thresholds and a deep learning model for quantifying aortic valve calcification using ex vivo gravimetric analysis as the reference standard.

June 20, 2026BMC Cardiovascular Disorders0 citationsOpen Access

Validation of aortic valve calcification quantification on contrast-enhanced computed tomography against ex vivo gravimetric analysis: comparison of fixed Hounsfield unit thresholds and deep learning segmentation

Q: What are the key findings of this study?

Against physically weighed calcium, nnU-Net deep learning segmentation provided the strongest correlation (Pearson r = 0.967) and lowest bias (+6.2 mg) for quantifying aortic valve calcification.

Q: What does this research mean for the field?

Deep learning segmentation provides the most accurate quantification of aortic valve calcification on contrast-enhanced CTA, while an 850 HU threshold is the preferred fixed-threshold method when validated against true physical calcium weight. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Q: What question did this study set out to answer?

This study aims to validate fixed Hounsfield unit thresholds and a deep learning model for quantifying aortic valve calcification using ex vivo gravimetric analysis as the reference standard.

DJD JiangSichuan University WZWei ZhangNanjing University of Chinese Medicine LLLulu LiuHebei Medical University

Key Result

Against physically weighed calcium, nnU-Net deep learning segmentation provided the strongest correlation (Pearson r = 0.967) and lowest bias (+6.2 mg) for quantifying aortic valve calcification.

Key Points

This study aims to validate fixed Hounsfield unit thresholds and a deep learning model for quantifying aortic valve calcification using ex vivo gravimetric analysis as the reference standard.
Retrospective cohort study including 400 patients with CT-confirmed aortic valve calcification and normal aortic valves.
Validation of two fixed Hounsfield unit thresholds (450 and 850) and a deep learning model against gravimetric calcium weight.
Statistical analysis included Pearson correlation, Spearman correlation, and Bland-Altman analysis.
nnU-Net achieved Pearson r = 0.967 with gravimetric weight and the lowest bias (6.2 mg).
450 HU threshold showed Pearson r = 0.864 with significant proportional bias (p=0.024).
850 HU threshold provided better results than 450 HU but was outperforming by nnU-Net in error metrics without reaching statistical significance.

Study Design

Type

Cohort (n=400)

Blinding

Blinded to reference standard

Multicenter

Yes

Structured PICO

Does nnU-Net deep learning segmentation improve the accuracy of aortic valve calcification quantification on contrast-enhanced CTA compared to fixed HU thresholds, using ex vivo gravimetric analysis as the reference standard?

Population

400 adults (≥18 years) undergoing contrast-enhanced CTA, including 50 patients with aortic valve calcification who underwent surgical aortic valve replacement to provide ex vivo gravimetric validation.

Exposure

nnU-Net deep learning segmentation and fixed Hounsfield unit thresholds (450 HU and 850 HU) for quantifying aortic valve calcification on contrast-enhanced CTA

Comparator

Ex vivo gravimetric calcium weight of excised native leaflets (obtained via high-temperature ashing and analytical weighing)

Outcome

Correlation and agreement (Pearson r, bias, RMSE) between CT-derived calcium mass-equivalent estimates and ex vivo gravimetric weightsurrogate

Deep learning segmentation (nnU-Net) and an 850 HU fixed threshold provide highly accurate quantification of aortic valve calcification on contrast-enhanced CTA when validated against physical gravimetric weight.

Main Result

Effect estimate: Pearson r 0.967 (95% CI 0.940-0.981)

Absolute Event Rate: 0.967% vs 0.929%

p-value: p=<0.001

Limitations

Surgical validation cohort was a selected subgroup of patients undergoing SAVR, potentially enriching for complex anatomy or severe calcification.
All sites used the same scanner vendor platform and a harmonised acquisition protocol, requiring further validation for generalisability.
Gravimetric ashing yields total mineral mass without distinguishing calcium hydroxyapatite from other calcium salts.
CT-derived calcium mass-equivalent estimates depend on the calibration equation.
Small surgical validation cohort (n=50) limits the precision of subgroup estimates.
Five-fold cross-validation for internal model selection introduces the possibility of minor overfitting.
Dynamic threshold strategies were not compared against the gravimetric standard.

Abstract

Accurate quantification of aortic valve calcification (AVC) on contrast-enhanced computed tomography angiography (CTA) is pivotal for planning surgical and transcatheter aortic valve replacement. The optimal Hounsfield unit (HU) threshold for calcification detection on contrast-enhanced images remains unresolved, and every prior validation study has relied on non-contrast Agatston scoring—itself an imaging estimate—as the reference standard. This study validated two widely used fixed HU thresholds (450 HU and 850 HU) and a self-configuring nnU-Net deep learning model against ex vivo gravimetric calcium weight as an absolute physical ground truth. Four hundred patients were included in a retrospective cohort study with a pre-specified temporal validation split: 300 with CT-confirmed AVC and 100 with normal aortic valves. Fifty chronologically later AVC patients who underwent elective open surgical aortic valve replacement (SAVR) within seven days of clinically indicated pre-operative contrast-enhanced CTA formed the locked surgical validation cohort; their excised native leaflets underwent standardised high-temperature ashing (550 °C, 12 h) and analytical weighing (precision 0.1 mg) to obtain gravimetric calcium mass. The remaining 350 cases served exclusively for nnU-Net development (280 training / 70 internal validation). CT-derived calcium mass-equivalent estimates were quantified on the validation cohort and compared with gravimetric weight using Pearson and Spearman correlation and Bland-Altman analysis. The nnU-Net achieved the strongest observed correlation with gravimetric weight (Pearson r = 0.967; bias + 6.2 mg; RMSE 13.7 mg), significantly outperforming the 450 HU threshold for correlation ( r = 0.864; bias + 36.2 mg; RMSE 42.1 mg; Steiger p < 0.001) and showing a non-significant trend toward stronger correlation than 850 HU ( r = 0.929; bias + 17.5 mg; RMSE 23.9 mg; Steiger p = 0.085). Compared with 850 HU, nnU-Net provided lower bias and RMSE, although the difference in Pearson r did not reach statistical significance. The 450 HU method exhibited significant proportional bias ( p = 0.024), whereas neither 850 HU nor nnU-Net did. The nnU-Net achieved a mean Dice coefficient of 0.873 and intersection-over-union of 0.812. Against physically weighed calcium, nnU-Net deep learning segmentation provided the most favourable overall performance profile on contrast-enhanced CTA, with the lowest bias and RMSE and the strongest observed correlation. The improvement in Pearson correlation over 850 HU represented a non-significant trend, whereas the error and agreement metrics favoured nnU-Net. Among fixed thresholds, 850 HU substantially outperformed 450 HU, offering direct physical-rather than surrogate imaging-evidence to support 850 HU as the preferred fixed threshold in standard contrast-enhanced protocols.

问 AI

Bookmark

View Full Paper