Reduced-Slice Left Ventricular Volumetry with Long-Axis Derived Annular Plane Reference: Agreement with Full Stack Segmentation in a Clinical Cohort

Background: CMR is the reference method for quantifying left-ventricular (LV) systolic function through measurements of End Systolic Volume (ESV), End Diastolic Volume (EDV), Stroke Volume (SV) and Ejection Fraction (EF). In routine practice, a cine short-axis (SAX) stack covering the LV from base to apex is acquired along with long-axis (LAX) views, often across multiple breath-holds. Even with the advent of novel acceleration techniques, this makes the acquisition time consuming. Previous work has explored simplified volume calculations based on reduced set of slices [1], but such approaches can bias volumes because the atrioventricular (AV) plane moves relative to the basal short-axis slice across the cardiac cycle. In this study, we address this limitation by using a single LAX view to define the AV plane and recover the annular-to-basal gap. We demonstrate agreement with full-stack volumetry in a clinical cohort.

Methods: We screened N=582 consecutive patients undergoing clinically indicated CMR at 1.5T (Siemens Healthineers). All patients were imaged using conventional bSSFP cine for both SAX and LAX (2, 3, and 4-chamber LAX). Images were analyzed with an automated pipeline (Segment 4.1.0.1 R14284b, Medviso) [2]. Of the 582 cases, N=32 were excluded for automatic-segmentation failure, N=19 for image quality limitations (implants/arrhythmias), and N=24 for improper slice planning. The final analysis cohort comprised N=507 patients (53 ± 16 years; 48.4% female).

We selected three SAX slices (basal, mid-ventricular and apical) and repeated for each of the LAX slices (2, 3, or 4ch). The basal-most slice was selected to remain in the ventricle cavity in end-systole. The mitral annulus was identified in the long-axis image and used to define the cap volume between the AV plane and the basal-most slice. The volume calculation is described in Figure 1. Agreement with the full-stack as reference was assessed using Bland-Altman bias and 95% Limits of Agreement (LoA) and Concordance Correlation Coefficient (CCC).

Results: Agreement with the full-stack segmentation is shown in Figure 2. Bland-Altman plots are shown in Figure 3. The 2ch + 3 SAX slice combination had the best agreement with the full-stack measurements, with a slight underestimation of EDV (-7.58 ml), ESV (-5.09 ml) and SV (-2.49 ml) on average and slight overestimation for EF (1.23%). The 3ch and 4ch variants showed slightly larger biases and wider LoA.

Conclusion: Among reduced-slice variants, 2ch + 3 SAX gave the best agreement with full-stack reference in a clinical cohort. The reason for this is open to further analysis. The biases observed for EDV, ESV and SV were low, and the bias observed for EF (+1.23%) is well within established CMR inter-observer reproducibility [3] and well below the minimal detectable change in a single patient [4]. The proposed method can reduce scan time by 73%, enabling faster scans without significantly sacrificing accuracy.

Figure 1: Volume calculation using three SAX slices and a two-chamber LAX slice. The red dots represent points along the segmentation line between the myocardium and the blood pool. H = Distance between the top slice and the basal sax slice, 𝐴_𝑇𝑜𝑝 = Area of the atrioventricular slice (circular region approximation), 𝐴_(𝐵𝑎𝑠𝑎𝑙 𝑆𝐴𝑋 ) = Area of the basal slice, before the atrioventricular slice V2, V3 and, V4 are calculated as per El-Rewaidy et al. [1]
Figure 2: Volumetric analysis comparison between the 2ch, 3ch, and 4ch + 3SAX methods while using full-stack values from SAX slices as reference for 507 patients. Here, CCC = Concordance Correlation Coefficient, Limits of Agreement in the Bland-Altman Analysis are +- 1.96 * Standard Deviation (SD). Bold text indicates the best results. EDV = End Diastolic Volume; ESV = End Systolic Volume; SV = Stroke Volume; EF = Ejection Fraction, CI = Confidence Intervals
Figure 3: Bland-Altman analysis for 2ch + 3SAX method with full-stack as reference. The dashed green line represents bias whereas the dotted red and the yellow lines represent the 95% Limits of Agreement.