Inter -observer variability in manual measurement of aortic root dimensions in pediatric patients: benefits of using a semi-automated tool

Abstract

Purpose or Case Report: Accurate measurement of maximal aortic root dimensions is important for informed decision making on the timing/nature of aortic valve replacement surgeries in pediatric patients. Currently, the observer manually measures the 1) Cusp to Commisure (Cu-Co), and 2) Cusp to Cusp (Cu-Cu) lengths to quantify maximum aortic root dimensions: This introduces significant inter-observer variability (IOV), especially if a followup study is performed in a different institution. Hypothesis: A simple post-processing tool that allows the observer to perform semiautomated measurements using reproducible landmarks will enable accurate quantification of maximal aortic root dimensions as well as decrease IOV, thereby leading to reproducibility of measurements. Methods & Materials: An image analysis tool was developed using MatlabTM that can automatically measure the Cu-Cu and Cu-Co lengths from easily reproducible landmarks namely: 1) location of the commissure, and 2) the external margin of the sinuses and aorta (Fig 1). The observer also performed manual measurements for comparison. Study Design: In this retrospective study, cine bSSFP CMR data (1.5∗1.5∗5 mm) of the aortic root of 11 patients with Tetrology of Fallot (Age: 16.6±7.5 years, LV EF: 59±5.5%) and 15 patients with Marfans' Syndrome (Age: 12.8±8 years, LV EF: 59±6%) were analyzed by two experienced observers. Data Processing: Anonymized MR images of the aortic root in the shortaxis view were analyzed in early systole and Cu-Co, and Cu-Cu lengths were measured. IOV was computed using Pearson's correlation coefficient (r) and Fisher's z-transformation (z). Results: The manual measurements under-estimated aortic root dimensions (Figure 2). The r values (inter-observer agreement) were better for the semi-automated in both ToF and MFS patients. The p-values for ztransformation were a) 0.32 for Cu-Co; 0.045 for Cu-Cumeasurements in ToF patients; and b) 0.27 for Cu-Co; 0.027 for Cu-Cu measurements for MFS patients. The p values for Cu-Cu measurements show a significant difference between the manual measurements and the semi-automated tool. Conclusions: Manual measurements under-estimate aortic root dimensions and have a significant higher IOV when compared to semiautomated measurements. We demonstrate that a simple semiautomated tool consistently captures the maximum dimension of the aortic root, and significantly improves IOV, and will be useful in detecting subtle interval change on serial studies conducted at varying imaging centers. (Figure Presented).