UC San Diego

Purpose

To investigate whether endocardial regional shortening computed from four-dimensional (4D) CT angiography (RS_CT) can be used as a decision classifier to detect the presence of left ventricular (LV) wall motion abnormalities (WMAs).

Materials and methods

One hundred electrocardiographically gated cardiac 4D CT studies (mean age, 59 years ± 14 [SD]; 61 male patients) conducted between April 2018 and December 2020 were retrospectively evaluated. Three experts labeled LV wall motion in each of the 16 American Heart Association (AHA) segments as normal or abnormal; they also measured peak RS_CT across one heartbeat in each segment. The data set was split evenly into training and validation groups. During training, interchangeability of RS_CT thresholding with experts to detect WMA was assessed using the individual equivalence index (γ), and an optimal threshold of the peak RS_CT (RS_CT*) that achieved maximum agreement was identified. RS_CT* was then validated using the validation group, and the effect of AHA segment-specific thresholds was evaluated. Agreement was assessed using κ statistics.

Results

The optimal threshold, RS_CT* of -0.19, when applied to all AHA segments, led to high agreement (agreement rate = 92.17%, κ = 0.82) and interchangeability with experts (γ = -2.58%). The same RS_CT* also achieved high agreement in the validation group (agreement rate = 90.29%, κ = 0.76, γ = -0.38%). The use of AHA segment-specific thresholds (range: 0.16 to -0.23 across AHA segments) slightly improved agreement (1.79% increase).

Conclusion

RS_CT thresholding was interchangeable with expert visual analysis in detecting segmental WMA from 4D CT and may be used as an objective decision classifier.Keywords: CT, Left Ventricle, Regional Endocardial Shortening, Wall Motion Abnormality Supplemental material is available for this article. © RSNA, 2023.