Accuracy of non-ECG-gated computed tomography angiography of the chest in assessment of left-sided cardiac chamber enlargement

Abstract

Purpose: To determine whether measurement of left ventricular (LV) and left atrial (LA) diameters on nongated computed tomography angiography (CTA) can detect left atrial and left ventricular enlargement (LAE and LVE) with high specificity.

Materials and methods: Ninety-nine patients who had undergone both nongated CTA of the chest and echocardiography within 1 week constituted the pilot group. On CTA, LA and LV diameters were measured in the axial plane and on a 3-chamber-view multiplanar reconstruction. These measurements were compared with echocardiography results. Receiver operating characteristic (ROC) curves were then generated to determine optimal cutoff values for diagnosis of LAE and LVE. These thresholds were applied to a validation group (n = 203), and sensitivities and specificities for cutoff values were determined. Simple κ-coefficients were calculated for interobserver agreement.

Results: In the pilot group, axial measurements of the LA and LV diameters were superior to multiplanar reconstruction measurements based on areas under the ROC (LV: 0.82 vs. 0.76, LA: 0.87 vs. 0.82). Using the ROC results, cutoff values of 5.5 and 4.5 cm were chosen for LVE and LAE, respectively. These thresholds were applied to the validation group, achieving an average sensitivity and specificity for LAE of 53% [confidence interval (CI): 37%-71% and 94% (89%-97%)], respectively. Average sensitivity and specificity for LVE were 41% (CI: 23%-59%) and 99% (CI: 96%-100%), respectively. Kappa coefficients for diagnosis of LAE and LVE were 0.70 and 0.81, respectively.

Conclusion: Thresholds for LA and LV diameter on nongated CTA can be chosen to provide specific, but not sensitive, detection of LAE and LVE.