Going beyond Cardiomegaly: Evaluation of Cardiac Chamber Enlargement at Non–Electrocardiographically Gated Multidetector CT: Current Techniques, Limitations, and Clinical Implications

Abstract

Cardiac chamber enlargement is important in the prediction of morbidity and mortality for a multitude of cardiovascular processes. Although non–electrocardiographically (ECG) gated multidetector CT is a commonly used cross-sectional imaging modality to evaluate a litany of cardiothoracic processes, a standardized method for evaluating and reporting cardiac chamber size does not exist. This has led to heterogeneity in the reporting of cardiac enlargement at routine multidetector CT with most readers often using gestalt assessment and the term cardiomegaly, which does not implicate the chamber or chambers that are enlarged. The purpose of this review article is to highlight advantages and limitations of several techniques used to assess cardiac chamber size at non–ECG-gated multidetector CT and to provide readers with reproducible and rapid measurements to determine if cardiac chamber size is present. The long-term aim would be to promote discussions between radiologists and institutions that would result in improved accuracy and decreased variability when commenting on cardiac chamber size.

© RSNA, 2019

Summary

Advantages and limitations of several techniques to assess cardiac chamber size at non–electrocardiographically gated multidetector CT are highlighted and some important clinical features of cardiac chamber enlargement are outlined.

Essentials

• ■ Simple unidimensional, bidimensional, and cross-sectional area techniques have been shown to be reliable methods to estimate individual cardiac chamber size at non–electrocardiographically (ECG) gated chest CT.

• ■ Several techniques to assess cardiac chamber size at non–ECG-gated multidetector CT are provided; while the selection of a technique is ultimately the reader’s responsibility, we specifically identify those that we find most useful in our clinical practice.

• ■ Discrepancy of chosen thresholds to define chamber enlargement between similar or identical techniques exists and requires further investigation to standardize measurements.

Introduction

Cardiac chamber enlargement has been implicated as an important biomarker in the prediction of morbidity and mortality for an array of cardiovascular processes, including atrial fibrillation, myocardial infarction, stroke, and heart failure (1–5). Despite the ubiquity of non–electrocardiographically (ECG) gated multidetector CT of the chest and the nearly universal comment of heart size in radiology reports, cardiac chamber assessment at non–ECG-gated multidetector CT has yet to be standardized, leading to heterogeneity and inaccuracies in its reporting.

Qualitative assessment of chamber size based on visual inspection alone is difficult, particularly when mild (6). While manual volumetric quantification is considered a precise method for chamber size assessment, it is time-consuming and impractical for daily clinical practice (6,7). Possibly because of underlying misconceptions regarding the utility of non–ECG-gated multidetector CT for assessment of chamber enlargement, such as cardiac motion artifact and the supposed inadequacy of nonobliquity corrected axial planes, there are currently no established measurement guidelines. This often confounds diagnosis or prompts supplementary imaging, delaying treatment. Several studies have described a variety of quantitative and semiquantitative techniques to determine individual cardiac chamber enlargement (Table 1) (6,8–17). The purpose of this article is (a) to introduce the reader to a variety of emerging practical techniques to efficiently and reproducibly determine individual cardiac chamber enlargement at non–ECG-gated multidetector CT, (b) to detail the specific advantages and limitations of each technique, and (c) to discuss the clinical implications of abnormal findings.

Table 1:

Various Techniques Currently Available to Assess for Individual Cardiac Chamber Size

Note.—Summary of the various techniques currently available to assess for individual cardiac chamber size. AP = anteroposterior, BSA = body surface area, ECG = electrocardiography, LA = left atrium, LA-MACSA = LA maximal axial cross-sectional area, LV = left ventricle, NPV = negative predictive value, PCWP = pulmonary capillary wedge pressure, PH = pulmonary hypertension, PPV = positive predictive value, RHC = right-sided heart catheterization, TTE = transthoracic echocardiography, WHO = World Health Organization.

Technical Considerations

In the absence of standardization, the reader must consider several factors when choosing the appropriate technique to assess chamber enlargement. Accuracy in the assessment of chamber size should be the primary consideration with several of the techniques described later in this article having varying degrees of sensitivity and specificity, positive predictive values, and negative predictive values. Because the vast majority of non–ECG-gated multidetector CT studies are performed for noncardiac indications, techniques with a high specificity are particularly desirable to reduce unnecessary workup and patient anxiety. Both intraobserver and interobserver reproducibility are important with near fully automated, semiautomated, and simpler measurements traditionally demonstrating higher reproducibility rates than complex manual measurements (11). Efficiency and impact on productivity and workflow should be considered. Although volumetric quantification by manual or semiautomated techniques may be more accurate, they often require separate software, which necessitates additional training that can lead to increased interpretation time and workflow disruption in high-volume environments (8,11). Simpler techniques based on nonreformatted images that use unidimensional, bidimensional, or cross-sectional area measurements may be more advantageous in such settings.

Although numerous options to assess individual chamber size exist, ranging from simple techniques to more complex volumetric quantification, this article will focus on the former (6,8–17). Several of the techniques described later in this article are readily applicable by using tools that are part of the standard picture archiving and communication system measurement palette (6,8–10,12–14,16,17).

Left Atrium

Left atrial enlargement (LAE) has been established as an important prognostic marker for adverse cardiac outcomes including atrial fibrillation, stroke, myocardial infarction, congestive heart failure, and subsequent pulmonary hypertension (PH) (18). Studies have shown that LAE is an independent risk factor for new-onset atrial fibrillation and flutter (18,19). In atrial fibrillation, LAE has been associated with anticoagulation failure, as well as higher rates of recurrence in patients following ablation (20). LAE has been shown to correlate with elevated mean pulmonary capillary wedge pressures in heart failure and to be linked with some pulmonary processes, with one recent longitudinal study reporting correlation to an increase in prevalence and severity of obstructive sleep apnea (21,22).

Anteroposterior Diameter Technique

Huckleberry et al conducted a retrospective study in patients (n = 203) who had undergone contrast material–enhanced 16- or 64-section non–ECG-gated multidetector CT chest angiography for suspicion of pulmonary embolism or acute aortic syndrome (8). On axial nonreformatted images, left atrium (LA) size was assessed independently by two readers by measuring the maximal anteroposterior (AP) intraluminal diameter of the LA at the level of the aortic root (Fig 1, A). Multiplanar reformatted CT images were also reconstructed by individual readers to emulate those of standard echocardiographic parasternal long-axis planes with maximal LA intraluminal diameter measured. Both axial nonreformatted and multiplanar reformatted multidetector CT measurements were validated to those obtained with transthoracic echocardiography (TTE). By using a threshold of greater than 45 mm for LAE on nonreformatted axial multidetector CT images, the overall sensitivity was 53%, specificity was 94%, positive predictive value was 62%, and negative predictive value was 91%. Simple κ coefficient for interobserver agreement to detect LAE was 0.81. Female patients demonstrated both a higher sensitivity (62% vs 45%) and specificity (99% vs 95%) when compared with male patients. Interestingly, the measurements made on multiplanar reformatted images proved to be less accurate in determining LAE; an unexpected finding likely attributed to the introduction of measurement error by reader image manipulation.

Figure 1:

Axial contrast-enhanced non–electrocardiographically gated CT images show, A, anteroposterior diameter, B, transverse diameter, and, C, maximum cross-sectional area techniques to assess the left atrium (LA) with values consistent with LA enlargement.

Recently, Eifer et al conducted a retrospective study in patients (n = 217) who had undergone non–ECG-gated contrast-enhanced 64- or 320-section multidetector CT (17). LA size was determined on axial nonreformatted images by measuring the maximal AP intraluminal diameter while excluding the LA appendage and pulmonary veins. Measurements were obtained by a single fellowship-trained radiologist and compared with the maximum LA area at cardiac MRI. LAE at cardiac MRI was defined as an area greater than or equal to 29 mm² for men and greater than or equal to 27 mm² for women (22). LA AP dimensions for the detection of LAE at multidetector CT were greater than or equal to 45 mm for women (sensitivity, 53.8%; specificity, 95.8%) and greater than or equal to 50 mm for men (sensitivity, 46.5%; specificity, 92.2%). When a second reader analyzed a random subset (n = 40), the absolute intraclass correlation coefficient between readers was 0.813.

Although the AP diameter measurement is an easy-to-use rapid technique for assessing LAE, several limitations for these studies exist. It has been reported that significant differences in LA volume and diameter are present in healthy patients when measurements are not normalized for age, sex, and body surface area (23). Only in Eifer et al’s study was a sex-specific threshold chosen and measurements were indexed to body surface area. Neither study indexed measurements to age. Although a single threshold for LAE for an entire population is appealing, applying demographic or body surface area normalization to these techniques may further improve accuracy.

Between both studies, there is mild discrepancy between the chosen thresholds for LAE (>45 mm vs ≥45–50 mm) (8,17). Eifer et al acknowledged using higher thresholds to increase specificity at the cost of sensitivity, which ranged from 34%–58% (17). Discrepancy in chosen thresholds may also result from differing standards used in both studies. TTE has been reported to result in underestimation of cardiac chamber volume when compared with MRI (8,17,24).

Transverse Diameter Technique and Rapid LA Volume Estimation

Sohrabi et al conducted a retrospective study in patients (n = 222) with a history of atrial fibrillation who underwent 16- or 64-section contrast-enhanced non–ECG-gated multidetector CT angiography (6). On axial nonreformatted images, the maximal intraluminal AP and transverse LA diameters were obtained by a single reader. These were compared with LA volume determined on the same multidetector CT scans by using a modified Simpson method. The cross-sectional areas of the LA appendage and LA, excluding the pulmonary veins, were obtained for one of every five axial images, and the sum of the cross-sectional areas was multiplied by five times the section thickness. By using a transverse diameter threshold of greater than 73 mm for LAE, sensitivity, specificity, and accuracy were all 84%, positive predictive value was 96%, and negative predictive value was 52% (Fig 1, B). By using an AP diameter threshold of greater than 43 mm, a similar sensitivity (84%), decreased specificity (61%), decreased accuracy (44%), and decreased positive predictive value (56%) were determined compared with the transverse diameter. LA volume was also estimated by the following formula: (transverse diameter)² × AP diameter; LA volume demonstrated reasonable agreement according to Bland-Altman analysis when compared with the volume determined by using the modified Simpson method, with a mean difference of 2.18 mL.

Like the AP diameter technique, the transverse diameter technique is advantageous for its use of a single measurement and threshold without the need for postprocessing (6,8). A benefit of the transverse diameter technique is rapid estimation of LA volume requiring only one additional measurement. However, this study is limited, as the authors used the modified Simpson method on non–ECG-gated multidetector CT images as the method of validation, which may not accurately reflect true LA size. Images were analyzed by a single reader, and interobserver variability could not be assessed.

Whether measuring LA AP diameter is superior to measuring the LA transverse diameter is unclear. Sohrabi et al reported that the transverse diameter demonstrated superior sensitivity and accuracy in determining LAE when compared with the AP diameter techniques used in their study, as well as in Huckleberry et al (6,8). However, Huckleberry et al’s AP diameter technique demonstrated superior specificity (6,8). In contrast, Eifer et al found that the area under the curve for discriminating LAE was not significantly different between LA AP and transverse diameters for either men or women but reported better interobserver agreement for AP measurements when compared with transverse measurements (17). An explanation for these discrepancies may be differing reference standards, and direct comparative analyses between these methods could be performed for further clarification.

LA Maximal Axial Cross-sectional Area Technique

Jivraj et al performed a retrospective study in patients (n = 165) with PH who had undergone non–ECG-gated multidetector CT techniques including chest CT with or without contrast and CT pulmonary angiography (9). Two independent readers analyzed images. Between the level of the left ventricular (LV) outflow tract and the height of the mitral valve leaflets, the axial nonreformatted image with the largest LA area was determined. The LA maximal axial cross-sectional area (LA-MACSA) was obtained by using a freehand region-of-interest tool and outlining the inner contours of the chamber excluding the LA appendage and pulmonary veins (Fig 1, C). Maximal AP and transverse LA diameters were obtained on the same image. Receiver operating characteristic curves were constructed to determine if this technique was adequate in differentiating group 2 PH (defined as pulmonary capillary wedge pressure > 15 mm Hg determined by right-sided heart catheterization standard) from other causes of PH, as the former is associated with LAE. A threshold of greater than 2400 mm² LA-MACSA demonstrated a sensitivity of 44%, specificity of 93%, and accuracy of 69%. Although there was similar capability in predicting group 2 PH using either LA-MACSA or the AP diameter measured on the same image, a higher degree of interobserver agreement was observed when using LA-MACSA. In a similar study performed by Katikireddy and colleagues, an LA-MACSA threshold of greater than 2000 mm² in combination with normal right ventricular (RV) size had a sensitivity of 77%, specificity of 94%, positive predictive value of 92.5%, and negative predictive value of 80% in predicting group 2 PH (10).

Although promising, these studies are limited in determining an exact threshold for LAE as receiver operating characteristic curves were designed to differentiate patients with group 2 PH from patients with non–group 2 PH. Moreover, LA-MACSA was not compared with an LA volume standard. Prior studies have shown an excellent correlation of LA-MACSA with LA volumes, albeit on ECG-gated multidetector CT studies (7). While encouraging, further studies comparing LA-MACSA derived with non–ECG-gated multidetector CT examinations to LA volumes determined with either ECG-gated multidetector CT or cardiac MRI are necessary to determine a threshold for LAE.

Left Ventricle

LV enlargement (LVE) is associated with a wide variety of cardiovascular processes, including cardiomyopathy, myocardial infarction, valvular heart disease, PH, and congestive heart failure (25–28). LVE is considered a modifiable risk factor in the development of myocardial infarction, stroke, and sudden death in both symptomatic and asymptomatic patients (25). As these entities can be clinically silent until their manifestation, early detection is crucial for disease-modifying therapy.

Transverse Diameter Technique

Huckleberry et al investigated the use of a single measurement to determine LV size at 16- or 64-section contrast-enhanced non–ECG gated multidetector CT angiography in patients (n = 203) suspected of having pulmonary embolism or acute aortic syndrome (8). Two readers independently obtained LV size by measuring the maximal intraluminal transverse diameter, defined as the short-axis distance between the septal and lateral walls perpendicular to the long axis of the LV (Fig 2). TTE was used as a reference standard for LV size measurement. A threshold of greater than 55 mm demonstrated a sensitivity of 41%, specificity of 99%, positive predictive value of 86%, and negative predictive value of 91%. Simple κ coefficient for interobserver agreement to detect LVE was 0.7. When this threshold was applied to individual sexes, female patients demonstrated a lower sensitivity (30% vs 48%) but similar specificity (99%) compared with male patients. A retrospective study performed by Gladish and Daher investigating LV transverse diameter in patients with cancer by using non–ECG-gated multidetector CT, with TTE as a reference standard, demonstrated a threshold of greater than 55 mm was effective in identifying LVE (12). A similar independent study conducted by Kathiria et al using 16- or 64-section contrast-enhanced non–ECG-gated multidetector CT (n = 155) read by a single reader found that an LV transverse diameter threshold measurement of greater than 56 mm demonstrated a sensitivity of 78% and a specificity of 100% in predicting LVE when using TTE as a reference standard (13). Bland-Altman analysis demonstrated a high interobserver agreement with a mean difference of −0.020 cm and a Pitman test P value of .39.

Figure 2:

Axial contrast-enhanced non–electrocardiographically gated CT image demonstrates the left ventricular (LV) transverse diameter technique with an LV short-axis measurement of 60 mm indicative of LV enlargement. LV end-diastolic volume at a concurrent cardiac MRI was 257 mL, indexed at 128 mL/m², which is severely dilated.

In a retrospective study, Murphy et al evaluated patients with cardiomyopathies (n = 49) compared with control subjects (n = 26) who had undergone non–contrast-enhanced (n = 5) or contrast-enhanced (n = 70) 64-section non-ECG-gated chest multidetector CT (14). Two readers independently measured intraluminal LV transverse diameter on nonreformatted axial images at the anterolateral papillary muscle belly level. Quantitative multidetector CT measurements were compared with those of cardiac MRI. An LV transverse diameter threshold of greater than 47 mm demonstrated a sensitivity of 93% and a specificity of 88% in predicting LVE. While Murphy et al used five non–contrast-enhanced CT examinations, these studies can be particularly limited when attempting to precisely measure the intraluminal diameter given the close attenuations of blood and normal myocardium in the absence of intravenous contrast material.

Recently, a retrospective study conducted by Eifer et al (n = 217) showed an LV transverse diameter of greater than or equal to 58 mm for men and greater than or equal to 53 mm for women on contrast-enhanced 64- or 320-section non–ECG-gated axial chest multidetector CT images to be 93% specific for LVE determined with cardiac MRI for both men and women. LVE at cardiac MRI was defined as LV end-diastolic volume indexed to body surface area greater than or equal to 117 mL/m² for men and greater than or equal to 101 mL/m² for women. When a second reader analyzed a random subset (n = 40), the absolute intraclass correlation coefficient between readers was 0.948. The sensitivity was lower for women (33%) and moderate for men (69%). The authors acknowledged choosing thresholds that would result in lower false-positive rates to reduce patient anxiety and unnecessary testing (17).

The unidimensional transverse diameter technique is a method to rapidly assess LVE. A major limitation is discrepancy in the thresholds determined by the aforementioned studies. This may be a result of differing reference standards (cardiac MRI vs TTE) with the former demonstrating higher accuracy in assessing chamber size in a separate study (24). Other potential confounders include a comparatively lower number of subjects in the study conducted by Murphy et al (n = 75 vs n = 203 and n = 217). Only one of the studies indexed LV size to sex and body surface area, which, in addition to age and race, has been shown to result in significant differences in LV size ranges (17,29). Further investigation is required to validate threshold measurements and to consider indexing LV size to additional demographic data for increased accuracy.

Right Side of the Heart

Right atrial enlargement (RAE) and right ventricular enlargement (RVE) have been associated with several cardiac conditions. RAE may result from congenital abnormalities of the tricuspid valve and interatrial septum, as well as from acquired causes such as PH (30,31). RAE has recently been implicated as an independent predictor of clinically significant supraventricular arrhythmia in the setting of pulmonary arterial hypertension (31). RVE has also been shown to be a risk factor for heart failure and death, independent of LV mass (5,32).

Right-sided Heart Techniques

RV assessment with non–ECG-gated multidetector chest CT is commonly performed by comparing the transverse diameter to that of the LV, with a normal RV:LV ratio of less than or equal to 0.9. However, this ratio does not correlate to an absolute measurement of RV size and is instead used as a surrogate for RV dysfunction (33–35). In the presence of LVE, this ratio can be unreliable in detecting RVE, necessitating a technique to determine absolute RV chamber size. In a retrospective study conducted by Nuffer et al, right atrial (RA) and RV measurements were determined on contrast-enhanced non–ECG-gated multidetector chest CT images in patients (n = 142) undergoing evaluation for pulmonary embolism or transcatheter aortic valve replacement (16). Unidimensional RA and RV measurements were obtained by measuring the greatest intraluminal diameters in both the long and short axes on nonaxial reformatted images. Findings were correlated with those at TTE. By using a longitudinal diameter threshold of greater than 50 mm obtained in the long axis, prediction of RAE had a sensitivity of 71% and a specificity of 76%. By using a RV transverse diameter threshold of greater than 47 mm obtained in the short axis, sensitivity was 68% and specificity was 67%. Chamber measurements had good agreement between observers according to Bland-Altman analysis. Although the study is limited by the use of TTE as a reference standard, these data are promising and further investigations are warranted.

Eifer et al retrospectively analyzed patients (n = 217) who underwent 64- or 320-section contrast-enhanced non–ECG-gated multidetector chest CT and cardiac MRI (17). RAE was defined as greater than or equal to 30 mm² on cardiac MR images for both men and women. RVE was based on RV end-diastolic volume indexed to body surface area greater than or equal to 128 mL/m² for men and greater than or equal to 110 mL/m² for women. RA measurements were obtained on nonreformatted axial chest multidetector CT images by a single fellowship-trained reader. RA transverse diameter measurements were determined by measuring the largest intraluminal RA diameter perpendicular to the interatrial septum and parallel to the tricuspid valve plane excluding the RA appendage and coronary sinus (Fig 3, A). RA longitudinal diameter measurements were measured in the long axis orthogonal to the transverse measurement (Fig 3, B). RV size was assessed in a similar fashion with maximum transverse measurement determined in a plane perpendicular to the interventricular septum extending from the septum to the lateral wall (Fig 4). RA transverse diameter measurements of greater than or equal to 67 mm for men and greater than or equal to 64 mm for women had sensitivities of 59%–75% and specificities of 92%–93% for determining RAE. RA longitudinal diameter measurements of greater than or equal to 65 mm for men and greater than or equal to 60 mm for women demonstrated sensitivities of 48%–63% and a specificity of 92% for determining RAE. RV transverse diameter measurements of greater than or equal to 60 mm for men and greater than or equal to 57 mm for women had sensitivities of 63%–67% and specificities of 91%–94% in determining RVE. When a second reader analyzed a random subset (n = 40), the absolute intraclass correlation coefficients between readers for RA transverse diameter, RA longitudinal diameter, and RV transverse diameter techniques were 0.881, 0.727, and 0.854, respectively.

Figure 3:

Axial contrast-enhanced non–electrocardiographically gated CT images demonstrate right atrial enlargement using the, A, transverse and, B, longitudinal diameter techniques.

Figure 4:

Axial contrast-enhanced non–electrocardiographically gated CT image demonstrates use of the transverse diameter technique with right ventricle (RV) short-axis measuring 62 mm, signifying RV enlargement. RV end-diastolic volume at a concurrent cardiac MRI was 370 mL, indexed at 166 mL/m², which is moderately dilated.

There is considerable variability between thresholds selected by Nuffer et al and Eifer et al (16,17). This is probably largely because of Eifer et al choosing thresholds that resulted in higher specificities and indexing of RV chamber size to body surface area. Differing standards were also used. Initial data remain promising, and comparative analysis is warranted.

In Our Practice

In our practice, we find the most effective methods to assess cardiac chamber enlargement to be those that are rapidly applicable, highly reproducible, demonstrate a high specificity, and have easy-to-remember thresholds. Currently, there are no societal recommendations as to which technique or threshold to apply. We have adopted thresholds similar to Eifer et al as these had a high specificity, were sex specific, and used cardiac MRI as the reference standard (17). We round to the nearest integer of 5 mm for ease of remembrance, with nearly all values being slightly higher than what was reported, further increasing specificity. We acknowledge that this further decreases sensitivity, and should a chamber appear enlarged with numbers near sex-specific thresholds, particularly for those patients with a smaller body surface area, we may suggest the presence of chamber enlargement and offer correlation with TTE.

In general, we have found transverse and AP diameter techniques to be the most reproducible and least time intensive. For LAE, we suggest using the AP diameter measurement with a threshold of greater than or equal to 45 mm for women and greater than or equal to 50 mm for men. For LVE, we suggest using the transverse diameter technique with a threshold of greater than or equal to 55 mm for women and greater than or equal to 60 mm for men. For RAE, we suggest using the transverse diameter measurement with a threshold of greater than or equal to 65 mm for women and greater than or equal to 70 mm for men. In the presence of a normal-sized LV, we recommend visually comparing the maximum transverse diameter of the RV to the LV with a ratio of greater than 0.9 supportive of RVE. In the setting of LVE or if there is suspicion for RVE, we recommend assessing RV size by using the transverse diameter measurement with a threshold of greater than or equal to 55 mm for women and greater than or equal to 60 mm for men.

An easy-to-remember sex-specific summary would be: LAE greater than or equal to 45–50 mm, LVE and RVE greater than or equal to 55–60 mm, and RAE greater than or equal to 65–70 mm (Table 2). Sex-specific thresholds for each chamber are separated by an integer of 5 mm with women having lower cutoffs and a numerical value ending with the digit 5. Threshold differences between chambers is 10 mm, with LAE having the lowest value, RVE and LVE having the same thresholds, and RAE having the highest threshold.

Table 2:

Recommended Sex-specific Measurements to Assess Cardiac Chamber Enlargement

Note.—These sex-specific measurements to assess cardiac chamber enlargement are used by the authors in their practice.

Conclusion

Cardiac chamber enlargement may precede the clinical features of and risk-stratify those with cardiovascular disease, making size assessment an important component of all CT reports. Despite the clinical importance and almost universal commentary on heart size in routine CT reports, there are currently no universally agreed on standardized methods to assess individual cardiac chamber size on non–ECG-gated multidetector CT studies. Recently, several techniques have emerged, each with specific advantages and limitations. Although initial results are promising, further investigation and agreed on standardization of these techniques is necessary.