European Journal of Radiology 59 (2006) 270–275
Multi-slice computer tomography of left ventricular function with automated analysis software in comparison with conventional ventriculography Martine Gilard b,∗ , Pierre-Yves Pennec b , Jean-Christophe Cornily b , Ulric Vinsonneau d , Gregoire Le Gal c , Michel Nonent a , Jacques Mansourati b , Jacques Boschat b a Departments of Radiology, Brest University Hospital, Brest, France Departments of Cardiology, EA 3878, Brest University Hospital, Brest, France c Department of Internal Medicine, Brest University Hospital, Brest, France Service de cardiologie, hopital d’instruction des armees Clermont-Tonnerre, 29240 Brest Armees, France b
d
Received 1 December 2005; received in revised form 4 February 2006; accepted 17 February 2006
Abstract Purpose: To evaluate the accuracy of left ventricular volumetric and functional parameters from multi-slice computed tomography using automated analysis software, and to correlate results with those of invasive left ventriculography. Materials and methods: In 145 consecutive patients (mean age, 61 years ± 12) known or suspected to have coronary artery disease, a 16channel multi-slice computed tomography (Philips Mx8000 IDT 16) was performed using a standard technique. Using short-axis secondary multi-slice computed tomography reformations, we determined end-diastolic and end-systolic left ventricular volumes and ejection fraction with a commercially available analysis software capable of automated contour detection. Conventional left ventriculography was performed according to standard techniques within the following 24 h. Bland-Altman analysis was performed to calculate the limits of agreement and systematic errors between multi-slice computed tomography and conventional left ventriculography. Results: As determined by computer tomography, mean end-systolic (53 ± 29 mL) left ventricular volumes had an acceptable correlation with conventional ventriculography (67 ± 50 mL; r = 0.74; p < 0.001) and mean end-diastolic (119 ± 33 mL) left ventricular volumes a poor correlation with conventional ventriculography measurements (154 ± 69 mL; r = 0.41). Left ventricular ejection fraction (57% ± 14 versus 55% ± 14 for conventional ventriculography; r = 0.79) showed a very good correlation (p < 0.001). Bland-Altman analysis showed acceptable limits of agreement (±9.2% for ejection fraction) without systematic errors. Conclusion: The use of a multi-slice computed tomography with an automatic calculation software has a good correlation with conventional ventriculography findings and could accurately assess left ventricular function, but should not be used for ventricular volumes, because of biased estimations. © 2006 Elsevier Ireland Ltd. All rights reserved. Keywords: Computed tomography; Conventional ventriculography; Left ventricular function
1. Introduction The assessment of volumetric (left ventricle volume: LVV) and functional (ejection fraction: EF) parameters ∗ Corresponding author at: Department of Cardiology, La Cavale Blanche Hospital, 29609 Brest Cedex, France. Tel.: +33 2 98 34 75 05; fax: +33 2 98 05 32 77. E-mail address:
[email protected] (M. Gilard).
0720-048X/$ – see front matter © 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2006.02.010
is important for diagnosis, prognosis and management in patients with cardiac diseases [1–4]. Several techniques have been used for analyzing LVV and EF. Although echocardiography is applied routinely, radionucleide-ventriculography has been used and magnetic resonance imaging (MRI) has evolved into a preferred technique allowing determination of left ventricular mass, LVV and EF. Conventional cineventriculography is considered a practicable standard and was used in several large multi-center studies to determine
M. Gilard et al. / European Journal of Radiology 59 (2006) 270–275
EF [5,6]. A promising new imaging technique for the noninvasive detection of coronary artery disease is multislice computed tomography (MSCT), which enables high-quality images of the entire heart to be acquired within a single breath-hold. Recently introduced, 16-slice MSCT with dedicated cardiac reconstruction algorithms has excellent longitudinal spatial resolution. Image reformation can be performed in any desired plane. From diastolic and systolic MSCT images, LVV can be measured, and consecutive assessment of EF is also possible. Studies assessing EF by means of MSCT are scarce, but initial results have demonstrated a good correlation between EF assessed by MSCT and two-dimensional echocardiography or MRI [7–9]. The purpose of the present study was to investigate the feasibility of EF assessment from MSCT using automated analysis software in patients with suspected coronary artery disease, and to compare the MSCT results to those of conventional cineventriculography.
2. Methods 2.1. Patients Between January 2004 and June 2004, all patients admitted to our center for coronary angiography (symptomatic and/or positive stress test, pre-operative clearance, coronary bypass graft follow-up, or coronary angioplasty) were considered for inclusion in this study that first aimed at assessing diagnostic performance of MSCT for coronary artery disease, as compared with coronary angiography. Exclusion criteria were: renal insufficiency, atrial fibrillation, or iodine allergy. All patients gave their written informed consent. During the study period, 145 (110 male and 35 female) patients were included. MSCT was performed one day after conventional coronary angiography. Mean age was 61 years ±12. Of the 145 patients included, 65 patients (45%) had history of myocardial infarction. Prior coronary revascularization procedures included percutaneous coronary intervention in 54 patients (37%) and coronary bypass surgery in 21 patients (14%). The patients’ mean height was 169 ± 3 cm, mean weight 75 ± 17 kg, and mean bodymass index 26 ± 0.5. Conventional coronary angiography and MSCT entailed no complications in any patient. 2.2. MSCT protocol Using a 16-slice MSCT (Philips Mx8000 IDT 16, Eindhoven, The Netherlands), a volume data set was acquired (16 mm × 0.75 mm cross-sections; gantry rotation time, 420 ms; table feed, 2.8 mm per rotation), covering the distance from the carina to the diaphragmatic side of the heart, except for patients with coronary artery bypass graft, in whom a longer scan-range was used. Tube current was 400 mA, with a tube voltage of 120 kv. The entire heart was scanned during a single breath-hold. 110 ml of contrast agent was continuously injected at a rate of 4 ml/s. Automated
271
peak enhancement detection in the aortic root was used to start the image acquisition. Cross-sectional images were reconstructed with a slice thickness of 0.8 mm, at 0.4 mm intervals. Retrospective gating was used. Axial images at 0%, 33%, 40% and 50% of the RR interval were reconstructed in each patient. Using the workstation’s standard three-dimensional software, we obtained multiplanar reformations along the long axis of the left ventricle. A double-oblique short-axis orientation was generated by tilting the cut-plane parallel to the plane of the mitral valve and parallel to vertical long axis of the left ventricle. Then, multiple short-axis multiplanar reformations (12–16 slices) with a section thickness of 3 mm and a gap of 6 mm were produced to encompass the entire left ventricle from the apex to the last slice in which the mitral valve did not appear. Automatic endocardial and epicardial border tracking (LV–RV analysis software, Philips, Eindhoven, The Netherlands) was performed separately for each short-axis slice at end-diastole and end-systole. Results were reviewed visually. Results were reviewed visually. For each patient, automatic contour detection could be corrected manually if there were no more than three slices with minor contour detection abnormalities. If there were more than three unreadable slices, no corrections were performed and the patient was excluded from the analysis. Papillary muscles were included in left ventricular cavity. The LV volumes and EF were calculated from short-axis views. The volumes were calculated as the sum of the cavity areas multiplied by the section interval (section thickness + section gap), using Simpson’s method. All data sets were independently analyzed by two physicians experienced in MSCT, blinded to the angiography data. 2.3. Cineventriculography protocol Standard cineventriculography was performed using a 30◦ right anterior oblique projection, with injection of at least 30 mL of contrast medium at a flow rate of 12 ml/s, using a 6 French pigtail catheter. Semi-automatic contour-tracking was used to define the end-diastolic image, based on the frame with the largest ventricular silhouette, and the end-systolic image, based on the frame with the smallest ventricular silhouette. Image calibration used a metal ball with a diameter of 4.0 cm, with identical X-ray tube positions. End-diastolic and end-systolic LVV was determined using Simpson’s method. Cineventriculography was analyzed by two experienced cardiologists. 2.4. Statistics For all parameters, means ± S.D.s are given. For linear correlation analysis, the Pearson correlation coefficient R was computed using SPSS analysis software, release 12.0 (SPSS, Chicago, Ill). Bland-Altman analysis [10] was performed for each pair of values of EF, to calculate limits of agreement and systematic errors between the two modalities. The resulting
272
M. Gilard et al. / European Journal of Radiology 59 (2006) 270–275
power of this study was about 0.999. A P value of 0.05 or less was considered statistically significant.
3. Results 3.1. MSCT The mean heart-rate in the MSCT studies was 61.6 beats per min ±1.4. The mean reconstruction window for diastolic images series was 0% and systolic image reconstruction were
performed at 50% of the RR interval in 55%, at 33% in 36% and at 40% in 9% (Fig. 1). Automated analysis was, in 20 cases (14%), prevented by poor contour detection in more than three slices. The results are summarized in Table 1, for the 125 patients analyzed. For volume, there were missing data at the base of the ventricle as cross-section acquisition stopped just short of the mitral valve, and there were also missing data at the apex and inside the ventricle, since slices with deficient contour detection were excluded. Mean time for post processing of MSCT images was 6 min (standard deviation 2 min).
Fig. 1. Images show software capabilities of automated contour detection for analysis of MSCT data sets, generated in short-axis orientation (A and B). Endocardial and epicardial borders of systolic (C and E) and diastolic (D and F) were traced automatically, papillary muscles were included in the LV cavity. The LV base next to the mitral valve was not taken into account (E). Table 1 Left ventricular volume and ejection fraction (EF)
Cineventriculography MSCT MSCT: multislice computed tomography.
End-diastolic volume (mL)
End-systolic volume (mL)
EF (%)
n
154 ± 69 (44–447) 119 ± 34 (52–261)
67 ± 50 (8–383) 53 ± 30 (12–164)
55 ± 14 57 ± 14
125 125
M. Gilard et al. / European Journal of Radiology 59 (2006) 270–275
3.2. Cineventriculography End-diastolic and end-systolic LVV was determined using Simpson’s method for all patients. The results are summarized in Table 1 (only for the 125 patients analyzed on MSCT). 3.3. Agreement between MSCT and cineventriculography in the determination of EF There was a significant correlation between the two modalities for the estimation of the ejection fraction, R = 0.79; p < 0.001. The dispersion of the differences between EF as assessed with MSTC and with cineventriculography imaging is illustrated with Bland-Altman plots (Fig. 2).
273
3.4. Agreement between MSCT and cineventriculography in the determination of LV volumes For systolic and diastolic volumes, the correlation between LVV as assessed by MSCT and by cineventriculography were r = 0.74 (p < 0.001) and (r = 0.41), respectively. The dispersion of these differences in LVV is illustrated with Bland-Altman plots (Fig. 2): mean difference was −1.6; 95% confidence interval: −20 to +17. 4. Discussion This study shows that overall left ventricular function can be accurately assessed by MSCT using automated analysis software, similar to conventional angiography in patients evaluated for coronary artery disease. In contrast,
Fig. 2. A–C, scatterplots show correlation between LV measurements and end systolic volume (A), end diastolic volume (B), ejection fraction (C), by means of MSCT and cineventriculography. Bland-Altman plots (D) of ejection fraction show relationship between differences and means with MSCT and cineventriculography: the difference (y axis) between each pair (mean MSCT value minus mean angiography value) is plotted against the average value (x axis) of the same pair (mean MSCT value plus mean angiography value divided by 2).
274
M. Gilard et al. / European Journal of Radiology 59 (2006) 270–275
estimation of left ventricle volumes with MSCT is not accurate. Sixteen-slice (MSCT) provides good coronary artery visualization [11–13]. MSCT can also provide diastolic and systolic image reconstructions by a retrospective ECG-gating technique. After reformations in a true short-axis orientation at diastolic and systolic windows and automated contour detection, LV ejection can be calculated by Simpson’s method. So far, only initial results in studies with small numbers of patients have been reported for LV volume and function assessed by MSCT in comparison to cineventriculography [14–16]. This is the first study to have used 16-slice MSCT with automated contour analysis software. EF values correlate well between MSCT and cineventriculography: R = 0.79, in agreement with Heuschmid and Juergens, who used four-slice MSCT without automated analysis software [14,15]. Juergens et al. [14] showed a better correlation for EF as assessed by MSCT and cineventriculography if the Simpson method, as opposed to the area-length method, was used. The LVV correlation was less good: R = 0.74 for mean end-systolic volume and R = 0.41 for mean end-diastolic volume. Such poor correlations agree with Heuschmid’s findings (respectively, R = 0.81 and 0.51) [15]. The poor LVV correlation found in our study may be explained by our volume calculation method: the LV base next to the mitral valve was not taken into account, and slices that failed to enable automated detection were excluded. Such excluded values were greater in diastole. Currently, MRI is the non-invasive diagnostic gold standard for the assessment of LV volume, EF and regional myocardial function [17,18]. Recent studies have reported good correlations between MRI and MSCT results using manual contour definition [8,9,19]. Our study used coronary angiography, but not MRI, as the reference test. However, coronary angiography is the most widely used technique. Although this is a limitation of our study, our data demonstrated the feasibility of using automated analysis software. This allows functional analysis of left ventricle by MSCT without being longer limited by time-consuming secondary reformations. The present MSCT technique has some limitations. The software did no provide left ventricle parameters estimation in all patients. In fact, in 14% of patients, automatic contour detection was not accurate enough to allow any analysis. In this case, manual correction would be too time-consuming, and physicians might be reluctant to use this technique to assess the left ventricle in such patients. However, our data suggest that in patients with a correct automatic contour detection, the estimation of LV function by MSCT is reliable and easily collected. Another limitation is that we only obtained a limited number of reconstruction windows, and only till 50% of the RR interval. This does not cover the entire cycle, and in some patients, end-systole could have occurred out of the analyzed windows. Similarly, MSCT is not as precise as MRI for the definition of end-diastole. However, we
believe that such limitations should not have caused a differential bias in our estimations. As MSCT requires overlapping image acquisitions for retrospective ECG-gating, it entails higher radiation exposure than do conventional angiography or radiation-free techniques such as MRI or echocardiography [20,21]. Nevertheless, functional analysis can be performed using the same MSCT data-set, providing valuable information as an adjunct to coronary MSCT angiography with no extra radiation exposure or examination time, thanks to the automated analysis software.
5. Conclusion Left ventricle analysis based on MSCT data sets with an automated analysis software shows good correlation with conventional ventriculography findings for the estimation of the LV ejection fraction, but not for the LV volumes, because of biased estimations with this technique. The combination of non-invasive coronary artery imaging and of the assessment of cardiac function with a single breath-hold MSCT study might be an interesting approach to conclusive cardiac workup in patients with suspected coronary artery disease, completing assessment for patients with multiple cardiovascular disorders.
References [1] Hofmann T, Meinertz T, Kasper W, et al. Mode of death in idiopathic dilated cardiomyopathy: a multivariate analysis of prognostic determinants. Am Heart J 1988;116:1455–63. [2] Volpi A, De Vita C, Franzosi MG, et al. Determinants of 6-month mortality in survivors of myocardial infarction after thrombolysis. Results of the GISSI-2 data base. Circulation 1993;88:416–29. [3] Emond M, Mock MB, Davis KB, et al. Long-term survival of medically treated patients in the Coronary Artery Surgery Study (CASS) Registry. Circulation 1994;90:2645–57. [4] St John Sutton M, Pfeffer MA, Moye L, et al. Cardiovascular death and left ventricular remodeling two years after myocardial infarction: baseline predictors and impact of long-term use of captopril: information from the survival and ventricular enlargement (SAVE) trial. Circulation 1997;96:3294–9. [5] Pfeffer MA, Braunwald E, Moye LA, et al. Effect of captopril on mortality and morbidity in patients with left ventricular dysfunction after myocardial infarction: results of the survival and ventricular enlargement trial—the SAVE Investigators. N Engl J Med 1992;327:669–77. [6] The GUSTO Angiographic Investigators. The effects of tissue plasminogen activator, streptokinase, or both on coronary-artery patency, ventricular function, and survival after acute myocardial infarction. N Engl J Med 1993;329:1615–22. [7] Dirksen MS, Bax JJ, de Roos A, et al. Usefulness of dynamic multislice computed tomography of left ventricular function in unstable angina pectoris and comparison with echocardiography. Am J Cardiol 2002;90:1157–60. [8] Mahnken AH, Spuentrup E, Niethammer M, et al. Quantitative and qualitative assessment of left ventricular volume with ECG-gated multislice spiral CT: value of different image reconstruction algorithms in comparison to MRI. Acta Radiol 2003;44:604–11.
M. Gilard et al. / European Journal of Radiology 59 (2006) 270–275 [9] Juergens KU, Grude M, Maintz D, et al. Multi-detector row CT of left ventricular function with dedicated analysis software versus MR imaging: initial experience. Radiology 2004;230:403–10. [10] Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307–10. [11] Nieman K, Cademartiri F, Lemos PA, et al. Reliable noninvasive coronary angiography with fast submillimeter multislice spiral computed tomography. Circulation 2002;106(16):2051–4. [12] Martuscelli E, Romagnoli A, D’Eliseo A, et al. Accuracy of thinslice computed tomography in the detection of coronary stenoses. Eur Heart J 2004;25:1043–8. [13] Mollet NR, Cademartiri F, Nieman K, et al. Multislice spiral computed tomography coronary angiography in patients with stable angina pectoris. J Am Coll Cardiol 2004;43:2265–70. [14] Juergens KU, Grude M, Fallenberg EM, et al. Using ECG-gated multidetector CT to evaluate global left ventricular myocardial function in patients with coronary artery disease. AJR 2002;179:1545– 50. [15] Heuschmid M, Kuttner A, Schroder S, et al. Left ventricular functional parameters using ECG-gated multidetector spiral CT in comparison with invasive ventriculography. Rofo 2003;175:1349–54.
275
[16] Schuijf J, Bax J, Jukema JW, et al. Non-invasive angiography and assessment of left ventricular function using multislice computed tomography in patients with type 2 diabetes. Diabetes Care 2004;27:2905–10. [17] Pattynama PMT, Lamb HJ, van der Velde EA, de Roos A. Left ventricular measurement s with cine and spin-echo MR imaging: a study of reproducibility with variance component analysis. Radiology 1993;187:261–8. [18] Lorenz CH, Walker ES, Morgan VL, et al. Normal human right and left ventricular mass, systolic function, and gender differences by cine magnetic resonance imaging. J Cardiovasc Magn Reson 1999;1:7–21. [19] Koch K, Oellig F, Bender P, et al. Assessment of global and regional left ventricular function with 16-slice spiral-CT using two different software tools for quantitative functional analysis and qualitative evaluation of motion changes in comparison with magnetic resonance imaging. Rofo 2004;176:1786–93. [20] Leung KC, Martin CJ. Effective doses for coronary angiography. Br J Radiol 1996;69:426–31. [21] Gilard M, Cornily JC, Mansourati J, et al. Estimation of radiation exposure with 16-detector row computed tomography of coronary artery. Eur Heart J 2004;25(Suppl.):472.
