Analyzing left ventricular function is the prelude to treating several cardiac diseases, and it seems that multidetector CT may now be equivalent to MRI for the task over a wide range of heart rates, according to researchers from Kyoto University Graduate School of Medicine in Japan. But another study gives MRI a slight edge still.
Neither ECG-gated SPECT nor 2D echocardiography did as well as MRI, or the segmented eight-slice MDCT protocol that was the focus of the Japanese study. CT's main problem was its high radiation dose, which the authors said they hoped to reduce significantly when a scanner with more detectors became available and scan times shortened.
A number of modalities are used to evaluate global left ventricular function, especially ejection fraction, but the clear leader has been MRI, wrote authors Dr. Masaki Yamamuro, Dr. Eiji Tadamura, and colleagues in the February issue of Radiology (February 2005, Vol. 234:2, pp. 391-398). Recently MDCT with a temporal resolution of 125-250 msec has been shown to be promising when compared to cineventriculography or MRI.
"However, it was indicated that reconstructed images obtained in patients with a high heart rate were of low quality because of motion artifacts; thus, manual tracing of endocardial contours had limited accuracy," the authors wrote. Segmental reconstruction, using data from several heartbeats to improve temporal resolution, was used to effectively shorten temporal resolution and reduce motion artifacts.
First, phantom studies were performed to evaluate reconstruction artifacts resulting from two retrospective ECG-gated reconstruction methods, including 1) a half reconstruction algorithm, and 2) a segmental reconstruction algorithm using data from several heartbeats. The studies were performed using heart rates at every 5% of the cardiac cycle from 40 to 100 beats per minute, and 1-mm reconstruction increments. The phantom results showed various artifacts at heart rates of 65 beats per minute or higher when the half reconstruction algorithm was used, but no substantial artifacts with the segmental reconstruction estimate, except when heart rates exceeded 80 beats per minute, the authors wrote.
For the human studies, the group acquired MDCT images in 50 patients at 0.5-sec rotation time, and 8 x 1-mm detector collimation on an Aquilion Multi V1.10 JR 002 scanner (Toshiba Medical Systems, Tokyo) in 50 patients with clinical indications for cardiac function tests (28 men and 22 women, ages 46-84, mean age 67).
The estimated effective radiation dose was 7.4 mSv, and before scanning all patients received 100-150 mL of nonionic contrast material at 3.0-3.5 mL/sec, using an automatic triggering system that activated when density in the ascending aorta reached 150 HU.
A half and a segmental reconstruction algorithm were applied, and retrospective reconstruction was performed for acquisition of phase images from 0% to 95% of the RR interval at 5% increasing steps, and 20 heart phases were obtained.
Within 10 days of CT the same 50 patients underwent cardiac MRI on a Symphony 1.5-tesla whole-body imager (Siemens Medical Solutions, Erlangen, Germany) using multiple surface coils. Breath-hold cine MR images were acquired using repetition and echo times of 3.6 msec and 1.8 msec, respectively, with a flip angle of 70 degrees, 7-15 lines per segment, 208 x 256 matrix, and 340 x 340 field-of-view. Cine MR images of 10-12 contiguous sections with 10-mm section thickness were obtained in the short-axis plane from the base to the apex of the left ventricle to acquire 3D left ventricular (LV) data.
Both CT and MR were analyzed by an experienced observer who did not have access to clinical information. Forty-one patients also underwent 2D echocardiography using a Sonos 555 system (HP, Palo Alto, CA) using a 3.5-MHz transducer in the parasternal and apical planes, saved in the cine loop format. The images were evaluated by the same observer who was blinded to clinical information.
Finally, a total of 27 patients underwent exercise thallium 201 SPECT, by means of 74 MBq of 201TI injected intravenously at the peak exercise level. Initial SPECT was performed 10 minutes after exercise ended. Three to four hours later, following injection of an additional 37 MBq of 201T1, reinjection SPECT was performed. Images were obtained on a Millennium dual-head gamma camera (GE Healthcare, Waukesha, WI).
CT measurements for end-diastolic volume, end-systolic volume, ejection fraction, and LV mass were compared with MRI, which served as the reference standard. Estimated ejection fraction values for MDCT, echocardiography, and SPECT were also compared with MRI.
"Twenty patients had aortic and/or mitral valve disease, 12 had myocardial infarction, 12 had angina pectoris, two had infectious endocarditis, two had sarcoidosis, one had pericarditis, and one had dilated cardiomyopathy," the authors wrote.
The results showed that the CT-estimated ejection fraction was well correlated to MR estimates (bias ± standard deviation, -1.2% ± 4.6, r = 0.96). Similar high agreement was seen in end-diastolic volume (-0.35 mL ± 15.2, r = 0.97), end-systolic volume (1.1 mL ± 8.6, r = 0.99), and LV mass (2.5 mL ± 15.0, r = 0.96). The standard deviation of the ejection fraction difference between MDCT and MR was significantly less than that of echocardiography and MRI (p < 0.001) or the difference separating SPECT and MRI (p < 0.001).
Even the segmental reconstruction approach wasn't robust enough to prevent artifacts in the phantom experiments at heart rates above 80 beats per minute, the authors acknowledged, though no artifact was seen in any of the human studies. And more rapid rotation times of 0.4 sec that have been recently reported will make it possible to stabilize and shorten temporal resolution using the segmental method, they noted.
"Our results demonstrate that the left ventricular values, including (ejection fraction, end-diastolic values, end-systolic values) and left ventricular mass obtained with multidetector-row CT and a segmental reconstruction algorithm correlated and agreed with those obtained with MR over a wide range of heart rates," Yamamuro and colleagues concluded. Moreover, MDCT "with a segmental reconstruction algorithm was more accurate than two-dimensional echocardiography or ECG-gated SPECT."
Different results in Europe
It was Juergens et al from the University of Muenster in Germany who suggested that reconstruction algorithms use data from several heartbeats to optimize cardiac function, Yamamuro and colleagues wrote (American Journal of Roentgenology, December 2002, Vol. 179:6, pp. 1545-1550). Thus, the principal achievement of the Japanese team's work was in the demonstration that a segmental reconstruction algorithm was more appropriate for reducing motion artifacts, particularly at high heart rates, than a half reconstruction algorithm, the authors stated.
Meanwhile, the German team has also been keeping busy, and has since found that MDCT is improved with multisegmental MDCT image reconstruction, but still not as good as cine MRI.
In a study just published in European Radiology, Jeurgens's group also used a multisegmental cardiac image reconstruction algorithm and MDCT to evaluate temporal resolution and determine LV volumes and global LV function. The results were compared with cine MRI in 12 patients with known or suspected coronary artery disease (European Radiology, January 15, 2005, Vol. 15:1, pp. 111-117).
Their standard adaptive cardiac volume algorithm and multisegmental used data from three, five, and seven rotations. The results showed that mean temporal resolution was 192 ± 24 msec using the adaptive cardiac volume (ACV) algorithm, and significantly better using the three-, five-, and seven-rotation segments at 139 ± 12, 113 ± 13, and 96 ± 11 msec (p < 0.001 for each), respectively. No significant differences were found in mean end-diastolic volume using the ACV algorithm, the multisegmental approach, and cine MRI.
However, the improved temporal resolution with multisegmental image reconstruction end-systolic volume estimates was still less accurate with MDCT compared to cine MRI (mean differences 3.9 ± 11.8 mL to 8.1 ± 13.9 mL). And a Bland-Altman analysis showed a consistent underestimation of the ejection fraction by 2.3% to 5.4% compared to MRI, the authors wrote in their abstract.
"Multisegment image reconstruction improves temporal resolution compared to the standard ACV algorithm, but this does not result in a benefit for determination of LV volume and function," Jeurgen and colleagues concluded.
By Eric Barnes
AuntMinnie.com staff writer
February 15, 2005
Related Reading
Cardiac MR center finds success with 'one-stop shopping' model, January 25, 2005
Left ventricular hypertrophy accompanies obesity in young, healthy women, April 23, 2004
Stress MRI tops echo for coronary disease detection, May 26, 2004
MR tagging identifies LV torsion-to-shortening ratio in AS, March 4, 2004
Coronary MS CTA quantifies LV function adequately, March 23, 2004
Copyright © 2005 AuntMinnie.com
