Two-segment CTA reconstruction aids images, not accuracy

May 20, 2007

For patients with faster heart rates, the use of two-segment reconstruction improves image quality at 64-slice coronary CT angiography (CTA), according to a study from Germany. Among all patients, however, accuracy was not improved using two-segment when compared to single-segment reconstruction.

Multisegment reconstruction, using data scanned from more than one heart cycle, is a popular way to improve temporal resolution at 64-slice coronary CTA, noted study authors Dr. Christopher Herzog and colleagues from Johann Wolfgang Goethe University in Frankfurt, Germany, and Dr. Joseph Schoepf from the Medical University of South Carolina in Charleston.

At typical gantry rotation times of 350-330 msec, multisegment reconstruction yields temporal resolution of 82.5-43 msec at some heart rates, compared to 175-165 msec for single-segment reconstruction, they explained.

"However, the practical value of segmented reconstruction is controversial," the group wrote. "Previously published data on potential improvements in image quality at coronary CT angiography with the use of multi- versus single-segment reconstruction disagree. More important, to our knowledge, the effect of segmented reconstruction algorithms on the actual diagnostic accuracy for lesion detection has not been addressed to date" (Radiology, May 10, 2007).

The prospective study sought to evaluate the effect of single- versus two-segment image reconstruction on image quality and diagnostic accuracy at 64-slice CTA, using catheter coronary angiography as the reference standard.

The researchers performed both 64-slice multidetector CT coronary angiography and conventional angiography in 40 consecutive patients (22 men, 18 women; mean age 61 years ± 8). Patients with unstable vital signs or other contraindications were excluded.

The 32 patients with heart rates higher than 65 beats per minute (bpm) were given one to two intravenous injections of beta-blockers of 5 mg of metoprolol (Lopressor, Novartis, East Hanover, NJ) before the scan. Patients were also given 70-90 mL of the nonionic contrast agent iopamidol (370 mg/mL) (Isovue, Bracco, Milan) by power injector (Stellant D, Medrad, Indianola, PA) followed by a 50-mL saline chaser.

CT images were acquired on a 64-slice scanner (Sensation 64 Cardiac, Siemens Medical Solutions, Malvern, PA) at 120 kV and 900 mAs, using 64 x 0.6-mm collimation, z-flying focal spot, 0.33-second rotation time, and pitch of 0.2. ECG was measured simultaneously with imaging to enable reconstruction with retrospective ECG gating. Each dataset was reconstructed twice -- once using a single-segment and once with a two-segment adaptive cardiac volume reconstruction algorithm included with the scanner.

"At single-segment reconstruction, each single transverse section contained data from only one RR cycle ... which resulted in a temporal resolution equivalent to half of the rotation time in a centered region of interest (e.g., 165 msec for the 0.33-second rotation time)," the authors wrote. "A multidetector spiral interpolation between the projections of adjacent detector rows was used to compensate for table movement and to provide a well-defined section sensitivity profile for images without spiral artifacts."

All CTA findings were compared with coronary angiograms, obtained by using the Judkin's technique.

Agreement between the two readers for stenosis grading and image quality was high at 0.81 (95% CI: 0.74, 0.89) and 0.78 (95% CI: 0.75, 0.81), the team reported.

In all, 560 of 600 (93.3%) coronary segments were seen at single-segment reconstruction and 561 (93.5%) at two-segment reconstruction (p = 0.35), Herzog and colleagues reported. The mean image quality scores were not significantly different (p = 0.22) for single- (3.1 ± 0.9) versus two-segment (3.2 ± 0.8) reconstructions.

"However, the plotting of differences in image quality (i.e., quality scores for two-segment reconstruction minus scores for one-segment reconstruction) versus heart rate showed larger differences (F = 20.358, df = 16, 1175; p < 0.001) at two-segment reconstruction in six (15%) patients who presented with heart rates of 80-82 beats per minute," they wrote. "At these heart rates, the temporal resolution at two-segment reconstruction is at its optimum at approximately 83 msec, as opposed to 165 msec at single-segment reconstruction, a value that remains stable at all heart rates."

Analyzed per segment, significant coronary artery disease (greater than 50% stenosis), was detected with 77.1% sensitivity and 98.6% specificity using single-segment image reconstruction, and with 79.2% sensitivity and 99.1% specificity by using two-segment image reconstruction, the group reported. Per-patient sensitivity and specificity were 100% and 87.5%, respectively, using single-segment image reconstruction, and 100% and 95.8%, respectively, using two-segment image reconstruction.

The superior temporal resolution of two-segment image reconstruction produced better image quality at 64-slice CTA at higher heart rates, but did not improve overall diagnostic accuracy in a patient population with a wide variety of heart rates, the group wrote. Similarly high image quality and diagnostic accuracy were seen with both reconstruction methods.

The results were "fairly consistent" with previous reports, for example by Kachelriess et al, who showed that "two-segment reconstruction for heart rates greater than 70 beats per minute yielded better results than single-segment reconstruction when the table feed was restricted," the authors wrote.

At multidetector-row CT (MDCT), diagnostic accuracy may be affected by broadening of the time sensitivity profile due to inconsistencies of subsequent heart cycles, and heart motion patterns cannot be expected to follow the same motion pattern with every beat, resulting in "spatial inconsistencies and blurring when data for a single transverse image are sampled from several heartbeats," they noted.

On the other hand, data from several cardiac cycles with slightly different RR interval lengths will be derived from slightly different cardiac phases, resulting in typical stair-step artifacts along the z-axis.

"At multisegment reconstruction, this timing shift is averaged out during reconstruction of each transverse image, which reduces the stair-step artifacts on reformations but increases blurring within individual images," the authors wrote. "Theoretically, such spatial inconsistencies impair diagnostic accuracy for stenosis detection and grading. This, however, was not observed in our patient population, likely because of improved temporal resolution at 64-section CT as compared with previous scanner generations used in other studies."

Yet image quality was improved at heart rates of 80-82 bpm. "At these heart rates, the gain in temporal resolution thus may outweigh any impairment of image quality due to broadened time sensitivity profiles or dynamic definition of temporal resolution," they wrote.

The most important study limitation was sample size, which included only nine patients with heart rates of 78-87 bpm.

"Routine use of segmented reconstruction algorithms in patients with arrhythmia or tachycardia appears recommendable," the group concluded.

By Eric Barnes
AuntMinnie.com staff writer
May 21, 2007

Calcium scoring, CTA combine for better diagnosis of coronary artery disease, December 26, 2006

Dual-source coronary CTA images the calcium-burdened, May 13, 2007

Calcification percentage higher in stable carotid plaques, March 28, 2007