Myocardial perfusion SPECT reconstruction: Receiver operating characteristic comparison of CAD detection accuracy of filtered backprojection reconstruction with all of the clinical imaging information available to readers and solely stress slices iteratively reconstructed with combined compensation

BACKGROUND: Past receiver operating characteristic (ROC) studies have demonstrated that single photon emission computed tomography (SPECT) perfusion imaging by use of iterative reconstruction with combined compensation for attenuation, scatter, and detector response leads to higher area under the ROC curve (A(z)) values for detection of coronary artery disease (CAD) in comparison to the use of filtered backprojection (FBP) with no compensations. A new ROC study was conducted to investigate whether this improvement still holds for iterative reconstruction when observers have available all of the imaging information normally presented to clinical interpreters when reading FBP SPECT perfusion slices. METHODS AND RESULTS: A total of 87 patient studies including 50 patients referred for angiography and 37 patients with a lower than 5% likelihood for CAD were included in the ROC study. The images from the two methods were read by 4 cardiology fellows and 3 attending nuclear cardiologists. Presented for the FBP readings were the short-axis, horizontal long-axis, and vertical long-axis slices for both the stress and rest images; cine images of both the stress and rest projection data; cine images of selected cardiac-gated slices; the CEQUAL-generated stress and rest polar maps; and an indication of patient gender. This was compared with reading solely the iterative reconstructed stress slices with combined compensation for attenuation, scatter, and resolution. With A(z) as the criterion, a 2-way analysis of variance showed a significant improvement in detection accuracy for CAD for the 7 observers (P = .018) for iterative reconstruction with combined compensation (A(z) of 0.895 +/- 0.016) over FBP even with the additional imaging information provided to the observers when scoring the FBP slices (A(z) of 0.869 +/- 0.030). When the groups of 3 attending physicians or 4 cardiology fellows were compared separately, the iterative technique was not statistically significantly better; however, the A(z) for each of the 7 observers individually was larger for iterative reconstruction than for FBP. Compared with results from our previous studies, the additional imaging information did increase the diagnostic accuracy of FBP for CAD but not enough to undo the statistically significantly higher diagnostic accuracy of iterative reconstruction with combined compensation. CONCLUSIONS: We have determined through an ROC investigation that included two classes of observers (experienced attending physicians and cardiology fellows in training) that iterative reconstruction with combined compensation provides statistically significantly better detection accuracy (larger A(z)) for CAD than FBP reconstructions even when the FBP studies were read with all of the extra clinical nuclear imaging information normally available.     

Three-dimensional fetal echocardiography: gated versus nongated techniques.

The purpose of this study was to compare gated with nongated three-dimensional fetal echocardiography in terms of the ability to demonstrate fetal cardiac anatomy. We examined nine fetuses in utero using conventional two-dimensional sonographic imaging equipment, an electromagnetic position sensor, and a computer-graphics workstation. Free-hand sweeps were performed through the fetal heart and great vessels in either transverse or sagittal orientations with respect to the fetal heart. Seven transverse and five sagittal sweeps were selected for reconstruction and analysis. Cardiac gating was performed by using a temporal Fourier transform to determine the fundamental frequency of cardiac motion. Two-dimensional data from each sweep were reprojected to a series of volume data sets. Each series was then condensed to a single volume, so that each two-dimensional sweep could be compared with its respective gated and nongated volume data sets. The two-dimensional data were reviewed utilizing a display with forward and backward cineloop capability. The gated and nongated volume data sets were displayed interactively as a series of three orthogonal planes, with the ability of the observer to control the location of each image plane within the volume. The gated data were animated with variable display frame rates. Conventional two-dimensional imaging provided a fairly complete evaluation of the fetal heart when scanning included the four-chamber view with a sweep across the outflow tracts. Nongated three-dimensional fetal echocardiography allowed visualization of some structures and views not demonstrated with two-dimensional ultrasonography. Gated three-dimensional fetal echocardiography provided significantly better visualization and comprehension of cardiac anatomy than nongated three-dimensional fetal echocardiography. The superiority of gated over nongated three-dimensional fetal echocardiography appears to come from both improved image quality and the anatomic clues that derive from the ability to view cardiac motion.     

Simultaneous assessment of cardiac perfusion and function using 5-dimensional imaging with Tc-99m teboroxime

BACKGROUND: Dynamic single photon emission computed tomography (SPECT) acquisition and reconstruction of early poststress technetium 99m teboroxime washout images has been shown to be useful in the detection of coronary disease. Assessment of poststress regional wall motion may offer additional use in assessing coronary disease. Our goal was to investigate the feasibility of simultaneously imaging myocardial ischemia and transient poststress akinesis using gated-dynamic SPECT. METHODS AND RESULTS: A gated-dynamic mathematical cardiac torso (MCAT) phantom was developed to model both teboroxime kinetics and cardiac regional wall motion. A lesion was simulated as having delayed poststress teboroxime washout together with a transient poststress wall motion abnormality. Gated projection data were created to represent a 3-headed SPECT system undergoing a total rotation of 480 degrees . The dynamic expectation-maximization reconstruction algorithm with postsmoothing across gating intervals by Wiener filtering, and the ordered-subset expectation maximization reconstruction algorithm with 3-point smoothing across gating intervals were compared. Compared with the ordered-subset expectation maximization with 3-point smoothing, the dynamic expectation-maximization algorithm with Wiener filtering was able to produce visually higher-quality images and more accurate left ventricular ejection fraction estimates. CONCLUSION: From simulations, we conclude that changing cardiac function and tracer localization possibly can be assessed by using a gated-dynamic acquisition protocol combined with a 5-dimensional reconstruction strategy.