Chuanyong Bai

Development and evaluation of a new fully automatic motion detection and correction technique in cardiac SPECT imaging

BackgroundIn cardiac SPECT perfusion imaging, motion correction of the data is critical to the minimization of motion introduced artifacts in the reconstructed images. Software-based (data-driven) motion correction techniques are the most convenient and economical approaches to fulfill this purpose. However, the accuracy is significantly affected by how the data complexities, such as activity overlap, non-uniform tissue attenuation, and noise are handled.MethodsWe developed STASYS, a new, fully automatic technique, for motion detection and correction in cardiac SPECT. We evaluated the performance of STASYS by comparing its effectiveness of motion correcting patient studies with the current industry standard software (Cedars-Sinai MoCo) through blind readings by two readers independently.ResultsFor 204 patient studies from multiple clinical sites, the first reader identified (1) 69 studies with medium to large axial motion, of which STASYS perfectly or significantly corrected 86.9% and MoCo 72.5%; and (2) 20 studies with medium to large lateral motion, of which STASYS perfectly or significantly corrected 80.0% and MoCo 60.0%. The second reader identified (1) 84 studies with medium to large axial motion, of which STASYS perfectly or significantly corrected 82.2% and MoCo 76.2%; and (2) 34 studies with medium to large lateral motion, of which STASYS perfectly or significantly corrected 58.9% and MoCo 50.0%.ConclusionsWe developed a fully automatic software-based motion correction technique, STASYS, for cardiac SPECT. Clinical studies showed that STASYS was effective and corrected a larger percent of cardiac SPECT studies than the current industrial standard software.

CsI(T1)/PIN solid state detectors for combined high resolution SPECT and CT imaging

We have developed a CsI(Tl)/PIN detector module for high resolution SPECT and low dose photon-counting CT imaging. Using the detector modules with 6.1mm pixels, we built a cardiac SPECT system with three detector heads. The detector heads form a triple-head (each 20×15cm), geometry for emission scans and reconfigure to form a large transaxial field-of-view (FOV) geometry for transmission scans using an x-ray based transmission source. Anthropomorphic phantom and patient data was used to evaluate the performance of the new cardiac SPECT camera. We then developed modules with 2.8mm pixel size to further improve the spatial resolution, improving energy resolution to 7.5%. Improved energy resolution was achieved by improving the crystal arrays, electronics and packaging of the module. Using the 2.8mm modules, we built a large FOV planar imager (39×31cm), and evaluated its emission performance with clinical studies and its CT performance with an anthropomorphic phantom. The count rate capability was 20kcps per detector pixel. The combined SPECT/CT scan was completed in 5 min, resulting in high quality and accuracy attenuation maps (μH20@140keV was 0.151+/−0.003/cm at ∼5μSv CT dose). The large FOV imager showed excellent clarity and high resolution for emission bone studies and demonstrated reconstructed CT image quality suitable for attenuation correction or image fusion and localization (∼150μSv CT dose).

A triple-head solid state camera for cardiac single photon emission tomography (SPECT)

The CardiusTM-3 (C-3) camera is a triple-head small field-of-view camera dedicated to cardiac SPECT imaging. It is built on Digirads solid-state detector technology with 6mm x 6mm CsI:Tl crystals. The system demonstrates upright imaging with the patient rotating in a SPECTourTM chair while the gantry keeps stationary during data acquisition. A region of interest (ROI) tool on the persistence scope (p-scope) is used to position the heart at the center of rotation to avoid cardiac-truncation. The tool also provides the count rate in the ROI so the users can determine the acquisition time for different patients to meet the American Society of Nuclear Cardiology (ASNC) guidelines. The intrinsic energy resolution, reconstructed spatial resolution with scatter and NEMA extrinsic planar sensitivity of the solid-state detector were measured and results are reported. C-3 can acquire high dose tomographic studies in 7 minutes at 20 seconds/projection (based on actual patient study). This short acquisition time (compared with conventional Anger style dual and single head systems) not only improves the patient comfort but also reduces patient motion, which in turn improves the image quality. Anthropomorphic phantom and patient studies performed in this work showed that C-3 image quality and diagnostic outcome were equivalent to those from a dual head camera, but the acquisition time could be reduced by 38%. The reduced acquisition time (compared with conventional Anger style dual and single head systems) not only improves the patient comfort but also reduces patient motion, which in turn can improve the image quality.

An operator-passive thoracic impedance approach for Respiratory motion gating in myocardial perfusion SPECT

Respiratory motion gating (R-Gating) in SPECT myocardial perfusion imaging (MPI) usually requires external devices and extra work of the technologist for patient setup. In this paper, we developed an operator-passive R-Gating technique without external devices. Method: Using the thoracic impedance approach, the respiratory motion signal was obtained simultaneously with the ECG signal from the same ECG leads. With list-mode acquisition, the respiratory motion signal was tagged to the SPECT data and was used to rebin the list-mode data into R-Gating data. Forty consecutive patients were scanned to evaluate and establish the R-Gating approach. Both the conventional data and the list-mode data were saved. Then forty eight consecutive patients were scanned for respiratory motion assessment using the established approach. Results: The summed R-Gating images were different from the summed ECG-gating images in only three of the forty patients (7.5%) and equivalent in the rest of the patients (92.5%). R-Gating with six respiratory phases showed optimal motion detection and image quality compromise. The R-Gating images illustrated different types of motion of the heart during respiratory cycles, including 3-D translation, apparent rotation of the heart along the long axis, and pivoting at the base. The observed motion of the heart in R-Gating images correlated well with the detected respiratory motion signal. For the motion assessment studies, 28/48 (58.3%) showed no motion or small motion, 20/48 (41.7%) of stress and 18/48 (37.3%) of rest studies showed medium motion, and 2/48 (4.2%) of rest studies showed large motion. In conclusion, we developed an effective operator-passive approach for respiratory motion gating for SPECT MPI.

Evaluation of a cardiac SPECT system using a common set of solid-state detectors for both emission and transmission scans and a ultras-low dose lead X-ray transmission line source

We developed a new cardiac SPECT system (X-ACT) with a low dose volume CT transmission-based attenuation correction (AC). Three solid-state detectors are configured to form a triple-head system for emission scans and reconfigured to form a large 27” (69 cm) field-of-view (FOV) detector arc for transmission scans. The transmission line source is the collimated fluorescence X-ray emitted from a lead target when the lead target is fluxed by a narrow X-ray beam from an X-ray tube. High quality transmission scans can be completed in as short as one minute with insignificant patient dose (~5 μSv). For evaluation, we scanned an anthropomorphic phantom with a uniform cardiac insert and the same anthropomorphic phantom with one 60º full defect and one 45º 50% defect in the cardiac insert. We also scanned an ACR phantom and a uniform cylindrical phantom. Results showed that image uniformity and defect contrast were improved by AC in the anthropomorphic phantom studies as compared to without AC (NAC). The inferior to anterior wall ratio and the septal to lateral wall ratio were 0.99 and 1.16 before and 1.02 and 1.00 after AC. The defect contrast of the full and 50% defect was 0.528 and 0.156 before and 0.628 and 0.173 after AC. The ACR phantom images with AC showed correction of the bowing effect due to attenuation in the NAC images. The reconstructed attenuation coefficient of water at 140 keV was 0.154±0.003/cm and 0.157±0.002/cm in the liver and cardiac regions of the anthropomorphic phantom, and 0.150±0.003/cm in the uniform region of the ACR phantom and the uniform cylindrical phantom. In conclusion, the X-ACT system generated accurate attenuation maps with one-minute transmission scans. Phantom studies showed that AC improved image quality and quantification over NAC.

Image characterization metrics for muon tomography

Muon tomography uses naturally occurring cosmic rays to detect nuclear threats in containers. Currently there are no systematic image characterization metrics for muon tomography. We propose a set of image characterization methods to quantify the imaging performance of muon tomography. These methods include tests of spatial resolution, uniformity, contrast, signal to noise ratio (SNR) and vertical smearing. Simulated phantom data and analysis methods were developed to evaluate metric applicability. Spatial resolution was determined as the FWHM of the point spread functions in X, Y and Z axis for 2.5cm tungsten cubes. Uniformity was measured by drawing a volume of interest (VOI) within a large water phantom and defined as the standard deviation of voxel values divided by the mean voxel value. Contrast was defined as the peak signals of a set of tungsten cubes divided by the mean voxel value of the water background. SNR was defined as the peak signals of cubes divided by the standard deviation (noise) of the water background. Vertical smearing, i.e. vertical thickness blurring along the zenith axis for a set of 2 cm thick tungsten plates, was defined as the FWHM of vertical spread function for the plate. These image metrics provided a useful tool to quantify the basic imaging properties for muon tomography.

New Daily Detector Uniformity Quality Control Methodology for Cardiac SPECT Using Solid-State Detectors

Detector non-uniformity can potentially introduce detectable artifacts into SPECT images. The degree of the non-uniformity and the extent and position of the non-uniform area on the detector surface determine the position and severity of the introduced artifacts. The commonly used daily uniformity quality control (QC) procedure follows the NEMA methodology but acquires fewer counts than the latter specifies. It has three major drawbacks: (1) it does not report the locations and extent of the non-uniform areas on the detector surface; (2) it may report a non-uniformity value that is lower than the true value due to the use of a 9-point filter, and it makes the reported non-uniformity value vary with the extent of the non-uniform area. These two drawbacks are inherited from the NEMA methodology. The third drawback is that the noise due to the relatively low counts collected in daily uniformity QC does not allow the measurement of certain degrees of non-uniformity with adequate statistical significance, yet such non-uniformity can potentially introduce observable artifacts. In this work we propose a new methodology for daily uniformity QC for cardiac SPECT imaging using solid-state detectors. The new QC procedure (1) reports the locations and extents of the non-uniform areas of the detectors and (2) can catch some detectors that pass the NEMA-based daily uniformity QC but are non-uniform enough to introduce detectable artifacts.

A new technique to systematically minimize misregistration introduced errors in cardiac perfusion studies with attenuation correction

In cardiac perfusion studies (both PET and SPECT) with CT-based attenuation correction, intensive efforts have been made to minimize the emission/transmission misregistration in order to minimize the errors introduced by the misregistration in the attenuation corrected images. In this work, we propose a new technique to systematically minimize the errors introduced by the misregistration. In general, the emission/transmission misregistration is a combination of two types of misregistration: (A) the myocardial wall is moved toward the mediastinum or diaphragm, the attenuation coefficient of which is approximately the same as that of the myocardium; and (B) the myocardial wall is moved into the lungs, the attenuation coefficient of which is significantly lower than that of the myocardium. In the proposed technique, after the emission/transmission registration and prior to the reconstruction with attenuation correction (AC), we first segment the heart in the emission image; then set the voxels of the transmission image corresponding to the segmented heart the attenuation coefficient of soft tissue. This technique is expected to have negligible effect on the accuracy of AC when type (A) misregistration exists, but systematically reduce the errors introduced by type (B) misregistration. A series of SPECT studies with attenuation correction were simulated to evaluate this technique using the mathematical MCAT phantom. Results showed that perfusion errors were significantly reduced for type (B) misregistration but not for type (A) misregistration of magnitude of 1.2 cm. In conclusion, we developed a new technique that systematically minimized the errors introduced by emission/transmission misregistration for cardiac perfusion studies with attenuation correction.

Handling of Bad Pixels on Pixelated Solid State Detectors

Bad pixel correction on pixelated solid-state detectors typically uses the average of the direct neighboring pixels (AVG) to derive the value of a bad pixel. However, the AVG approach was suboptimal for high resolution imaging. Therefore, we developed a least gradient approach (LGA) in this work. In the LGA approach, the gradients of the image in a 5 × 5 box centered at the bad pixel were calculated along the two orthogonal and two diagonal directions. The value of the bad pixel was derived from the average of the two neighboring pixels along the direction in which the gradient was the least. For 18 cardiac SPECT studies, we added to the data randomly generated bad pixels and bad pixels in a specially designed pattern and then corrected the bad pixels using the AVG approach. Images reconstructed from the bad-pixel-free data and the bad-pixel-corrected data were compared. For high resolution imaging, we used line and bar phantom studies to evaluate the AVG and LGA approaches on a pixelated solid-state gamma camera. Patient studies showed no visible qualitative or significant quantitative difference between the images reconstructed from the bad-pixel-free and bad-pixel-corrected data. The maximum segment change ranged from 0% to 7.4% with average of 3.6 for data with randomly generated bad pixels. Blind reading of the images by an expert nuclear cardiologist showed no diagnostic difference for any of the patients. The line phantom studies showed two bad pixels not corrected by the AVG approach but corrected by the LGA approach. Bar phantom studies showed ten bad pixels not corrected by the AVG approach. But 9 out the 10 bad pixels were corrected using the LGA approach. The commonly used averaging approach (AVG) was effective for cardiac SPECT imaging but the least gradient approach (LGA) developed in this work was more effective for high resolution imaging.

Using myocardium-to-background ratio to get the optimal starting angle for a non-360-degree upright cardiac SPECT acquisition

Background: Myocardium counts in the projection data and myocardium wall ratios in the reconstructed images have been used to find the optimal angular sampling for 180° supine SPECT imaging. However, supine and upright images demonstrate different soft tissue attenuation patterns. In this work, we propose the use of myocardium-to-background ratio (MBR) in the projection data as a figure-of-merit to determine the optimal starting angle for upright cardiac SPECT imaging. We hypothesize that higher MBR in the projection data means better signal-to-noise ratio and leads to better quality reconstructed images. Methods: We performed MBR analysis for 76 patient images (38 stress-rest studies) acquired with upright SPECT camera. Out of the 76, 22 were acquired with starting angle of right anterior oblique (RAO) 45° in 180°, 26 with RAO 30° in 180° and 22 with RAO 38° in 202°. We separated the male (34) and female (42) data and compared the MBR at each projection angle. For the MBR analysis, the myocardium wall and the background in each projection were identified in four steps: (1) a volume image was reconstructed using FBP algorithm, (2) left ventricle was segmented from the volume image, (3) the segmented image was reprojected to obtain the region of interest (ROI) to calculate the average counts in the myocardium, and (4) the ROI in step (3) was blurred then subtracted by the ROI in step (3) to derive the ROI to calculate background counts. Results: MBR reached a local peak at RAO 30° for all the data. For males, the MBR at RAO 45° and RAO 38° was lower than RAO 30° but higher than left posterior oblique (LPO) 60° (towards the end of the acquisition). For females, the MBR at RAO 38° and RAO 45° was lower than RAO 30° and LPO 60°. Conclusions: Using MBR as the figure of merit, we conclude that for males, starting angle of RAO 38° or RAO 45° is preferred over RAO 30° for upright cardiac SPECT imaging. For females, however, starting angle of RAO 30° is preferred.