Taek Soo Lee

Generation and evaluation of a simultaneous cardiac and respiratory gated Rb-82 PET simulation

The goal is to generate and evaluate a simulated 4D Rb-82 PET dataset that realistically models simultaneous respiratory and cardiac motions for use to study the effects of the motions and their compensation using various gating schemes. Normal cardiac and respiratory (C&R) motions were simulated separately using the realistic 4D XCAT phantoms. The C&R motion cycles were divided into 24 and 48 equally-spaced time points, respectively. The simultaneous dual motions were modeled by 24×48 phantoms with different combinations of C&R motion phases. Almost noise-free projections of the heart, blood pool, lungs, liver, stomach, spleen, and the remaining body were simulated separately using the combined SimSET and GATE Monte Carlo simulation program which is 12 times faster than GATE alone. The projections were scaled and combined to simulate a typical Rb-82 myocardial perfusion (MP) PET patient study. The no gating, 6-frame respiratory gating only, 8-frame cardiac gating only, and simultaneous 6-frame respiratory and 8-frame cardiac gating schemes were applied. Each gated projection dataset was reconstructed using a 2D OS-EM without and with attenuation correction (AC) using an averaged and gated attenuation maps. The reconstructed images were evaluated in terms of artifactual non-uniformity in the MP polar map. Significant artifactual non-uniformity was found in the MP polar map over all gating scheme without AC. With AC, the artifactual decreases in both the anterior and inferior regions were reduced with respiratory gating. Cardiac motion alone did not cause significant artifactual non-uniformity. In addition, the combination of dual gating and AC using the gated attenuation map provided the most uniform MP polar map. We demonstrated the flexibility and utility of the 4D XCAT phantom set with simultaneous C&R motions. It is a powerful tool to study motion effects on MP PET studies and to evaluate C&R gating schemes, AC and quantitative 4D PET image reconstruction methods.

Realistic simulation of regional myocardial perfusion defects for cardiac SPECT studies

The current 3D XCAT phantom allows users to manually define the regional myocardial perfusion defect (MPD) as a simple pie-shaped wedge region with reduced activity level in the myocardium of left ventricle. To more accurately and realistically model the MPD, we have developed a new regional MPD model for the 3D XCAT phantom for myocardial perfusion SPECT (MP-SPECT) studies based on the location and the severity of the stenosis in a computer generated coronary arterial tree. First, we generated a detailed coronary arterial tree by extending the large proximal branches segmented from patient CT images to cover the whole heart using an iterative rule-based algorithm. Second, we determined the affected downstream vascular segments of a given stenosis. Third, we computed the activity of each myocardial region as a function of the inverse-distance-weighted average of the flow of the neighboring vascular segments. Fourth, we generated a series of bulls-eye maps of MP-SPECT images of different coronary artery stenosis scenarios. Fifth, we had expert physician readers to qualitatively assess the bulls-eye maps based on their similarity to typical clinical cases in terms of the shape, the extent, and the severity of the MPDs. Their input was used to iteratively revise the coronary artery tree model so that the MPDs were closely matched to those found in bulls-eye maps from patient studies. Finally, from our simulated MP-SPECT images, we observed that (1) the locations of the MPDs caused by stenoses at different main arteries were different largely according to their vascular territories, (2) a stenosis at a proximal branch produced a larger MPD than the one at a distal branch, and (3) a more severe stenosis produced a larger MPD than the less severe one. These observations were consistent to those found in clinical cases. Therefore, this new regional MPD model has enhanced the generation of realistic pathological MP-SPECT images using the XCAT phantom. When combining with the mechanical model of the myocardium, the new model can be extended for the simulation of 4D gated MP-SPECT simulation of a pathological heart with both perfusion and motion defects.

Comparison of 3D OS-EM and 4D MAP-RBI-EM Reconstruction Algorithms for Cardiac Motion Abnormality Classification Using a Motion Observer

Using a heart motion observer, we compared the performance of two image reconstruction techniques, a 3D OS-EM algorithm with post Butterworth spatial filtering and a 4D MAP-RBI-EM algorithm. The task was to classify gated myocardial perfusion (GMP) SPECT images of beating hearts with or without regional motion abnormalities. Noise-free simulated GMP SPECT projection data was generated from two 4D NCAT beating heart phantom models, one with normal motion and the other with a 50% motion defect in a pie-shaped wedge region-of-interest (ROI) in the anterior-lateral left ventricular wall. The projection data were scaled to the clinical GMP SPECT count level before Poisson noise was simulated to generate 40 noise realizations. The noise-free and noisy projection data were reconstructed using the two reconstruction algorithms, parameters chosen to optimize the tradeoff between image bias and noise. As a motion observer, a 3D motion estimation method previously developed was applied to estimate the radial motion on the ROI from two adjacent gates. The receiver operating characteristic (ROC) curves were computed for radial motion magnitudes corresponding to each reconstruction technique. The area under the ROC curve (AUC) was calculated as an index for classification of regional motion. The reconstructed images with better bias and noise tradeoff were found to offer better classification for hearts with or without regional motion defects. The 3D cardiac motion estimation algorithm, serving as a heart motion observer, was better able to distinguish the abnormal from the normal regional motion in GMP SPECT images obtained from the 4D MAP-RBI-EM algorithm than from the 3D OS-EM algorithm with post Butterworth spatial filtering.

The Direct Incorporation of Perfusion Defect Information to Define Ischemia and Infarction in a Finite Element Model of the Left Ventricle

This paper describes the process in which complex lesion geometries (specified by computer generated perfusion defects) are incorporated in the description of nonlinear finite element (FE) mechanical models used for specifying the motion of the left ventricle (LV) in the 4D extended cardiac torso (XCAT) phantom to simulate gated cardiac image data. An image interrogation process was developed to define the elements in the LV mesh as ischemic or infarcted based upon the values of sampled intensity levels of the perfusion maps. The intensity values were determined for each of the interior integration points of every element of the FE mesh. The average element intensity levels were then determined. The elements with average intensity values below a user-controlled threshold were defined as ischemic or infarcted depending upon the model being defined. For the infarction model cases, the thresholding and interrogation process were repeated in order to define a border zone (BZ) surrounding the infarction. This methodology was evaluated using perfusion maps created by the perfusion cardiac-torso (PCAT) phantom an extension of the 4D XCAT phantom. The PCAT was used to create 3D perfusion maps representing 90% occlusions at four locations (left anterior descending (LAD) segments 6 and 9, left circumflex (LCX) segment 11, right coronary artery (RCA) segment 1) in the coronary tree. The volumes and shapes of the defects defined in the FE mechanical models were compared with perfusion maps produced by the PCAT. The models were incorporated into the XCAT phantom. The ischemia models had reduced stroke volume (SV) by 18-59 ml. and ejection fraction (EF) values by 14-50% points compared to the normal models. The infarction models, had less reductions in SV and EF, 17-54 ml. and 14-45% points, respectively. The volumes of the ischemic/infarcted regions of the models were nearly identical to those volumes obtained from the perfusion images and were highly correlated (R² = 0.99).

Task-based evaluation of motion compensated reconstructed images using 4D channelized Hotelling observer in dual gated SPECT

We developed a 4D SPECT image reconstruction method with respiratory motion (RM) and cardiac motion (CM) compensation, and evaluated its performance using a previously developed 4D channelized Hotelling observer (CHO) model for the task based evaluation of 4D medical images. A series of 4D XCAT (eXtended CArdiac Torso) phantoms with and without a regional CM abnormality was generated that divided a respiratory cycle into 24 frames and a cardiac cycle into 48 frames for each respiratory phase. Almost noise-free projection data were generated from the 1152 3D XCAT phantoms at each respiratory and cardiac (R&C) gated time frame using the SimSET simulation modeling a typical 99mTc Sestamibi gated myocardial perfusion (GMP) SPECT study. They were scaled and combined to form 6 equal-amplitude respiratory gates and 8 equal-time cardiac gates. Poisson noise was then added before image reconstruction using the 3D OS-EM with or without RM and CM compensation. Using a group-wise B-spline non-rigid image-based registration method, the deformation field (DF) of the RM were estimated and applied to each cardiac phase of the R&C gated SPECT images for RM correction. Then, the RM compensated GMP SPECT images were transformed using the estimated DF of the CM. The 2D reconstructed image slices were extracted and reorganized into a set of cine images and the space-time channels of the 4D CHO were applied to produce the space-time feature vectors to which the receiver operating characteristic (ROC) was applied and areas under the ROC curve (AUC) values calculated. The result demonstrated that the RM and CM compensated images showed drastically reduced noise level without blurring and its AUC values were significantly higher than those without compensation. We conclude that the RM and CM compensation allow significant reduction of image blurring and noise level in dual gated SPECT images resulting in significant improvement in detecting a regional CM abnormality using a 4D CHO model.

Evaluation of a 4D MAP-RBI-EM Image Reconstruction Method for Gated Myocardial SPECT using a Human Observer Study

Previously, we investigated a 4D maximum a posteriori rescaled-block iterative (MAP-RBI)-EM image reconstruction method with corrections of image degrading factors for gated myocardial SPECT. It provided a significantly improved trade-off between normalized mean squared error (NMSE) and normalized standard deviation (NSD) of the reconstructed images as compared to conventional reconstruction methods allowing more timing gates per cardiac cycle for better detection of wall motion abnormalities. In this study, we investigate the use of human observer study to evaluate the 4D MAP-RBI-EM method. We used a population of realistic 4D NURBS-based Cardiac-Torso (NCAT) phantoms modeling variations in cardiac motion. Half the population was normal; the other half had hypokinetic cardiac motion abnormalities. Noise-free and noisy projection data with 16 cardiac gates were generated using an analytical projector that included the effects of attenuation, collimator-detector response and scatter (ADS). The projection data were reconstructed using the 3D FBP and 3D OS-EM methods with corrections for ADS followed by a linear filter and the 4D MAP-RBI-EM method with ADS corrections. The reconstructed images were used in a human observer study. The observers were trained to the simulated gated SPECT images animated with a realistic real-time frame rate and were instructed to rate their confidence on the absence or presence of a motion defect on a continuous scale from 1 to 5. We applied receiver operating characteristic (ROC) analysis and used the area under the ROC curve as an index of comparison. The result showed significant differences in detection performance among the different NMSE-NSD combinations. Images obtained from the optimized 4D MAP-RBI-EM with corrections gave better human observer detection performance among the other image reconstruction methods. We conclude that the optimized 4D MAP-RBI-EM method with corrections of image degrading factors provides improvement in detecting wall motion abnormalities in gated myocardial SPECT.

Development of 4D mathematical observer models for the task-based evaluation of gated myocardial perfusion SPECT.

This paper presents two 4D mathematical observer models for the detection of motion defects in 4D gated medical images. Their performance was compared with results from human observers in detecting a regional motion abnormality in simulated 4D gated myocardial perfusion (MP) SPECT images. The first 4D mathematical observer model extends the conventional channelized Hotelling observer (CHO) based on a set of 2D spatial channels and the second is a proposed model that uses a set of 4D space-time channels. Simulated projection data were generated using the 4D NURBS-based cardiac-torso (NCAT) phantom with 16 gates/cardiac cycle. The activity distribution modelled uptake of (99m)Tc MIBI with normal perfusion and a regional wall motion defect. An analytical projector was used in the simulation and the filtered backprojection (FBP) algorithm was used in image reconstruction followed by spatial and temporal low-pass filtering with various cut-off frequencies. Then, we extracted 2D image slices from each time frame and reorganized them into a set of cine images. For the first model, we applied 2D spatial channels to the cine images and generated a set of feature vectors that were stacked for the images from different slices of the heart. The process was repeated for each of the 1,024 noise realizations, and CHO and receiver operating characteristics (ROC) analysis methodologies were applied to the ensemble of the feature vectors to compute areas under the ROC curves (AUCs). For the second model, a set of 4D space-time channels was developed and applied to the sets of cine images to produce space-time feature vectors to which the CHO methodology was applied. The AUC values of the second model showed better agreement (Spearmans rank correlation (SRC) coefficient = 0.8) to human observer results than those from the first model (SRC coefficient = 0.4). The agreement with human observers indicates the proposed 4D mathematical observer model provides a good predictor of the performance of human observers in detecting regional motion defects in 4D gated MP SPECT images. The result supports the use of the observer model in the optimization and evaluation of 4D image reconstruction and compensation methods for improving the detection of motion abnormalities in 4D gated MP SPECT images.

A simulation study of the effect of phase-shift on dual gated myocardial perfusion ECT

We evaluated the effect of relative phase-shift of cardiac and respiratory (C&R) motions in myocardial perfusion (MP) ECT. Previously, we generated a set of realistic 3D XCAT (eXtended CArdiac Torso) phantoms that model simultaneous C&R motions for use in the study of new data acquisition methods and corrective image reconstruction techniques for improved gated MP ECT, including PET and SPECT. The respiratory motion (RM) over a respiration cycle was modeled using 24 equally-spaced time frames while the cardiac beating motion (CBM) over a cardiac cycle was divided into 48 equally-spaced time frames for each of the 24 RM phases. Almost noise-free projection datasets were generated separately from the heart, blood pool, lungs, liver, kidneys, stomach, gall bladder and remaining body at each of the 24×48 time points using Monte Carlo simulation techniques that include the effect of collimator detector response, photon attenuation and scatter. To demonstrate the effect of relative phase-shift, a typical 99mTc Sestamibi MP SPECT projection dataset were generated. They were then scaled and combined to model different degrees of relative C&R phase shifting and grouped into 6 respiratory-gates with 8 cardiac-gates. Each projection was reconstructed using a 3D OS-EM without and with attenuation correction using an averaged and phase-mismatched gated attenuation maps. The image artifacts of the reconstructed images were compared by visual inspection of the MP polar maps. The results showed significant changes of artifactual non-uniformity in the polar maps for off-phase of RM compared to those of CBM. The changes in the polar maps also demonstrated the effect of phase shifting accordingly. We conclude that the 4D XCAT phantom dataset with simultaneous C&R motions provides a powerful tool in the study of the effects of C&R motions with relative phase shifts, and development of C&R gating schemes and motion correction methods for improved ECT/CT imaging.

Initial evaluation of a state-of-the-art commercial preclinical PET/CT scanner

We performed an initial evaluation of a state-of-the-art commercial preclinical PET/CT scanner (SuperArgus 4R, SEDECAL, Madrid, Spain). The PET unit consists of 4 rings of 96 detector modules each with an array of 338 1.45 × 1.45 × 15 mm3 pixelated LYSO and GSO phoswich crystals with DOI information. It has a maximum axial FOV of 100 mm and transaxial FOV of 120 mm. The CT unit consists of an x-ray source with variable micro focal spot size and a large 229 × 145 mm2 flat-panel detector that allows imaging of a volume-of-view (VOV) at three different magnifications, at a smallest voxel of 15 micron. We measured the sensitivity and uniformity of the PET unit using a calibrated Na-22 point source and a cylindrical phantom filled with a homogeneous FDG solution. The system resolution was determined from the reconstructed images of a thin FDG-filled capillary tube in air and inside a plastic cylinder, and a hot-rod phantom using the FBP and the 3D iterative reconstruction algorithm at different iteration numbers. The uniformity of the CT unit was evaluated from a summed reconstructed image with low statistical image noise of and at three magnifications. The resolution was determined from the edge functions of images of a set of precision-machined acrylic rods of different diameters at three magnifications and with different acquired and reconstructed pixel sizes. The dual-modality image co-registration was assessed using a set of CT and PET images obtained from a phantom consisting of a Ge-68 annulus ring phantom with an attached Na-22 point source. Finally, the preclinical imaging performance of the PET/CT system were evaluated from sample images several small animal studies. Our preliminary results showed the PET unit was able to achieve a system resolution of 0.85 mm and sensitivity of ∼8.9%, and the CT unit a highest resolution of ∼20 micron. We conclude the preclinical PET/CT system meet the stated specifications and is suitable for high performance preclinical molecular imaging of small animals.

Application of post reconstruction dual respiratory and cardiac motion compensation for 4D high-resolution small animal myocardial SPECT

We investigated the performance of a post reconstruction dual respiratory and cardiac (R&C) motion compensation method for improved image quality of 4D cardiac gated small animal myocardial perfusion (MP) SPECT images. A normal mouse was injected with ~8 mCi of Tc-99m sestamibi, anesthetized, fitted with ECG leads for cardiac gating signal acquisition, and placed on top of a pressure gauge bellow for respiratory motion measurements. A 2-hour list-mode dataset was acquired using a MILab small animal SPECT system fitted with a multi-pinhole collimator with 0.4 mm resolution in 5-minute sections. They were subsequently sorted for different acquisition times and reconstructed using a vendor provided OS-EM algorithm with simultaneous 6 respiratory and 8 cardiac equal-time gates over each motion cycle. Using a group-wise B-spline non-rigid image-based registration method, the deformation fields of the respiratory motion (respiratory motion) were estimated and applied to each cardiac phase for respiratory motion correction. Then, the respiratory motion compensated cardiac gated SPECT images were similarly used to estimate cardiac motion (cardiac motion) and later transformed to a reference frame and summed. Finally, the reference frame was inverse-transformed using the estimated cardiac motion to each of the 8 cardiac frames. The cardiac gated images with dual R&C motion compensation were compared to those without correction but with post-smoothing filter. The results showed the dual R&C motion compensation significantly reduced image noise level. At the same time, the image resolution was improved by 10% to 40% depending on the different acquisition times when compared with that obtained without motion compensation at the same image noise level. We conclude that dual R&C motion compensation provides significant reduction of noise level in 4D cardiac gated small animal MP SPECT images with minimum degradation of resolution. The improved image quality can be traded for reduction of acquisition time or radiation dose to the animal.