Saturday 10:30-12:00, Pier 2 & 3 & Salon C

Chair: Paul Kinahan, University of Pittsburg

S. Alenius¹, U. Ruotsalainen², J. Astola¹

^{1Tampere Univ. of Tech., Finland, 2Turku PET Centre, Finland}

The attenuation correction (AC) in PET using blank and transmission measurements suffers from statistical noise due to restricted scan times. Smoothed attenuation data blurs emission images. AC factors can be calculated using transmission images reconstructed from blank and transmission scans. For preventing noise to propagate the transmission image can be segmented, which requires assumptions of typical attenuation values of tissue types. High quality transmission images are desired for accurate correction. The median root prior (MRP) iterative reconstruction algorithm favors locally monotonic images. MRP was previously applied to PET emission studies. Because the assumption of local monotonicity is valid for both emission and transmission, MRP is used for reconstruction of transmission images. The attenuation factors can be determined with short transmission scan times. No smoothing of measured data is needed, and the transmission image needs no segmentation. The method is object independent, which is desirable for quantitative whole body PET studies.

S.J. Wilderman¹, N.H. Clinthorne¹, J.A. Fessler¹, W.L. Rogers¹

^{1Univ. of Michigan}

Images have been reconstructed from simulated Compton scattering camera data using a list mode likelihood technique. List mode methods are appealing in the Compton camera reconstruction problem because the total number of detected photons is significantly smaller than the number of possible combinations of position and energy measurements, leading to a much smaller problem than that faced by traditional iterative reconstruction techniques. For a realistic device, the number of counted photons is typically less than a tenth of a percent of the number of possible detector bins. The primary difficulty in applying list mode techniques is in determining the system response matrix. In this work, a list mode back-projection is performed first, and system matrix coefficents assigned only for tallied pixels. Projection data has been generated for a representative Compton camera system using Monte Carlo simulation for various source distributions and energies, and reconstructions performed. Less approximated methods of determining the coefficients have also been investigated.

T. Hasegawa¹, Y. Wada², H. Murayama³, T. Nakajima³

^{1Kitasato Univ., School of Allied Health Sci., 2Siemens-Asahi Medical Technologies Ltd., 3National Inst. for Radiological Sci. (NIRS)}

This study reports performance measurement of a PET scanner, SIEMENS/CTI ECAT EXACT HR+, with attention on some of the adjustable data-acquisition parameters in the operation software. In later PET scanners in capable of three dimensional (3D) mode, due to larger space variance of the acceptance angle, basic scanner performance such as sensitivity, spatial resolution, etc. varies more prominently in the field-of-view. This variation changes its manner when an adjustable parameter is changed. In HR+, there are some adjustable parameters (e.g. maximum ring difference, span, angular compression, overlap, etc.), on which scanner performance depends. Although the manufacturer's default values aregiven, they are not necessarily optimum for all the clinical applications. The results clarified the importance of understanding scanner performance as depending on various adjustable parameters. Theresults are useful for scan protocol optimization for each clinical application, and for precise interpretation of PET measurements.

R.S. Miyaoka¹, S.G. Kohlmyer¹, T.K. Lewellen¹

^{1Univ. of Washington, Seattle, Washington}

Phantom studies were performed to investigate detectability for a coincidence imaging system. Detectability is defined as:

D = contrast/noise (CNR) = ([sph]-[back])/stdev(back)

where [sph] and [back] are average pixel values for sphere and background regions of interest (ROIs) and stdev(back) is the standard deviation of the background ROIs. Spheres of 3.2, 2.5, 1.8, 1.6, 1.3, and 0.8cm diameter were placed at a radius of 5.75cm in a 22cm diameter cylindrical phantom. Data were acquired for sphere to background concentration ratios of 10.3, 5.4 and 2.9. The data were binned to 2D slices and reconstructed using filtered back projection. Attenuation and scatter corrections were not performed. Transverse views were used for ROI placement. For the image sets studied, detectability is dominated by contrast rather than noise. There were no measurable differences in CNRs between photopeak only (pp) and photopeak+Compton (pp+pC) images, using equal acquisition times. The smallest detectable spheres were 0.8cm, 1.6cm and 1.8cm for the 10.3:1, 5.4:1 and 2.9:1 image sets, respectively.

M. Bentourkia¹, A. Bol¹, J. Bodart¹, C. Michel¹, A. Coppens¹, M. Sibomana¹, D. Labar¹, A. DeVolder¹

^{1Positron Tomography Laboratory, Univ. of Louvain, 2 Chemin du Cyclotron, Louvain-la-Neuve B-1348, Belgium}

In order to characterize physiological cerebral activation or tissue response to a treatment, at least two FDG-PET studies are mandatory. In this work, a study of one hour duration with two injections at 30 min apart is reported. The separation of the two input curves (IC) consisted of fitting the first injection in the blood curve using spectral analysis, then estimating the second injection by removing the remnant of the first. Tissue time activity curves (tTAC) were fitted for the first 30 min using the first IC and extrapolated till 60 min. This extrapolated part was removed from the tissue response to the second injection before being fitted using the second IC. Other data were obtained for simple injection from which metabolic rates for glucose (MRG) estimated from 0-30 min was compared to that obtained from 0-60 min. Maximal MRG differences in double injection with activation were found to be five times higher than those from 0-30 and 0-60 min baseline simple injection. The method is expected to be more accurate to observe drug uptake or tumor response to a treatment.

R. Trebossen¹, B. Bendriem¹, M. Ribeiro¹, M. Sarazin¹, D. Strul¹, S. DuPont¹, F. Semah¹, P. Remy¹

^{1Service Hospitalier F. Joliot, CEA, 4 place du General Leclerc, Orsay France.}

The improvement of the PET sensitivity and resolution allows the measurement of the input function using dynamic scanning of the internal carotid arteries. Five subjects were scanned on an ECAT EXACT HR+ with the following sequence (number x length in seconds): 12 x 10, 2 x 30, 2 x 150, 3 x 300, 1 x 600, 1 x 1200 and 1 x 150. The scan started with the injection of the 18-FDG (168 MBq in average for the 5 subjects). The later scan was used to calculate rCMRglc images with the autoradiographic model using the 4 "PHELPS constants". A plasma input function was measured in a radial artery. Venous samples were also taken 5 minutes after the injection at the same rate as the arterial samples (venous function). The effects of the limited resolution on the carotid measurements were corrected using geometrical factors computed from each MRI registered to the PET images. The carotid input function was then obtained after correction for blood to plasma difference using each venous function. The relative differences between the rCMRglc values calculated with an input function measured in the carotid and the radial arteries were 6%, 2%, 27%, 20% and 8% respectively for the 5 subjects. In conclusion, for tracer modelling sensitive to the integral of the input function (such as FDG modelling), this method is adequate for quantification.

T.K. Lewellen¹, R.L. Harrison¹, S.K. Kohlmyer¹

^{1Univ. of Washington Medical Center}

Simulations of the scatter distributions in the Advance tomograph have been performed using the SimSET software package. The Zubal phantom was used, centered at the heart with activity in the heart or in all soft tissues. Data were binned axially for: 1) -0.844 to 0.844 cm; 2) 2.53 to 4.22 cm; and 3) -7.6 to 7.6 cm and sorted into 0 and 90 degree projections, and all lines of response (AL). For AL and the full field of view, the total scatter to trues ratio varies from 1.05 (300 keV) to 0.51 (425 keV). The ratio of multiple to total scatters (M%) varies slightly between the three axial zones and two isotope distributions. The average M% for all data was 38%, 31%, 28%, 24% and 20% for 300, 350, 375, 400, and 425 keV respectively. The M% vary somewhat with angle, and the 0 and 90 degree single scatter profiles show more structure than the multiple scatter profiles. This may limit accuracy of multiple scatter estimates derived from the single scatter distribution using a convolution model.

M. Bentourkia¹, A. Bol¹, C. Michel¹, A. Coppens¹, M. Sibomana¹, D. Labar¹, A. DeVolder¹

^{1Positron Tomography Laboratory, Univ. of Louvain, 2 Chemin du Cyclotron, Louvain-la-Neuve B-1348, Belgium}

PET images suffer from statistic fluctuations where pixels representing the same tissue present different amplitudes. These fluctuations become more important whith short frames acquired in dynamic studies. Another difficulty is the time consuming in fitting images pixel by pixel in order to obtain parametric images. In the present work, a flexible method, based on pixel variance, to focalize on specific tissue structures or to compute parametric images is reported. In the case of a fluorodeoxyglucose study, a template image is obtained from the last frame where the gray matter present the higher uptake. Three independent operations can be performed: thresholding, estimation of windowed pixel mean and variance and finally scaling. Pixels representing a certain structure or only noise can be discriminated from others by setting levels on the variance, while the scaling permits to group pixels of comparable amplitude. From the template image, pixels identified to have comparable dynamic behavior are averaged and fitted, thus reducing statistic fluctuation and computation time.

C.I. Hu¹, C.E. Huang², Y.E. Xiaoli²

^{1C Huang, 2Xiaoli Yu}

Principle component analysis (PCA) is a well-known multivariate image analysis technique to reduce the dimensions and noise level in dynamic PET data. However, the contrast of small lesions is often reduced by PCA, due to the lack of a priori physiological knowledge of lesions. In this paper, a new method, modified from the generalized sidelobe canceler (GSC), is proposed to incorporate the physiological features extracted from the visible tumor and normal tissue in order to enhance the non-palpable lesions in the background noise. Two different schemes are adopted to implement the modified GSC. One uses both physiological features of lesions and normal tissue, while the other employs only that of lesions. The multidimensional maximum likelihood estimation is applied to find physiological factors which span time activity curve subspaces of lesions and normal tissue from dynamic PET data. Results show that the proposed modified GSC can substantially stretch the contrast of small lesions, thus is able to aid in detecting non-palpable tumors.

D. Tsai¹, S. Watanabe¹

^{1Dept. of Electrical Eng., Gifu National College of Tech.}

This work presents a method for optimization of fuzzy reasoning by genetic algorithms (GA's) for discrimination of myocardial heart disease from ultrasound images. In the present study, Gaussian-distributed membership functions are used, and the standard deviations of the functions acting as optimum parameters are optimized through learning process using the GA's. In the GA-based learning, a two-step fitness function in the selection operation is employed in order to increase the precision and accuracy of the discrimination. The results of the experiments are very promising. In the best case, we achieve a discrimination rate of 95.8% with high reliability.

A.J. Da Silva¹, H. Tang², B.H. Hasegawa^1,2

^{1Dept. of Radiology, Univ. of California San Francisco, 2Bioengineering Graduate Group, Univ. of California San Francisco and Berkeley}

Partial volume errors cause SPECT to underestimate true radionuclide content by approximately 50% in absolute radionuclide quantitation of regional myocardial perfusion. We have developed a new technique for compensating myocardial SPECT images for partial volume errors using coregistered X-ray CT images. The CT-derived myocardial mass defines a template that can be assigned unit activity and mathematically projected with a realistic physical model of the radionuclide imaging process including non-ideal collimation and incorporating an object-specific attenuation map from CT. The template projections then are reconstructed using the SPECT reconstruction algorithm to obtain a pixel-by-pixel partial-volume correction for the myocardial SPECT image. Experiments in phantoms demonstrate that this technique achieves absolute quantitation of myocardial radionuclide concentration with an accuracy error of less than 8%. This method also has the potential for correcting background effects such as "spill in" of background counts from uptake in the liver.

J. Kim¹, K. Kim²

^{1POSTECH Information Research Laboratory, Pohang Univ. of Sci. and Tech., 2Dept. of Mathematics, Pohang Univ. of Sci. and Tech.}

A multiresolution filtering method based on MDLSs (minimum detectable lesion sizes) provides an optimal solution for SPECT reconstruction problem. From m MDLSs determined from SPECT lesion detectability curve, we generated a class of the Cj filters (j=1,2,.,m), which produce stronger resolution recovery effect as j approaches 1, and stronger noise smoothing effect as j approaches m. For every pixel of the images obtained using these Cj filters, a local homogeneity is determined in a window centered at the pixel by comparing the local variance and its robust global estimate. Among the windows declared as `homogeneous', we select the window of the Cj filter image corresponding to the largest j, and substitute the pixel with the spatial average computed in the selected window. If no homogeneous window exists, an adaptive least squares smoothing method is applied for the window of the C1 filter image. From the results of simulated phantom studies, we observed good performance of this filtering algorithm in effectively reducing image noise and enhancing image structures.

R.R. Fulton¹, S. Eberl¹, S.R. Meikle¹, B.F. Hutton², M. Braun³

^{1Dept. of PET & Nuclear Medicine, Royal Prince Alfred Hospital, 2Dept. of Medical Physics, Westmead Hospital, 3Dept. of Applied Physics, Univ. of Tech. Sydney}

Patient motion during brain SPECT studies can degrade resolution and introduce distortion. We have developed a correction method which incorporates a motion tracking system to monitor the position and orientation of the patients head during acquisition. Correction is achieved by spatially repositioning projections according to measured head movements and reconstructing the projections with a fully three-dimensional (3D) algorithm. The method has been evaluated in SPECT studies of the Hoffman 3D brain phantom performed on a triple head camera with fan beam collimation. Movements were applied to the phantom and recorded by a head tracker during SPECT acquisition. Fully 3D reconstruction was then performed using the motion data provided by the tracker. Correction accuracy was assessed by comparison with a motion free study, visually and by calculating mean squared error (MSE). In all studies, motion correction reduced distortion and improved MSE by a factor of 4 - 6. We conclude that this method can compensate for head motion under clinical SPECT imaging conditions.

A. Sitek¹, E. Di Bella¹, G.T. Gullberg¹

^{1Dept. of Radiology, Univ. of Utah}

Factor analysis of dynamic structures (FADS) is a technique used in the automatic extraction of time activity curves (TACs) from dynamic sequences. Although it has been reported that factor analysis with non-negativity constraints can, in certain cases, correlate with region of interest (ROI) measurements in SPECT and PET heart studies, the method does not ensure a unique solution. In this work it is shown that the FADS solution can be made unique by including an entropy term in the objective function. Both the FADS technique and the new maximum entropy method were tested on simulated data and experimental 99mTc-teboroxime canine cardiac data. The results showed that the FADS technique using only non-negativity constraints produced curves that did not closely approximate the true curves, while the new method resulted in TACs that closely resembled the true curves. Thus, the inclusion of an entropy term is a useful approach to extracting accurate and unique TACs from dynamic SPECT data.

C.J. Dykstra¹, A.M. Celler², R.J. Jaszczak¹

^{1Duke Univ. Medical Center, Box 3949, Durham, North Carolina, , 2Division of Nuclear Medicine, VGH, 855 W. 12th Ave., Vancouver, Canada, V5Z 1M9}

Two of the limitations on the utility of SPECT and planar scintigraphy for the non-invasive detection of breast carcinoma are small sizes of many tumors and low contrast between tumor uptake and background. Use of image processing can improve the visibility of tumors which are at the limit of hardware resolution. Smoothing, by some form of image averaging, either during or post-reconstruction, is widely used to reduce noise and thereby improve the detectability of regions of elevated activity. However, smoothing degrades resolution and, by averaging together closely spaced noise, can make noise appear as information.

Image morphing by erosion and dilation does not average image values; instead removing small variations without degrading contrast or resolution. It may, therefore, be better than smoothing for noise removal in breast images. In this work we investigate the utility of image morphing for improved breast tumor detection. Phantom and clinical data of tumors at the limit of visibility from both SPECT and planar imaging will be used to evaluate the technique.

P.M. Mansour^1,2, M.F. Smith¹, W.C. Barker¹, S.L. Bacharach¹

^{1National Institutes of Health, Bethesda, MD , 2Univ. of Maryland at College Park, College Park, MD}

Gated Blood Pool (GBP) SPECT has the potential to assess left-ventricular (LV) regional wall motion (RWM) more accurately than planar scans. However, due to limited scan times for exercise studies, GBP SPECT produced images that appeared too noisy for regional analysis. We investigated a counts-based scheme for quantifying 2 RWM parameters, regional EF (REF) and relative regional phase (RRP). We divided the LV cavity into 12 3-d regions. We used GBP SPECT rest (~30 min) studies from 10 normals, and 2 patients with known RWM defects. By simulating the effect of reduced scan times (~3 min), we estimated the variability among normals in the parameters from a short exercise study. We show that the SD due only to increased noise is ~15% of the total SD(REF) and ~30% of the total SD(RRP). This resulted in only a very small loss of sensitivity of the methods ability to detect RWM defects in the 2 patients. The results indicate that this scheme for RWM analysis may be relatively unaffected by greatly reduced scan times, making short exercise studies potentially feasible.

C. Bai¹, G.L. Zeng¹, D.J. Kadrmas¹, G.T. Gullberg¹

^{1Dept. of Radiology, Univ. of Utah}

A variety of effects may make the apical region appear colder than it actually is in SPECT reconstructions. In this paper, simulations were performed using fan-beam collimators to study the effect of geometric point response and attenuation corrections on the apical region in cardiac SPECT. The images were reconstructed using OS-EM algorithm. The smoothing introduced by interpolations in the algorithm was also studied. It was shown that reconstructions without performing the geometric point response correction did not further decrease the activity of the apical region. Use of the attenuation map reconstructed from the truncated transmission data when performing the attenuation correction might result in an artificially cool apical region. Comparison of the images reconstructed with the OS-EM and the FBP algorithms demonstrated that the partial volume effect related to the interpolations in the OS-EM algorithm did not introduce visible activity decrease in the apical region for the cardiac orientation used.

Z.I. Liu¹, T.E. Obi¹, M.E. Yamaguchi¹, N.E. Ohyama¹

^{1Zheng Liu, Takashi Obi, Masahiro Yamaguchi and Nagaaki Ohyama}

The scatter in SPECT contains activity information and some reports have shown that it is possible to use this information for improving image quality. The general method is modeling scatter in multiple energy windows in reconstruction process. However, what the performance will be and how noise will behave when using scatter in reconstruction have not been well answered. In this work, we used singular value decomposition to investigate the method of modeling scatter in multiple energy windows. The system performance and image quality were evaluated in terms of resolution kernels and normalized mean-square error (NMSE). For noise behavior investigation, reconstruction was accomplished by estimating the variance of reconstructed voxel values and the effectiveness of using scatter was evaluated by the resolution kernel analysis. The results show that the improvements were shown in the NMSE of the images and the resolution kernels. When photon counts fall below about one million, there still is effectiveness of using scatter for some fewer projection cases.

T. Hasegawa¹, M. Suzuki², H. Murayama³, T. Irie³, K. Fukushi³, Y. Wada⁴

^{1Kitasato Univ., School of Allied Health Sci., 2Inagi Municipal Hospital, 3National Inst. for Radiological Sci. (NIRS), 4Siemens-Asahi Medical Technologies Ltd.}

This study assessed influence of out-of field-of-view (FOV) radioactivity on brain acetyl-cholinesterase (AChE) activity measurement with tracer [11C] N-methyl-4-piperidyl-acetate ([11C]MP4A) by Positron Emission Tomography (PET) in the three dimensional (3D) mode. From dynamic scans on the breast and head parts of a volunteer, it was found that the breast part radioactivity could be much higher than the head one immediately after the tracer injection with a drastic time dependence. In order to measure quantitative influence of the out-of FOV radioactivity, dynamic phantom measurement was performed with two kinds of radioactivity sources, 18F in FOV and 11C for out-of FOV. As a result, expected errors in the AChE activity parameter, a rate-constant parameter in a compartment model analysis, were found to be less than one percent.

R.J. Ott¹, M.A. Flower¹, A.J. Reader¹, K. Erlandsson¹

^{1Physics Dept., Royal Marsden Hospital/Inst. of Cancer Research, Sutton, Surrey, SM2 5PT, UK}

Images from the PETRRA positron camera will be produced directly from list mode data acquisition. A method of scatter correction in backprojection space has been developed by deriving from experimental measurements an algorithm relating attenuation and scatter. A 3D image space attenuation matrix and a measured scatter response function are used to generate a 3D scatter matrix. The attenuation and scatter matrices are convolved with an in-air PRF to produce back-projection space distributions. The scatter distribution is subtracted from the acquired back-projected image, attenuation corrected and deconvolved with the in-air PRF. The method has been applied to a Monte Carlo simulated phantom, an experimental phantom and a patient study. The predicted scatter distribution agrees with the simulation to ~3%. The correction improves image contrast for both the experimental phantom and patient data. The method could be applied to any large-area positron camera acquiring data in list mode.

T. Schmitz¹, J. Noetzel¹, H. Herzog¹

^{1Inst. für Medizin, Forschungszentrum Jülich, D- Jülich}

The scatter fraction in 3D volume PET data may be up to 40 % for brain and over 50 % for body scans. In order to get quantitatively accurate results, the data have to be scatter corrected. In case of Siemens/CTI ECAT EXACT HR+ scanners, a single scatter simulation based correction is applied. It needs a preliminary reconstruction of the measured data without scatter correction, the estimated image. The type and width of the estimated image reconstruction filter can be set at runtime. We asked, whether and how the estimated image filter selection influences the final, scatter corrected image. Different filter types with two filter widths (2 mm and 8 mm) were applied in the reconstruction of experiments with the EEC thorax phantom and hollow sphere insert. The study showed that the estimated image filtering has an influence on the outer 1/3 of the planes. Using a Hanning instead of a Ramp filter (the standard setting) reduces the standard deviation in ROIs in the outer planes. Further analysis is being performed in order to find the reason for these effects.

N. Wainer¹, J. Bouhnik¹, M. Wilk¹, P. Chouraqui²

^{1Elgems Ltd., 2Tel-Aviv Univ., Haim Sheeba Medical Center}

The performance of simultaneous transmission-emission SPECT using moving gadolinium-153 line source depends in part on the proportion of the activities of the line source and the emission background. The goal of the study was to evaluate the influence of transmission source decay on the corrected SPECT images obtained at three different times, in a period of 27 months. Simultaneous transmission-emission SPECT images of a cardiac phantom filled with Tl-201 were obtained using a dual-head 90O gamma camera equipped with two moving collimated Gd-153 line-sources. Gd-153 activities in this studies were calculated to be 230 mCi, 49 mCi, and 24 mCi. Visual comparison between the corrected SPECT images permit to evaluate the correction quality. At very low activities, the spine (teflon) of the phantom is not seen in the transmission images and some degradation of the correction, mainly in the apex is seen. The corrected images display acceptable correction for Gd-153 activities over 50 mCi.

E.F. Hollinger¹, M.K. West¹, W. Chang¹

^{1Dept. of Diagnostic Radiology and Nuclear Medicine, Rush-Presbyterian-St. Luke's Medical Center, Chicago, IL}

Attenuation correction (AC) with short, sequential asymmetric fan-beam (AsF) TCT improves the uniformity of normal cardiac SPECT. The intent of this study was to assess whether AC from short AsF-TCTs also improves lesion size and severity estimates in abnormal myocardium. An anthropomorphic cardiac-thorax phantom was used to acquire Tl-201 SPECT of a transmural defect with activity 0-75% of normal tissue. ECTs were obtained in the phantom and in air for 8 defect positions and corrected with 1-20 min TCTs. Defect size was measured where activity was <80% of that of normal tissue. Defect severity was assessed by the ratio of activity in the defect and nearby myocardium. Without AC, septal defect size is most increased (175-205% relative to in-air) because the defect seems to extend into the (attenuated) inferior wall. Applying AC (TCTs 34 min) improves defect size to 85-110% relative to in-air. Anterior defects show the least change. AC alters lesion severity by less than 15%, but yields better contrast between adjacent normal and defect regions.

I. Laurette¹, A. Welch², F. Natterer³, R. Clackdoyle¹, G.T. Gullberg¹

^{1Dept. of Radiology, the Univ. of Utah, USA, 2Dept. of Biomedical Physics and Bioengineering,, 3Univ. of Aberdeen, Scotlandd, 4Universitat Münster, Münster, Germany}

Recently investigators have studied the potential of using the consistency between emission data and the corresponding transmission data to perform attenuation correction. This technique allows one to compute the attenuation distribution without acquiring transmission data. However, because the system is highly under-determined, it is necessary to apply a priori information to the attenuator. This study presents three different ways to define the attenuator. The first method uses an affine deformation of an initial boundary. The second is based on the assumption that the unknown object is defined by a spline. The third technique utilizes the same information as the second except that the untruncated part of the transmission sinogram is used. Results of a heart simulation study are presented. It is shown that the first two methods provide equivalent results because these methods give an uniform estimation of a non-uniform attenuator. These results show that the third method provides better results since information from the real transmission sinogram is included.

S. Jang¹, J.R. Saffer¹, R.J. Smith¹, J.S. Karp¹

^{1Univ. of Pennsylvania Medical Center}

The purpose of this study was to evaluate factors affecting the quality of the measured attenuation map and nonuniform attenuation compensation (AC) for simultaneous transmission-emission system, especially when using fan beam geometry in Tc-99m myocardial perfusion imaging. We investigated factors including the effect of truncation, number of reconstruction iterations, sampling, filtering, scatter correction (SC) of ECT data and noise in the TCT data including transmission source strength and scan duration. Each of the factors studied is influential and needs to be optimized for a specific imaging task. Severe truncation of the lungs in obese patients or by mispositioning reduces the uniformity of the myocardial images and may produce artifacts. Eight to twenty iterations of OS-EM produce images with similar %RMS noise. When the accuracy of quantitation is pursued, both AC and SC are necessary. In conclusion, clinical protocols should be determined after carefully evaluating these factors to minimize noise and artifacts as well as optimize the spatial resolution in studies.

J. Nuyts¹, P. Dupont¹, S. Stroobants¹, A. Maes¹, L. Mortelmans¹, P. Suetens²

^{1Nuclear Medicine, K.U.Leuven, B3000 Leuven, Belgium, 2ESAT-PSI, K.U.Leuven, B3000 Leuven, Belgium}

Transmission images reconstructed with a maximum likelihood (ML) algorithm are often reported to be visually and/or quantitatively superior when compared to filtered backprojection (FBP) reconstructions. We report on a simulation experiment which verifies whether this feature of ML leads to improved attenuation correction for positron emission tomography (PET). In this experiment, the quality of the images is assessed with a detection task based on a series of simulated PET images. In 50% of these images, a tumor is present. Tumor location and contrast are varied. The images are presented in random order to human observers. The cumulative number of wrong answers as a function of contrast is regarded as a measure of image quality. Although ML reduced the streak artifacts in the attenuation map, tumor detection score was best for the classical attenuation correction approach (ratio of blank and transmission scan).

H. Erdogan¹, J.A. Fessler¹

^{1Univ. of Michigan}

Previous methods for optimizing the scan times for PET transmission and emission scans under a total scan time constraint were based on linear non-statistical methods and used noise equivalent counts (NEC) criteria. The scan times determined by NEC analysis may be suboptimal when nonlinear statistical image reconstruction methods are used. For statistical image reconstruction, the predicted variance in selected regions of interest is more appropriate than NEC analysis. We propose a new method for optimizing the relative scan times (fractions) based on analytical approximations to the covariance of images reconstructed by conventional as well as penalized-likelihood methods. We perform simulations to compare predicted standard deviations with empirical ones. Results show that for statistical transmission image reconstruction, the optimal fraction of the scan time devoted to transmission scanning is shorter than for conventional transmission smoothing.

T.G. Turkington¹, C.M. Laymon¹, N. Wainer², R.E. Coleman¹

^{1Radiology Dept., Duke Univ. Medical Center, 2Elgems, Ltd.}

We have investigated multiple point source transmission scanning for attenuation correction in rotating gamma camera coincidence imaging. The scanner is a two-headed system with coincidence capability. The cameras are equipped with transaxial lead septa (3 mm thick, 10 mm gaps, extending 6 cm from graded absorber on camera surface). These septa exist primarily to reject emission photons which deviate substantially from the transaxial plane. For this work, a Cs-137 point source was placed between two septa at the edge of a camera, providing an offset fan-beam flux to the opposite camera. Axial resolution and missing data issues were evaluated with varying amounts of collimation (thicker septa near the source) and multiple point sources (to achieve a large axial field of view), which were simulated with multiple acquisitions of phantoms stepped through the axial direction on the scanner table. Reconstruction was done approximating the data as 2D. It was determined that sufficient sampling and resolution could be obtained with source in every other septum slot.

M.T. Madsen¹

^{1Univ. of Iowa, Iowa City, IA}

It is desirable to use mean population attenuation coefficients reduce the statistical fluctuations in measured attenuation maps. In addition, the use of mean attenuation coefficients is also expected to facilitate algorithmic approaches to attenuation correction. Because of the large variation in lung attenuation coefficients both within and among individuals, there is concern about the errors using mean attenuation might create. To test the magnitude of this problem, myocardial metabolic studies were simulated. Attenuated projections were generated from 20 realistic emission and transmission distributions spanning the heart volume. The projections were corrected with both the actual attenuation map and with a mean attenuation map. Quantitative comparisons of the reconstructed images demonstrate no significant differences in the myocardial distributions. However, errors in the lung field exceeded 300% in some regions.

T. Satoh¹, K. Ogawa¹

^{1Hosei Univ., Tokyo, Japan}

In Monte Carlo simulation, photon transport requires calculating the distance from the location of the photon to the nearest boundary of each uniform attenuating medium along its path of travel, and comparing the distance with the length of its path generated in the simulation. In this paper, we propose a new method which omits the calculation of the location of the exiting point of the photon and of the distance between the exit point and the original position. The method only checks the medium of the voxel existing along with the path of the photon transport. If the medium is different from that in the voxel which the photon emitted, then we calculate the location of the entry point in the voxel, and the length of the path is compared with the mean free path length generated by a random number. Simulations using an MCAT phantom show that the ratios of the calculation time were 1.0 for voxel based method, 0.375 for the proposed method with a 256x256 matrix image and the effectiveness was confirmed.