Reversible cerebral hypoperfusion in Lyme encephalopathy.

Lyme encephalopathy (LE) presents with subtle neuropsychiatric symptoms months to years after onset of infection with Borrelia burgdorferi. Brain magnetic resonance images are usually normal. We asked whether quantitative single photon emission computed tomography (SPECT) is a useful method to diagnose LE, to measure the response to antibiotic therapy, and to determine its neuroanatomic basis. In 13 patients with objective evidence of LE, SPECT demonstrated reduced cerebral perfusion (mean perfusion defect index [PDI] = 255), particularly in frontal subcortical and cortical regions. Six months after treatment with 1 month of intravenous ceftriaxone, perfusion significantly improved in all 13 patients (mean PDI = 188). In nine patients with neuropsychiatric symptoms following Lyme disease, but without objective abnormalities (e.g., possible LE), perfusion was similar to that of the treated LE group (mean PDI = 198); six possible LE patients (67%) had already received ceftriaxone prior to our evaluation. Perfusion was significantly lower in patients with LE and possible LE than in 26 normal subjects (mean PDI = 136), but 4 normal subjects (15%) had low perfusion in the LE range. We conclude that LE patients have hypoperfusion of frontal subcortical and cortical structures that is partially reversed after ceftriaxone therapy. However, SPECT cannot be used alone to diagnose LE or determine the presence of active CNS infection.

Nonuniform collimator sensitivity: improved precision for quantitative SPECT.

UNLABELLED: Attenuation of photons degrades both the accuracy and the precision of SPECT images; attenuation correction algorithms correct the bias but cannot improve precision. Increased noise due to photon attenuation is most pronounced in regions deep in solid body sections, such as the brain and abdomen. We have quantified the degradation in performance in several estimation tasks that can be attributed to photon attenuation and determined the degree to which performance might be improved by a collimator with a nonuniform sensitivity profile. METHODS: The analysis used ideal-observer models of performance in tasks involving estimation of the activity and size of a focal lesion. The models were based on the Cramer-Rao lower bound on the variance with which lesion activity and size can be estimated by an unbiased procedure. To quantify the effects of attenuation, values of the Cramer-Rao bound were calculated for each estimation task as a function of location of the lesion in circularly-shaped attenuators of 10- and 20-cm radii, with and without attenuation. Values of the bound were also determined for two nonuniform sensitivity profiles, one of which was designed to equalize (or nearly equalize) task performance throughout the image. RESULTS: For 99mTc, photon attenuation increased the variance of the estimates by factors of up to 4.5 for the 10-cm radius attenuator and up to 20.0 for the 20-cm radius attenuator. A collimator with a nonuniform sensitivity function reduced variance by factors of up to 1.8 for the 10-cm radius attenuator and up to 2.8 for the 20-cm radius attenuator. These gains in estimation performance were insensitive to the imaging task and to deviations from the assumed attenuator size and shape. CONCLUSION: Performance in estimation tasks using images from SPECT systems with uniform sensitivity collimators is considerably lower than the theoretical optimum. We have derived a sensitivity function, realizable using existing technology, that improves performance substantially.

Quantitative brain SPECT in Alzheimer's disease and normal aging.

To improve the diagnostic utility of brain single-photon emission computed tomography (SPECT) in Alzheimer's disease (AD), we have developed and evaluated an objective method of differentiating patients and healthy elderly controls using a quantitative image analysis protocol. HMPAO-SPECT image datasets from 29 patients with probable AD and 78 age-matched controls were registered to a common anatomic frame of reference. Activity levels within 120 standardized cortical volumes were determined by an automated procedure. Subjects were classified into normal and AD groups by quadratic discriminant analysis using two features: global average activity level and average normalized activity levels within the two clusters of standardized volumes identified as most significantly different in AD by analysis of covariance. The classification used split-half replication to ensure valid results. Classification performance quantified by the area under a binormal ROC curve fitted to the data was 0.923 +/- 0.036; at a threshold likelihood ratio of 1, the sample sensitivity was 91% and specificity was 86%. We conclude that quantitative SPECT accurately distinguishes AD patients from elderly controls.

SPECT image noise power: effects of nonstationary projection noise and attenuation compensation.

The effects of nonstationary projection noise and attenuation compensation are included in a theoretic calculation of the radial noise power spectrum (NPS) of single photon emission computed tomographic (SPECT) images. The nonstationary projection noise is shown to cause a relatively large d.c. component in the NPS, especially for small objects; whereas, attenuation compensation increases the total noise variance while only changing the d.c. component slightly. The theoretic calculation agrees well with a radial NPS estimated from 1,250 SPECT images simulated from projections of random Gaussian noise, even though the effects of discrete data collection and reconstruction were not included in the theoretical model.

Absolute activity quantitation in simultaneous 123I/99mTc brain SPECT.

UNLABELLED: Dual-isotope imaging can allow simultaneous assessment of brain perfusion using a 99mTc-labeled tracer and neurotransmission using an 123I-labeled tracer. However, the images are affected by scatter, cross talk, attenuation, distance-dependent collimator response (DCR), and partial-volume effect. We determined the accuracy and precision of activity quantitation in simulated normal and pathologic studies of simultaneous 123I/99mTc brain SPECT when compensating for all degrading phenomena. METHODS: Monte Carlo simulations were performed using the Zubal brain phantom. Contamination caused by high-energy 123I decay photons was incorporated. Twenty-four 99mTc and 123I activity distributions were simulated on the basis of normal and pathologic patient activity distributions. Cross talk and scatter were corrected using a new method based on a multilayer perceptron artificial neural network (ANN), as well as by the asymmetric window (AW) approach; for comparison, unscattered (U) photons of 99mTc and 123I were recorded. Nonuniform attenuation and DCR were modeled in an iterative ordered-subset expectation maximization (OSEM) algorithm. Mean percentage biases and SDs over the 12 normal and 12 pathologic simulated studies were computed for each structure with respect to the known activity distributions. RESULTS: For 123I, AW + OSEM yielded a bias of 7% in the cerebellum, 21% in the frontal cortex, and 36% in the corpus callosum in the simulated normal population. The bias was increased significantly in the striata of simulated pathologic studies (P < 0.05). The bias associated with ANN was significantly lower (<9% in these brain structures, P < 0.05). For 99mTc with AW + OSEM, the bias was 60% in the corpus callosum, 36% in the striata, and 18%-22% in the cortical lobes in the simulated normal population. This bias was <11% in all brain structures with ANN. In the simulated pathologic population, the bias associated with AW increased significantly in the cortical lobes to 55% (P < 0.05), although it did not change significantly with ANN. CONCLUSION: The accuracy and variability over simulated normal and pathologic studies of both 99mTc and 123I activity estimates were very close with ANN to those obtained with U + OSEM. ANN + OSEM is a promising approach for absolute activity quantitation in simultaneous 99mTc/123I SPECT.

Maximum-likelihood estimation: a mathematical model for quantitation in nuclear medicine.

In a stimulation study, we investigated the limitations of quantitation in nuclear medicine using a maximum-likelihood (ML) estimation model. We estimated activity, size, and position of a disk-shaped object on a circular, uniform background of unknown activity. The parameter estimates were unbiased, and their standard error was proportional to the square root of the total image counts. The estimates of object activity and size were strongly (negatively) correlated; the position estimates, however, were not correlated with estimates of any other parameters. This implies that a priori knowledge of object location does not improve precision. The minimal model of quantitation tasks should incorporate unknown object activity and size as well as unknown background activity. The ML estimation procedure was used to investigate the trade-off between resolution and sensitivity in gamma camera collimator design. The results implied that for complex tasks such as the multiparameter estimation task investigated here, optimum performance is achieved at a better resolution than that previously found optimal for detection of a well-specified object in a known background.

Limitations of dual-photopeak window scatter correction for brain imaging.

UNLABELLED: A method for performing scatter corrections that would directly use the photopeak information and would be straightforward for use in clinical practice would be attractive in SPECT imaging. The dual-photopeak window method may be such a method. It relates the scatter fraction to the ratio of the lower to the total parts of a split-photopeak window. We investigated the use of this scatter correction method on a dedicated brain camera. METHODS: Calibration curves for the Ceraspect, a dedicated brain imaging camera, were obtained for four split-window combinations using point sources in air and water. Simulations of the Ceraspect calibration curves at several energy resolution values were obtained using a Monte Carlo simulation of the instrument. RESULTS: The calibration curves, experimental and simulated, revealed an ambiguous and unstable relationship between lower-to-total ratio and scatter fraction. CONCLUSION: The unsatisfactory calibration curves can be attributed to the limited scatter produced in a brain-sized phantom during the calibration process and inherent stability problems in the calibration process. The dual-photopeak window method is not usable for small-field imaging systems and may even be unstable for larger-field systems.

Collimator selection for SPECT brain imaging: the advantage of high resolution.

We compared a prototype long-bore (LB) high-resolution collimator with a low-energy, general-purpose collimator (LEGP) using 99mTc and 123I. The LB collimator provided a 56% improvement in tomographic resolution (autocorrelation width) over the LEGP for 99mTc; for 123I, the gain was 79%, providing substantially improved contrast for small structures. The sensitivity of the LB collimator, however, is only 32% of that of the LEGP. The imaging tasks to be performed on [123I]IMP brain scans involve localization and discrimination of small, high-contrast brain structures and detection of abnormalities in shape, size, or uptake, rather than simple detection of lesions. Observer performance in such higher-order imaging tasks is known to depend on high spatial resolution, even at the cost of sensitivity. Patient studies confirmed that, for resolution-limited tasks, the increase in resolution outweighs the increased noise due to a loss in sensitivity. When the tomographic resolution of the LB collimator was degraded by smoothing to that of the LEGP, the noise in the LB images was lower than that of the LEGP by a factor of 2.9 for the same imaging time, demonstrating the advantage of high-resolution detectors and a smooth reconstruction filter over low-resolution detectors without smoothing. Therefore, collimators designed for high resolution, even at substantial cost in sensitivity, are expected to yield significant improvements for brain SPECT. Geometric calculations show that commercially available low-energy, high-resolution cast collimators promise to meet these requirements.

The effects of misregistration of the projections on spatial resolution of CT scanners.

Misregistration of the projections in 360 degrees computed tomographic (CT) scanners has been found to blur the image without generating artifacts. The effects of this error were investigated by analytical methods and by reconstruction of real and simulated data. The point-spread function which results from shifting each projection by a constant distance epsilon consists of a two-dimensional impulse function surrounding a region of negative density. The locus of the impulse function is a circle for parallel-beam geometry and a sixth-order curve for fanbeam geometry. The anisotropy and position dependence of the point-spread function in fanbeam geometry have been characterized. The line-spread function due to the error in parallel-beam geometry consists of two delta functions located at +/- epsilon. In fanbeam geometry, the line-spread function consists of two delta functions separated by approximately 2 epsilon, with the locations of the impulses dependent on the position and orientation of the line. This error, combined with other sources of blurring, results in a system edge-response function which contains a flat region at one-half the maximum density.

Image quality of an analog radiation therapy simulator-based tomographic scanner.

The spatial resolution and noise level of images produced by a commercial analog tomographic scanner have been measured and compared to those of images reconstructed digitally from projections from the same detector. The full width at half maximum of the line spread function was 3.6 mm for images from the analog scanner and 1.1 mm for the digitally reconstructed images. The standard deviation of the CT numbers over a 10-cm2 circular area at the center of a large water phantom, calculated as a percentage of the linear attenuation coefficient of water, was 3.5% for the analog images, 15.4% for high-resolution digital images, and 3.2% for digital images reconstructed using a convolution filter which reduced the resolution to that of the analog images. The data contributing to each digital image were fewer than those contributing to each analog image by a factor of 10. The noise level did not depend on tube current in either the analog or the digital images. The utility of this analog device in radiation therapy planning will depend upon whether errors in contour localization resulting from transferring data from diagnostic CT scanners exceed the errors due to its poorer image quality.

Collimator optimization for lesion detection incorporating prior information about lesion size.

A Bayesian estimator has been developed as a paradigm for human observer performance in detecting lesions of unknown size in a uniform noisy background. The Bayesian observer used knowledge of the range of possible lesion sizes as a prior; its predictions agreed well with the results of a six-observer perceptual study. The average human response to changes in collimator resolution, as measured by the detectability index, dA, was tracked by the Bayesian detector's signal-to-noise ratio (SNR) somewhat better than by two other estimation models based, respectively, on lesser and greater degrees of lesion size uncertainty. As the range of possible lesion sizes increased, the Bayesian detector's SNR decreased and the optimal collimator resolution shifted towards better resolution. An analytic approximation for the variance of lesion activity estimates (which included the same prior) was shown to predict the variance of the Bayesian estimator over a wide range of collimator resolution values. Because the bias of the Bayesian estimator was small (< 1%), the analytic variance estimate permitted a rapid and convenient prediction of the Bayesian detection SNR. This calculation was then used to optimize the geometric parameters of a two-layer tungsten collimator being constructed from crossed grids for a new imaging detector. A Monte Carlo program was first run to estimate all contributions to the radial point-spread function for collimators of differing tungsten contents and spatial resolution values, imaging 140-keV photons emitted from the center of a 15-cm-diameter, water-filled attenuator. The optimal collimator design for detecting lesions with unknown diameters in the range 2.5-7.5 mm yielded a system resolution of approximately 8.5-mm FWHM, a geometric collimator efficiency of 1.21 x 10(-4), and a single-septum penetration probability of 1%.

Evaluation of scatter compensation methods by their effects on parameter estimation from SPECT projections.

Three algorithms for scatter compensation in Tc-99m brain single-photon emission computed tomography (SPECT) were optimized and compared on the basis of the accuracy and precision with which lesion and background activity could be simultaneously estimated. These performance metrics are directly related to the clinically important tasks of activity quantitation and lesion detection, in contrast to measures based solely on the fidelity of image pixel values. The scatter compensation algorithms were (a) the Compton-window (CW) method with a 20% photopeak window, a 92-126 keV scatter window, and an optimized "k-factor," (b) the triple-energy window (TEW) method, with optimized widths of the photopeak window and the abutting scatter window, and (c) a general spectral (GS) method using seventeen 4 keV windows with optimized energy weights. Each method was optimized by minimizing the sum of the mean-squared errors (MSE) of the estimates of lesion and background activity concentrations. The accuracy and precision of activity estimates were then determined for lesions of different size, location, and contrast, as well as for a more complex Bayesian estimation task in which lesion size was also estimated. For the TEW and GS methods, parameters optimized for the estimation task differed significantly from those optimized for global normalized pixel MSE. For optimal estimation, the CW bias of activity estimates was larger and varied more (-2% to 22%) with lesion location and size than that of the other methods. The magnitude of the TEW bias was less than 7% across most conditions, although its precision was worse than that of CW estimates. The GS method performed best, with bias generally less than 4% and the lowest variance; its root-mean square (rms) estimation error was within a few percent of that achievable from primary photons alone. For brain SPECT, estimation performance with an optimized, energy-based, subtractive correction may approach that of an ideal scatter-rejection procedure.

The noise power spectrum of CT images.

An expression for the noise power spectrum of images reconstructed by the discrete filtered backprojection algorithm has been derived. The formulation explicitly includes sampling within the projections, angular sampling, and the two-dimensional sampling implicit in the discrete representation of the image. The effects of interpolation are also considered. Noise power spectra predicted by this analysis differ from those predicted using continuous theory in two respects: they are rotationally asymmetric, and they do not approach zero at zero frequency. Both of these properties can be attributed to two-dimensional aliasing due to pixel sampling. The predictions were confirmed by measurement of noise power spectra of both simulated images and images from a commercial x-ray transmission CT scanner.