B.C. Penney

A Wiener filter for nuclear medicine images.

To improve the quality of digital nuclear medicine images, we have developed a new implementation of the Wiener restoration filter. The Wiener filter uses as its optimality criterion the minimization of the mean-square error between the undistorted image of the object and the filtered image. In order to form this filter, the object and noise power spectrums are needed. The noise power spectrum for the count-dependent Poisson noise of nuclear medicine images is shown to have a constant average magnitude equal to the total count in the image. The object power spectrum is taken to be the image power spectrum minus the total count, except in the noise dominated region of the image power spectrum where a least-squares-fitted exponential is used. Processing time is kept to a clinically acceptable time frame through use of an array processor. Pronounced noise suppression and detail enhancement are noted with use of this filter with clinical images.

Implementation and testing of the dual photopeak window scatter correction method on a multi-headed SPECT system

The authors describe the implementation and testing of the dual photopeak window (DPW) scatter correction method on a multiheaded single photon emission computed tomography (SPECT) system. The DPW scatter correction method assumes that variations in the ratio of counts within two abutted energy windows which span the photopeak region is due solely to changes in the scatter fraction, and that there is a unique relationship between the ratio and the scatter fraction which is stable with time and location across the face of the cameras. A small variation in the window count ratio was observed across the camera faces, with rotation, and with time. The regression relation between window count ratio and scatter fraction was determined to have significant variation with time and location across the camera faces at large scatter fractions. The method was observed to give significant improvement in the contrast of spheres in acquisitions of a SPECT phantom. The variation in the regression relations at large scatter fractions raises concern for the quantitative accuracy of the contrast improvement when the DPW method is applied to energy windows which split the photopeak at the emission energy.<<ETX>>

Quantitative SPECT imaging: compensation for nonuniform attenuation, scatter, and detector divergence

Reconstruction methods which compensate for nonuniform attenuation, scatter, and detector divergence in SPECT (single photon emission computed tomography) imaging have been developed. The attenuation map is reconstructed from transmission data collected on one of the cameras of a three-headed SPECT system, with a line source of activity positioned at the focal line of its fan beam collimator. The other two cameras simultaneously collect the emission data. To control the noise in the attenuation map, the reconstruction algorithm employs a robust estimation technique based on an iteratively reweighted least squares criterion. The position-dependent point spread functions (PSFs) modeling the spatially variant detector divergence and scatter response are obtained from acquisitions taken of a point source of activity inside an attenuating medium. The nonuniform attenuation map and the PSFs are incorporated in the projector used for implementing a modified Chang algorithm, which also utilizes the same robust estimation technique for noise control. Using measurements taken on a physical phantom of the human torso it is demonstrated that this modified iterative Chang algorithm substantially improves the quality and quantitative accuracy of the reconstructed SPECT images of the emission densities.<<ETX>>

The Impedance Plethysmographic Sampling Field in the Human Calf

Impedance plethysmography has been widely used clinically because of its convenience and sensitivity, but its use has been questioned by some investigators because of uncertainty concerning the source of the signal. Quantitative definitions of the sensitivity to conductivity changes as a function of position have been lacking. The present study defines the sampling efficiency for impedance measurements made with four electrodes on a uniform cylinder model of the calf. Predictions based on electric field theory compare favorably with results from an in vitro model (average rms error = 8 percent). These studies demonstrate that for a given voltage electrode separation, the proximity of the current electrodes determines the relative sensitivity to conductivity changes within the sampling field. For voltage and current electrode separations of 10 and 18 cm on a 37 cm circumference calf (typical measurements for clinical testing of the lower leg), the sampling field extends about 3 cm beyond the voltage electrodes and the deep vessels are sampled with at least 80 percent of the efficiency of the superficial vessels. These studies also indicate that it is possible to predict the sampling field associated with various electrode configurations, a possibility which should aid the design of electrode configurations for specific applications of impedance plethysmography.

A fast projector backprojector pair for use in iterative reconstruction of SPECT images

Reconstructed SPECT images suffer from decreased image quality due to photon attenuation within the patient and distance-dependent blurring introduced by the collimated gamma camera. Several iterative reconstruction algorithms have been proposed to compensate for these degradations, as well as to suppress noise. These methods require the use of a realistic projector and back-projector pair which can accurately model the physics of photon transport. Here, the authors describe a projector/backprojector pair which accurately models both photon attenuation and the distance-dependent detector response, and provides a substantial decrease in computational complexity over previous methods. The computational savings are obtained by using the frequency distance principle (FDP) to incorporate distance-dependent blurring into the projector/back-projector pair. Since this procedure can be implemented using FFT methods, distance-dependent blurring can be modeled with a substantial reduction in the number of computations required. Simulation studies using a paint source object and the Hoffman bit-map brain phantom have shown that use of the FDP based projector-backprojector pair can reduce the computation time required for iterative reconstruction, with minimal loss of accuracy.<<ETX>>

Reduction of truncation artifacts in fan-beam transmission imaging using a spatially varying gamma prior

Fan beam transmission imaging used to obtain the attenuation map of an object for attenuation correction is limited because the object may not be completely in the field of view of the camera at all angles. This truncation of object data leads to the creation of artifacts in the attenuation map during reconstruction which may result in problems in the attenuation correction of the emission image. To improve the quality of the attenuation map, we have used information taken from the scatter window reconstruction of projections from a parallel hole collimator on one of the heads of the SPECT system to create a segmentation based attenuation map of the object. We use this map as a pixel by pixel gamma prior to aid in the iterative reconstruction of the final attenuation map. In the fully sampled central region, the use of the segmentation map is unnecessary; but in the poorly sampled outer region, the use of the segmentation map is essential to obtain an accurate reconstruction. We use a spatially varying weight function for the gamma prior to reflect our confidence in the transmission data in the central region while emphasizing the prior in the outer region. The result is a high quality density map over both the fully sampled and poorly sampled regions. Furthermore, since the line integrals for the transmission data are preserved in the iterative reconstruction, the algorithm does not require precise determinations of the attenuation coefficients in the segmentation maps to create an accurate attenuation map.<<ETX>>

Estimation of primary plus small angle scatter in scintigrams using two photopeak energy windows and a linear model

In scintigraphic imaging, scattered photons decrease image contrast and can lead to quantitative errors in single photon emission computed tomographic (SPECT) reconstructions. The point spread function for scattered photons can be characterized as having a narrow and a wide component. A method for estimating the wide scatter component has been developed. Subtracting this estimate from the original image yields an estimate of the primary plus the narrow scatter component. The method is based on the difference between the energy spectrum for the wide scatter component and that for primary plus the narrow scatter component. The assumptions of this method are tested using measurements of line sources at four depths. It is then applied to images of the Rollo contrast-detail phantom taken at a number of depths. Increased contrast and quantitative accuracy are obtained with no significant change in a lesion detectability index (with ratio of mean contrast to standard deviation in contrast for repeat acquisitions).<<ETX>>

An accelerated 3D maximum likelihood SPECT image reconstruction

A three-dimensional maximum likelihood estimation (MLE) algorithm for single photon emission computed tomography (SPECT) image reconstruction is developed. This expectation maximization MLE algorithm is implemented using a projector-backprojector pair which approximates the two-dimensional, position-dependent blurring in SPECT. The resulting iterative algorithm is integrated with a recently developed acceleration scheme based on vector extrapolation techniques adapted to three dimensions. Using the minimal polynomial and reduced rank vector extrapolation techniques, the accelerated 3-D MLE algorithm is evaluated on patient data from a SPECT system. It is shown that the acceleration scheme results in substantial savings in computation time.<<ETX>>