M. Rafecas

Characterization and readout of MADPET-II detector modules: validation of a unique design concept for high resolution small animal PET

MADPET-II is a novel high-resolution, high-sensitivity three-dimensional lutetium oxyorthosilicate avalanche photodiode small animal PET scanner with a unique detector design and readout that creates new possibilities for data processing and analysis. The scanner consists of a ring of dual-layer detector modules (O71 mm), each containing a 4/spl times/8 array of 2/spl times/2/spl times/6mm/sup 3/ (front) and 2/spl times/2/spl times/8 mm/sup 3/ (back) LSO crystals that are each optically isolated and coupled one-to-one to a monolithic 4/spl times/8 APD array. Signals from each channel are individually processed using fully-integrated 16-channel, low noise, charge sensitive preamplifiers and custom integrated electronics that include a four channel shaping amplifier, peak detector and non-delay line CFD. Analog-to-digital converters and time-to-digital converters digitise the pulse height and time-stamp each event, and data are stored exclusively in list mode format. Two 32-channel detector blocks were tested in a coincidence and in a dual-layer configuration. Coincidences were sorted post-acquisition in software, allowing for maximum flexibility in the data processing. Monte Carlo simulations have calculated the system spatial resolution to be 1.1 mm FWHM. Average energy resolution for a single detector block ranged from 20.2% FWHM to 23.9% FWHM. System pulse height linearity was measured, and all channels responded with R/sup 2/>0.9988 (Pearson correlation coefficient). Intrinsic uniformity for each detector block ranged from 2.0% to 4.8% (400 keV threshold). Overall system timing resolution was measured to be 10.2 ns FWHM, with individual LORs having time resolutions as low as 4.5 ns FWHM.

A 4/spl times/8 APD array, consisting of two monolithic silicon wafers, coupled to a 32-channel LSO matrix for high-resolution PET

Avalanche photodiodes (APDs) offer new design options for high-resolution positron emission tomography (PET) detectors. The goal of this study was the design of a very compact PET detector module with two layers. For this purpose, new monolithic arrays of 4/spl times/8 densely packed APDs (each with a 1.6/spl times/1.6 mm/sup 2/ sensitive area) were developed and evaluated for their suitability to read out 32 small (2/spl times/2/spl times/6 mm/sup 3/) lutetium oxyorthosilicate (LSO) crystals. The gain of the APDs was about 100, its standard deviation within an array was less than 12%. The average dark current was less than 45 nA per pixel, and the detector capacitance was about 12 pF. Energy resolution was 15% for gamma-rays with 511 keV. The time resolution was 2.5 ns (full-width at half-maximum). A high-reflectivity foil was used to hold all 32 individual crystals in a matrix. This matrix was glued directly on the APD array. The light crosstalk between the channels was only 7.9%. Together with a specially developed 16-channel low-noise JFET-CMOS preamplifier chip, a compact front-end detector module was built. The monolithic APD arrays proved to be stable and may, in combination with lutetium oxyorthosilicate (LSO) crystals, facilitate the design of various detector geometries, such as panels or multilayer modules.

Joint Reconstruction of Image and Motion in Gated Positron Emission Tomography

We present a novel intrinsic method for joint reconstruction of both image and motion in positron emission tomography (PET). Intrinsic motion compensation methods exclusively work on the measured data, without any external motion measurements. Most of these methods separate image from motion estimation: They use deformable image registration/optical flow techniques in order to estimate the motion from individually reconstructed gates. Then, the image is estimated based on this motion information. With these methods, a main problem lies in the motion estimation step, which is based on the noisy gated frames. The more noise is present, the more inaccurate the image registration becomes. As we show both visually and quantitatively, joint reconstruction using a simple deformation field motion model can compete with state-of-the-art image registration methods which use robust multilevel B-spline motion models.

Effect of noise in the probability matrix used for statistical reconstruction of PET data

Iterative reconstruction algorithms require prior computation of the system probability matrix P. This matrix is usually estimated from approximated calculations. The approach employed to determine P in this work, however, was based on Monte Carlo simulations. While this technique allows P to be described more accurately, the number of simulated events may limit the statistical quality of P, thus affecting the reconstructed image. The goal of this study was to quantify this effect for OSEM and PWLS applied to the small animal PET system MADPET (/spl phi/=86 mm). System matrices with different statistical quality were obtained by using subsets of the simulated data, and these were then used to reconstruct two simulated phantoms. The results showed that simulations with more than /spl ap/20 000 detected coincidences per pixel barely improved the accuracy of P, but when less than 4 000 detected coincidences per pixel were used, the statistical quality of P deteriorated strongly. PWLS was more sensitive to the inaccurate description of P than OSEM. For PWLS and 1 mm/sup 2/ pixels, any slight increase of the mean relative error for P in the range 23%-30% strongly affected the image properties, while OSEM in combination with any matrix characterized by a mean relative error below /spl ap/40% (obtained from simulations with more than a mean of /spl ap/680 detected counts per pixel) resulted in reasonable images. Good SNR and contrast was assured when using OSEM and a matrix characterized by a mean relative error below 25% (at least 7 000 detected coincidences per pixel). Discarding elements of P that had very small magnitudes reduced the size of the matrix (storage of nonzero elements) and improved the relative error of P and signal-to-noise ratio, especially if OSEM was employed.

Geometrical and Monte Carlo projectors in 3D PET reconstruction

PURPOSE In the present work, the authors compare geometrical and Monte Carlo projectors in detail. The geometrical projectors considered were the conventional geometrical Siddon ray-tracer (S-RT) and the orthogonal distance-based ray-tracer (OD-RT), based on computing the orthogonal distance from the center of image voxel to the line-of-response. A comparison of these geometrical projectors was performed using different point spread function (PSF) models. The Monte Carlo-based method under consideration involves an extensive model of the system response matrix based on Monte Carlo simulations and is computed off-line and stored on disk. METHODS Comparisons were performed using simulated and experimental data of the commercial small animal PET scanner rPET. RESULTS The results demonstrate that the orthogonal distance-based ray-tracer and Siddon ray-tracer using PSF image-space convolutions yield better images in terms of contrast and spatial resolution than those obtained after using the conventional method and the multiray-based S-RT. Furthermore, the Monte Carlo-based method yields slight improvements in terms of contrast and spatial resolution with respect to these geometrical projectors. CONCLUSIONS The orthogonal distance-based ray-tracer and Siddon ray-tracer using PSF image-space convolutions represent satisfactory alternatives to factorizing the system matrix or to the conventional on-the-fly ray-tracing methods for list-mode reconstruction, where an extensive modeling based on Monte Carlo simulations is unfeasible.

A Monte Carlo study of high resolution PET with granulated dual layer detectors

New arrays of Avalanche Photodiodes (APD) allow the design of novel highly granulated detector modules. Monte Carlo simulations were used to evaluate to what extent this feature can be used for high resolution, high sensitivity PET. Based on a fixed crystal front face of 2 mm/sup 2/ and a fixed number of crystals, sensitivity and scatter fraction for three different geometries were determined: (a) Ring with 143 mm diameter; (b) Ring with only 71 mm diameter but double the axial extent (37 mm); (c) Ring with 71 mm diameter and two radial crystal layers. The sensitivity (a:b:c) was 0.3%:1.1%:1.5% for a line source in air. Studies using a simple mouse-like phantom showed the highest scatter fraction for (b) and comparable sensitivities for (b) and (c). The large diameter of (a) reduced the scatter fraction at the expenses of high sensitivity losses. Line source simulations showed a resolution of about 1.6 mm for (c) at the center for the field of view (FOV). Within a region of 20 mm within the FOV, the resolution of (c) remained close to 2 mm. Geometry (c) is being implemented in the new tomograph MADPET.

Comparison of basis functions for 3D PET reconstruction using a Monte Carlo system matrix

In emission tomography, iterative statistical methods are accepted as the reconstruction algorithms that achieve the best image quality. The accuracy of these methods relies partly on the quality of the system response matrix (SRM) that characterizes the scanner. The more physical phenomena included in the SRM, the higher the SRM quality, and therefore higher image quality is obtained from the reconstruction process. High-resolution small animal scanners contain as many as 10³-10⁴ small crystal pairs, while the field of view (FOV) is divided into hundreds of thousands of small voxels. These two characteristics have a significant impact on the number of elements to be calculated in the SRM. Monte Carlo (MC) methods have gained popularity as a way of calculating the SRM, due to the increased accuracy achievable, at the cost of introducing some statistical noise and long simulation times. In the work presented here the SRM is calculated using MC methods exploiting the cylindrical symmetries of the scanner, significantly reducing the simulation time necessary to calculate a high statistical quality SRM and the storage space necessary. The use of cylindrical symmetries makes polar voxels a convenient basis function. Alternatively, spherically symmetric basis functions result in improved noise properties compared to cubic and polar basis functions. The quality of reconstructed images using polar voxels, spherically symmetric basis functions on a polar grid, cubic voxels and post-reconstruction filtered polar and cubic voxels is compared from a noise and spatial resolution perspective. This study demonstrates that polar voxels perform as well as cubic voxels, reducing the simulation time necessary to calculate the SRM and the disk space necessary to store it. Results showed that spherically symmetric functions outperform polar and cubic basis functions in terms of noise properties, at the cost of slightly degraded spatial resolution, larger SRM file size and longer reconstruction times. However, we demonstrate that post-reconstruction smoothing, usually applied in emission imaging to reduce the level of noise, can produce a spatial resolution degradation of ~50%, while spherically symmetric basis functions produce a degradation of only ~6%, compared to polar and cubic voxels, at the same noise level. Therefore, the image quality trade-off obtained with blobs is higher than that obtained with cubic or polar voxels.

Effect of inter-crystal scatter on estimation methods for random coincidences and subsequent correction

Random coincidences can contribute substantially to the background in positron emission tomography (PET). Several estimation methods are being used for correcting them. The goal of this study was to investigate the validity of techniques for random coincidence estimation, with various low-energy thresholds (LETs). Simulated singles list-mode data of the MADPET-II small animal PET scanner were used as input. The simulations have been performed using the GATE simulation toolkit. Several sources with different geometries have been employed. We evaluated the number of random events using three methods: delayed window (DW), singles rate (SR) and time histogram fitting (TH). Since the GATE simulations allow random and true coincidences to be distinguished, a comparison between the number of random coincidences estimated using the standard methods and the number obtained using GATE was performed. An overestimation in the number of random events was observed using the DW and SR methods. This overestimation decreases for LETs higher than 255 keV. It is additionally reduced when the single events which have undergone a Compton interaction in crystals before being detected are removed from the data. These two observations lead us to infer that the overestimation is due to inter-crystal scatter. The effect of this mismatch in the reconstructed images is important for quantification because it leads to an underestimation of activity. This was shown using a hot-cold-background source with 3.7 MBq total activity in the background region and a 1.59 MBq total activity in the hot region. For both 200 keV and 400 keV LET, an overestimation of random coincidences for the DW and SR methods was observed, resulting in approximately 1.5% or more (at 200 keV LET: 1.7% for DW and 7% for SR) and less than 1% (at 400 keV LET: both methods) underestimation of activity within the background region. In almost all cases, images obtained by compensating for random events in the reconstruction algorithm were better in terms of quantification than the images made with precorrected data.

Integrated low noise, low power, fast charge-sensitive preamplifier for avalanche photodiodes in JFET-CMOS-technology

To take advantage on the compactness of APD arrays, low noise, power efficient, fast charge sensitive preamplifier chips with differential current drivers have been developed. A 16-channel and a single channel version are available. The chips were adapted for low capacitance 4/spl times/8 APD arrays produced by Hamamatsu, Japan. A mixed JFET-CMOS production process yielded high quality integrated JFETs for the input stage of the amplifiers folded cascode. Thus, the 1/f-noise corner is kept at 4 kHz. The JFET has a transconductance of 11 mS at a drain current of 3 mA. The serial noise of the input transistor was found to be 0.8 nV//spl radic/Hz. The signal rise-time of the driver outputs is 20 ns. The rms noise of the preamplifier was found to be 480 e/sup -/ with a 25 e/sup -//pF noise slope for a shaping time of 50 ns. The serial input noise of the preamplifier is about 1.7 nV//spl radic/Hz from 200 kHz up to 40 MHz and the 1/f-noise corner is at 70 kHz. The power consumption is 30 mW per preamplifier, including the differential driver. The linearity is better than 1.3% over 48 dB dynamic range. For the 16 channel chip, the gain variation is less than 3.5%. Performance similar to PMTs can be achieved with APDs in combination with this integrated preamplifier chip.

Implementation of Monte Carlo coincident aperture functions in image generation of a high-resolution animal positron tomograph

For imaging animals, small-diameter tomographs are designed for increased sensitivity. As a consequence, it is optimal to reconstruct a large fraction of the detector aperture. This is a challenge if resolution degradation, especially at the edge of the field-of-view, is to be minimized. Accounting for system-dependent resolution functions could therefore greatly enhance the image generation process in these tomographs. Monte Carlo techniques were used to simulate the high-resolution animal tomograph MADPET and to determine coincidence aperture functions (CAFs) as a function of position in field-of-view. Sinogram rebinning was applied with spatially varying CAFs followed by filtered backprojection. An analytical rebinning method based on the intrinsic resolution of the system was used for comparison. Furthermore, CAFs were implemented into iterative list-mode reconstruction without sinogram rebinning. Simulated and measured line sources at positions within 88% of the system diameter were analyzed to determine image resolution. Image quality was assessed with a simulated and measured Derenzo-style structure phantom. Since no deconvolution methods were used for sinogram rebinning with simulated CAFs, this method could not improve resolution degradation. Nevertheless, it provided enhanced image quality by removing sampling artifacts introduced by common rebinning techniques. The use of Monte Carlo derived probability matrices combined with iterative list-mode reconstruction proved to be adequate to improve image quality and restore spatial resolution within 88% of the detector ring aperture in a small animal positron tomograph.