Martin Janecek

Design and Implementation of a Facility for Discovering New Scintillator Materials

We describe the design and operation of a high-throughput facility for synthesizing thousands of inorganic crystalline samples per year and evaluating them as potential scintillation detector materials. This facility includes a robotic dispenser, arrays of automated furnaces, a dual-beam X-ray generator for diffractometry and luminescence spectroscopy, a pulsed X-ray generator for time response measurements, computer-controlled sample changers, an optical spectrometer, and a network-accessible database management system that captures all synthesis and measurement data.

Simulating Scintillator Light Collection Using Measured Optical Reflectance

To accurately predict the light collection from a scintillating crystal through Monte Carlo simulations, it is crucial to know the angular distribution from the surface reflectance. Current Monte Carlo codes allow the user to set the optical reflectance to a linear combination of backscatter spike, specular spike, specular lobe, and Lambertian reflections. However, not all light distributions can be expressed in this way. In addition, the user seldom has the detailed knowledge about the surfaces that is required for accurate modeling. We have previously measured the angular distributions within BGO crystals and now incorporate these data as look-up-tables (LUTs) into modified Geant4 and GATE Monte Carlo codes. The modified codes allow the user to specify the surface treatment (ground, etched, or polished), the attached reflector (Lumirror®, Teflon®, ESR film, Tyvek®, or TiO paint), and the bonding type (air-coupled or glued). Each LUT consists of measured angular distributions with 4° by 5° resolution in theta and phi, respectively, for incidence angles from 0° to 90° degrees, in 1° -steps. We compared the new codes to the original codes by running simulations with a 3 × 10 × 30 mm3 BGO crystal coupled to a PMT. The simulations were then compared to measurements. Light output was measured by counting the photons detected by the PMT with the 3 × 10, 3 × 30, or 10 × 30 mm2 side coupled to the PMT, respectively. Our new code shows better agreement with the measured data than the current Geant4 code. The new code can also simulate reflector materials that are not pure specular or Lambertian reflectors, as was previously required. Our code is also more user friendly, as no detailed knowledge about the surfaces or light distributions is required from the user.

Optical Reflectance Measurements for Commonly Used Reflectors

When simulating light collection in scintillators, modeling the angular distribution of optical light reflectance from surfaces is very important. Since light reflectance is poorly understood, either purely specular or purely diffuse reflectance is generally assumed. In this paper we measure the optical reflectance distribution for eleven commonly used reflectors. A 440 nm, output power stabilized, un-polarized laser is shone onto a reflector at a fixed angle of incidence. The reflected lights angular distribution is measured by an array of silicon photodiodes. The photodiodes are movable to cover 2pi of solid angle. The light-induced current is, through a multiplexer, read out with a digital multimeter. A LabVIEW program controls the motion of the laser and the photodiode array, the multiplexer, and the data collection. The laser can be positioned at any angle with a position accuracy of 10 arc minutes. Each photodiode subtends 6.3deg, and the photodiode array can be positioned at any angle with up to 10 arc minute angular resolution. The dynamic range for the current measurements is 10 5:1. The measured light reflectance distribution was measured to be specular for several ESR films as well as for aluminum foil, mostly diffuse for polytetrafluoroethylene (PTFE) tape and titanium dioxide paint, and neither specular nor diffuse for Lumirrorreg, Melinexreg and Tyvekreg. Instead, a more complicated light distribution was measured for these three materials.

Reflectivity Spectra for Commonly Used Reflectors

Monte Carlo simulations play an important role in developing and evaluating the performance of radiation detection systems. To accurately model a reflector in an optical Monte Carlo simulation, the reflectors spectral response has to be known. We have measured the reflection coefficient for many commonly used reflectors for wavelengths from 250 nm to 800 nm. The reflectors were also screened for fluorescence and angular distribution changes with wavelength. The reflectors examined in this work include several polytetrafluoroethylene (PTFE) reflectors, Spectralon, GORE diffuse reflector, titanium dioxide paint, magnesium oxide, nitrocellulose filter paper, Tyvek paper, Lumirror, Melinex, ESR films, and aluminum foil. All PTFE films exhibited decreasing reflectivity with longer wavelengths due to transmission. To achieve >;0.95 reflectivity in the 380 to 500 nm range, the PTFE films have to be at least 0.5 mm thick-nitrocellulose is a good alternative if a thin diffuse reflector is needed. Several of the reflectors have sharp declines in reflectivity below a cut-off wavelength, including TiO2 (420 nm), ESR film (395 nm), nitrocellulose (330 nm), Lumirror (325 nm), and Melinex (325 nm). PTFE-like reflectors were the only examined reflectors that had reflectivity above 0.90 for wavelengths below 300 nm. Lumirror, Melinex, and ESR film exhibited fluorescence. Lumirror and Melinex are excited by wavelengths between 320 and 420 nm and have their emission peaks located at 440 nm, while ESR film is excited by wavelengths below 400 nm and the emission peak is located at 430 nm. Lumirror and Melinex also exhibited changing angular distributions with wavelength.

Optimization of a LSO-Based Detector Module for Time-of-Flight PET

We have explored methods for optimizing the timing resolution of an LSO-based detector module for a single-ring, “demonstration” time-of-flight PET camera. By maximizing the area that couples the scintillator to the PMT and minimizing the average path length that the scintillation photons travel, a single detector timing resolution of 218 ps fwhm is measured, which is considerably better than the ~385 ps fwhm obtained by commercial LSO or LYSO TOF detector modules. We explored different surface treatments (saw-cut, mechanically polished, and chemically etched) and reflector materials (Teflon tape, ESR, Lumirror, Melinex, white epoxy, and white paint), and found that for our geometry, a chemically etched surface had 5% better timing resolution than the saw-cut or mechanically polished surfaces, and while there was little dependence on the timing resolution between the various reflectors, white paint and white epoxy were a few percent better. Adding co-dopants to LSO shortened the decay time from 40 ns to ~30 ns but maintained the same or higher total light output. This increased the initial photoelectron rate and so improved the timing resolution by 15%. Using photomultiplier tubes with higher quantum efficiency (blue sensitivity index of 13.5 rather than 12) improved the timing resolution by an additional 5%. By choosing the optimum surface treatment (chemically etched), reflector (white paint), LSO composition (co-doped), and PMT (13.5 blue sensitivity index), the coincidence timing resolution of our detector module was reduced from 309 ps to 220 ps fwhm.

2D linear and iterative reconstruction algorithms for a PET-insert scanner

We are developing novel insert devices for existing whole body PET scanners to achieve better resolution in selected regions of interest such as the head, neck, breast or abdomen. The insert considered here is a full ring of high resolution detectors, which can be placed around the object of interest. Adding the insert inside the scanner leads to three different types of coincidences: insert-insert, insert-scanner and scanner-scanner. The insert-insert and scanner-scanner coincidences are similar to the coincidences obtained in a traditional PET system. The insert-scanner coincidences have an inherent fan-beam geometry for which a spatially variant system matrix is proposed. The system matrix is computed using the intersection of a fan beam with a pixel. A filtered back-projection (FBP) algorithm for the insert-scanner geometry is presented. This FBP algorithm yields images with significantly reduced artifacts compared to FBP reconstructions on insert-scanner data rebinned into parallel beams. This is demonstrated using simulated point source data acquired using SimSET. It is proposed to use a penalized ML-EM (PML-EM) algorithm using a log-cosh roughness penalty function to reconstruct a single activity distribution from all three data sets. This is demonstrated qualitatively using simulated point source data. A quantitative comparison of PML-EM and FBP was performed on data acquired from insert-scanner coincidences using a phantom with hot and cold tumors imaged in an experimental setup. The quantitative studies demonstrate that the resolution/noise tradeoff of PML-EM is improved relative to FBP.

Initial study of an asymmetric PET system dedicated to breast cancer imaging

We are developing a positron emission tomography (PET) system dedicated to breast cancer imaging and based on a pseudo-pinhole PET geometry. The system consists of a small half-ring (r=153 mm) of high-resolution LSO detectors and a large half-ring (R=413 mm) of medium-resolution LSO detectors. The coincidence detection profile between a small LSO crystal and a large LSO crystal reveals that, when the object imaged is located close to the small detector, the image resolution is primarily determined by the intrinsic spatial resolution of the small detector, and is only slightly affected by the resolution of the large detector. The dimensions of the crystals are 1.6/spl times/1.6/spl times/20 mm/sup 3/ for the small half-ring and 4.3/spl times/4.3/spl times/25 mm/sup 3/ for the large half-ring. Monte Carlo simulation of the system was performed using a modified SimSET package. The detectors in the small half ring are modeled as 2- or 4-layer detectors with DOI capability. One-mm point sources were simulated to measure the image resolution of the system at 0, 2, 6, and 10 cm offset along the +X, +Y and -Y directions. Results suggest image resolutions of 1.7-2.5 mm FWHM for breast tissues and 1.8-3.1 mm FWHM for tissues near the chest wall, including internal and axillary lymph nodes. Sensitivity of the system ranges from 5.4% to 7.8% and can be significantly increased if two planar detectors are attached to the system along the axial direction.

A simulation study for the design of a prototype insert for whole-body PET scanners

We are developing an insert device that will improve image resolution within a smaller field-of-view for clinical whole-body PET scanners. We modified SimSET (Simulation System for Emission Tomography) to simulate the insert and a PET scanner. The system consists of two detector rings. The inner ring represents an insert (r=153 mm) with high-resolution detectors using 10 mm thick LSO. The outer ring represents a PET scanner (R=413 mm) with 25 mm thick LSO. Events were binned into three sets of sinograms assuming a 2.4 and 6.75 mm crystal-pitch for the insert and the PET scanner, respectively. The detectors in the insert are modeled as 1, 2, or 4 layers with different offset configurations to evaluate the corresponding system resolution with the depth-of-interaction (DOI) effect. Results show that image resolution at 1 cm radial offset is improved from 5.6 mm full-width-at-half-maximum (FWHM) of the original PET scanner to 2.0 mm with the insert. At 12 cm offset, the resolution of the original system is 5.9 and 5.5 mm for radial and tangential directions, respectively. With the insert, the radial resolution is 5.0 mm FWHM for a single-layer detector design, but improves to 2.7 and 2.2 mm for 2and 4-layer DOI detectors, respectively. Different offsets for multi-layer detectors have negligible effect on resolution. Sensitivity of the device is, assuming the insert has a 2 cm axial extend, estimated to be 3.3%, including coincidence events from the insert-alone and insert-to-scanner sinograms. In contrast, if the insert is used as a stand-alone microPET scanner, its sensitivity is 1.3%.

Measuring Light Reflectance of BGO Crystal Surfaces

A scintillating crystals surface reflectance has to be well understood in order to accurately predict and optimize the crystals light collection through Monte Carlo simulations. In this paper, we measure the inner surface reflectance properties for BGO. The measurements include BGO crystals with a mechanically polished surface, rough-cut surface, and chemically etched surface, and with various reflectors attached, both air-coupled and with coupling compound. The measurements are performed with a laser aimed at the center of a hemispherical shaped BGO crystal. The hemispherical shape eliminates any non-perpendicular angles for light entering and exiting the crystal. The reflected light is collected with an array of photodiodes. The laser can be set at an arbitrary angle, and the photodiode array is rotated to fully cover 2pi of solid angle. The current produced in the photodiodes is readout with a digital multimeter connected through a multiplexer. The two rows of photodiodes achieve 5-degree by 4-degree resolution, and the current measurement has a dynamic range of 105:1. The acquired data was not described by the commonly assumed linear combination of specular and diffuse (Lambertian) distributions, except for a very few surfaces. Surface roughness proved to be the most important parameter when choosing crystal setup. The reflector choice was of less importance and of almost no consequence for rough-cut surfaces. Pure specular reflection distribution for all incidence angles was measured for polished surfaces with VM2000 film, while the most Lambertian distribution for any surface finish was measured for titanium dioxide paint. The distributions acquired in this paper will be used to create more accurate Monte Carlo models for light reflection distribution within BGO crystals.

A High-Speed Multi-Channel Readout for SSPM Arrays

Solid-state photomultiplier (SSPM) arrays are a new technology that shows great promise to be used in PET detector modules. To reduce the number of channels in a PET scanner, it is attractive to use resistor dividers, which multiplex the number of channels in each module down to four analog output channels. It is also attractive to have SSPMs with large pixels (3×3 or 4×4 mm2). However, large area SSPMs have correspondingly large capacitances (up to 1 nF) and directly coupling them to a resistive network will create a low-pass filter with a high RC time constant. In order to overcome this, we have developed an application specific integrated circuit (ASIC) that “hides” the intrinsic capacitance of the SSPM array from a resistive network with current buffers, significantly improving the rise time of the SSPM signals when connected to the resistive network. The ASIC is designed for a wide range of SSPM sizes, up to 1 nF (equivalent to 4×4 mm2), and for input currents of 1 to 20 mA per channel. To accommodate various sizes of SSPM pixels, the ASIC uses adjustable current sources (to keep the feedback loop stable). A test ASIC has been fabricated that has 16 input channels, an internal resistor divider array that produces four analog outputs, 16 buffers that isolate the SSPM capacitance from the resistor array, and four output buffers that can drive 100 ohm loads. Thus, detector modules based on SSPMs and this ASIC should be compatible with the block detector readout electronics found in many PET cameras. Tests of this ASIC show that its rise time is <; 2 ns (and it will thus not significantly degrade the ~7 ns rise time of the SSPM pixels) and that the analog decoding circuitry functions properly.