MicroPET

Performance Evaluation of the microPET Focus: A Third-Generation microPET Scanner Dedicated to Animal Imaging
Yuan-Chuan Tai, PhD1; J Nucl Med 2005; 46:455� 463


MicroPET P4和R4是 Concorde 公司基于UCLA的prototype microPET生产的两个型号的microPET. P4给non-human primates, R4用来给rodents.


R4使用LSO和PS-PMT技术。LSO的优点是：high stopping power, high light output and fast decay time

The detector for this system is formed from a block of LSO (19x19x10 mm3) cut 14 times to divide the block into an 8x8 crystal array with 9-mm depth cuts, such that the block is still heldtogether by a 1-mm-thick LSO layer at the bottom.

The R4 uses a 10mm-multi-clad optic bundle to couple the entire detector block to a Hamamatsu R5900-C8 PS-PMT.

In the microPET R4 model there are a total of 96 block detectors, 32 full crystal rings with 192crystals in each ring.


The Focus inherits the fundamental design and geometry of its predecessor, P4 (Primate model), with particular efforts devoted to the redesign of detector modules and pulse processing electronics. The system consists of 168 lutetium oxyorthosilicate (LSO) detectors arranged in 4 contiguous rings with a ring diameter of 25.8 cm and an axial extent of 7.6 cm. Each detector consists of a 12x12 array of LSO crystal elements coupled to a position-sensitive photomultiplier tube via an optical fiber bundle. Each LSO crystal
measures 1.51x1.51x10.00 mm3. Thin reflective material envelops the LSO crystals on all, but one, sides to improve the light collection efficiency and to provide better optical isolation between adjacent elements. The crystal pitch is 1.59 mm in both axial and transverse directions, resulting in a packing fraction of over 91% for a detector block.

The fiber optic bundle consists of 8x8 elements of square, multiclad plastic fibers each measuring 2.2x 2.2x 100.0 mm3. The same reflective material is also placed between individual fibers to provide optical isolation and to improve light collection efficiency. The use of optical fibers of cross-section greater than that of the LSO elements improves the light collection efficiency. These design changes were critical for good energy resolution as the LSO crystals have a relatively small cross-section that limits the light collection efficiency (18,19). The position-sensitive photomultiplier tube continues to be the Hamamatsu R5900-C12. A simple resistor network was used to convert the 12 anode outputs to 4 position-encoded signals (20). These 4 signals are fed into the pulse processing circuits and subsequently sent to the coincidence processor for coincidence determination. To improve the linearity of the analog-to-digital converters, each event is digitized twice and averaged for event positioning and energy determination. The new electronics doubles
the data transfer speed and effectively doubles the counting capability of the system under high counting rate situations.

The system acquires data in list mode to permit maximum flexibility in the postprocessing and reconstruction. Coincidence events can be sorted into 3-dimensional (3D) sinograms with different combinations of span and ring differences (21) or directly into 2-dimensional (2D) sinograms by single-slice rebinning (SSRB) (22). Images can be reconstructed using filtered backprojection (FBP) or ordered-subsets expectation maximization (23) (OSEM) algorithms.

The system was set to acquire data in singles mode with its energy window wide open (153� 814 keV). A 68Ge point source was placed at the center of the field of view (CFOV) to acquire a 2D position histogram of each detector. Lookup tables that map locations in the flood images into crystal identifications in LSO arrays were created using system software. The same 68Ge source was used to acquire 500 million events (

Allegro的calibration

Imaging Characteristics of a 3-Dimensional GSO Whole-Body PET Camera
Suleman Surti, PhD; and Joel S. Karp, PhD
THE JOURNAL OF NUCLEAR MEDICINE • Vol. 45 • No. 6 • June 2004

For an accurate measurement of interaction position in the detector and the deposited energy, a PMT gain matching needs to be performed to compensate for the PMT gain differences. Additionally, the collected light has a maximum of 20% difference between events at a PMT center and its edges. This is due to loss of collected light in the gaps between the hexagonally packed PMTs and small variations between light outputs of indi- vidual crystals. To achieve good system energy resolution, a position-based energy correction (scaling) is performed through a lookup table. The final data correction before sinogram binning involves identification of crystal boundaries. This must take into consideration the spatial nonlinearity of the calculated position, which is a characteristic of all Anger-logic detectors. An automated search algorithm previously developed for the G-PET (1) brain scanner has been adapted for Allegro. This algorithm detects the minima (between crystals) within a high-count flood histogram, defines the crystal boundaries around each crystal, and assigns a real position to all events that occur within each bound- ary region.

After all of these corrections have been performed, the acquired data from the scanner are binned into a sinogram with 161 angles
and 295 rays for every ring combination (total of 29^2=

Gamma Camera的linearity correction

今天看到一个patent的说明: http://www.freepatentsonline.com/20060065826.html, 里面概述了Gamma Camera的linearity correction问题.

Existing scintillation cameras experience spatial distortion that requires linearity correction (LC). The spatial distortion arises from the fact that the spatial coordinates of light events occurring either at the edges of or between adjacent photomultiplier tubes in a photodetector array will be computed differently than the coordinates of events occurring directly over the center of a photomultiplier tube, due to the physical limitations of the photomultiplier tube. A significant amount of effort has been made to developing correction schemes for spatial or linearity distortion (along with, e.g., the companion energy and flood corrections). Existing LC methods can be generally divided into two categories.

A first category is illustrated in U.S. Pat. No. 3,745,345 (the '345 patent) entitled Radiation Imaging Device, the entire disclosure of which is incorporated herein by reference. Here, a camera head is covered by a lead mask having a uniform grid of pinhole apertures. A sheet source of uniform radiation placed adjacent to the mask causes each aperture to illuminate a scintillation crystal located on the opposite side of the mask. The camera then records the detected location of events in the crystal. There is a difference between the (known) location of the pinholes and the detected location of the events as computed by the camera, which is representative of the degree of spatial distortion at the respective locations on the camera face. Accordingly, a correction factor is computed for each location point so as to move the apparent location of an event as detected to its actual location, as determined by the difference computed in the flood source calibration procedure. The correction factors are then stored in an array for later use during acquisition of clinical images.

[0007] A second category is illustrated in U.S. Pat. No. 4,212,061 entitled Radiation Signal Processing and U.S. Pat. No. 4,316,257 entitled Dynamic Modification Of Spatial Distortion Correction Capabilities Of Scintillation Camera, which pertain to spatial correction (both the '061 and '257 patents also are incorporated herein in their entirety by reference). For calibration, a lead mask having elongated slit apertures is used. The camera is exposed to a radiation source, first with the mask oriented in x lines and then with the mask oriented in y lines. For each such exposure orientation, a series of transverse peak measurements at select intervals is developed. An analytical expression is generated to represent event coordinates between calibration intervals. Each orientation exposure, thus, produces one of a pair of calibration coordinates, which in turn permit direct correspondence to associated spatial coordinates. Among other deficiencies in this method, this method can take more than one hour of time by itself. It also requires additional preparation such as `centering and gain`. Moreover, this method requires use of multiple masks wastes time and money and increases equipment downtime.

The present inventors have co-developed a new type of flood calibration mask having a much denser population of pinhole apertures in a non-uniform grid pattern, which is used in conjunction with a novel Gaussian fit algorithm to obtain a complete pinhole mask image model for LC coefficient generation. A LC coefficient represents a displacement vector of a point from its detected location in an acquired image to an ideal location.

Allegro的电子学设计

Imaging Characteristics of a 3-Dimensional GSO Whole-Body PET Camera
Suleman Surti, PhD; and Joel S. Karp, PhD
THE JOURNAL OF NUCLEAR MEDICINE • Vol. 45 • No. 6 • June 2004

The electronics architecture for the Allegro is derived from the original PENN PET design (4) with appropriate upgrades to use new advancements in technology. The output from each PMT enters an analog preamplifier channel for pulse amplification, followed by signal digitization with 50-MHz, flash (asynchronous) analog-to-digital converters. Concurrently, the analog preamplifier outputs of the PMTs are also summed into 28 overlapping trigger channels, each consisting of a group of 20 PMTs. Each trigger signal passes through a constant fraction discriminator (CFD) to obtain trigger timing information for high-energy deposition events in the detector. The overlapping triggering scheme and good system energy resolution provide the capability to raise the trigger CFD threshold as high as 400 keV, thereby reducing the scanner dead time significantly. The trigger signals are formed before PMT gain matching and, since the gains can vary by as much as a factor of 4, the CFD threshold is set lower than the final software energy lower level discriminator (ELLD) that sets a gate for events that are histogrammed in the sinogram. With a CFD threshold of around 400 keV, we are able to achieve uniform event distribution over the entire scanner area. The CFD signals are then checked for coincident events and, once a coincidence is detected (within a timing window, 2t

PMT的大小和Light Spread

文献:A count-rate model for PET scanners using pixelated  Anger-logic detectors with different scintillators
S Surti and J S Karp

In a large continuous scintillator, the spreading of the scintillation photons depends on the thickness of the lightguide as well as the scintillator.

scintillator受激后产生的光子会在scintillator和lightguide里面扩散, 扩散的范围取决于它们的厚度.

However, the scintillator thickness also determines the sensitivity of the detector which is fixed by the scanner requirements. As a result, there is not enough control over reducing the scintillation light spread within the detector by using thinner crystals without compromising the detector sensitivity.

scintillator的厚度决定了detector的sensitivity, 所以不能只靠减少crystal的厚度来降低scintillation light spread

Depending upon the light spread within the detector which increases with the scintillator thickness, large PMTs are needed to achieve optimal sampling of scintillation light and thus attain good spatial resolution. Using many small PMTs degrades spatial resolution due to the Poisson noise in each PMT signal, while using fewer small PMTs will lead to a reduced sensitivity of the positioning algorithm to detect small changes in interaction position.

However, increased light spread and the use of large PMTs lead to increased pulse pileup or deadtime in the detectors at high count rates. Thus, in order to achieve high sensitivity and good spatial resolution in a continuous Anger-logic detector, there is a drawback due to increased pulse pileup at high count rates.

NaI(Tl)-based PET scanners like the C-PET (Philips Medical Systems) use large 1 inch thick NaI(Tl) crystals coupled to an array of 63 mm diameter PMTs via a lightguide. The crystal and lightguide thickness are well optimized to the PMT size to derive the best achievable spatial resolution and count-rate performance from these detectors.

下面是他们的G-PET(与A-PET及Mosaic接近)的设计:

Each of the GSO crystals is optically isolated from its neighbours by using a high reflectivity material such as PTFE which will restrict the scintillation photons from entering adjacent crystals. The only place where these photons can spread is in the lightguide. Essentially, with such a design we can control the light spread through the lightguide without changing the detector sensitivity. The lightguide thickness can now be adjusted to vary the light spread within the detector and thus match it to a given PMT size , thus providing the flexibility to achieve minimal pulse pileup for any PMT size.

Calibration of Mosaic System

Preamp and Sumamp Offsets Time required: 15-30 minutes
Preamp offset calibration adjusts each PMT's preamp offset to get a specific baseline value. The sumamp offset calibration adjusts each trigger channel's sumamp offset to maximize the counts out of the trigger channel. Preamp & Sumamp Offset calibration can be performed independently of all other calibrations.

Automated Gain (preliminary) Time required: 4 hours
This procedure is normally performed at the factory only.

CFD Threshold Time required: 15-60 minutes
Constant Fraction Discriminator (CFD) Threshold calibration sets the threshold at the minimum energy required to trigger an event. Threshold calibration ensures that the CFDs gate out events at the proper point relative to the photopeak. Before running CFD threshold calibration a Preamp and Sumamp Offset calibration must be run.

Automated Gain (Final) Time required: 4-15 hours
Gain calibration adjusts the gain for each PMT, such that the output is approximately the same for all PMTs. Before running gain calibration, the following calibrations must be run: Preamp and Sumamp Offset, CFD Threshold, and Coincidence Timing. If this is not a new scanner or installation, after Gain calibration, the system linearity must be checked. If the system linearity does not pass, a Distortion removal (gdr) calibration must be run. If this is a new scanner or installation, Distortion removal (gdr) calibration must be run after the
Gain calibration. After running Gain calibration, the following calibrations must be run: Energy Correction, EC mock scans, Blank scans and Normalization.

Coincidence Timing Time required: 30-90 minutes
Coincidence Timing calibration adjusts the timestamp Least Significant Bit (LSB) and then all of the trigger channel timing delays to align all pulses. Before running Coincidence Timing calibration, the following calibrations must be run: Preamp and Sumamp Offset and CFD Threshold.

Distortion Removal (gdr) Time required: 28 hours - Data Collection 10 hours - Calculation 2 hours - Creve Process 16 hours
The GSO distortion removal (gdr) calibration is a spatial calibration that converts measured event positions to real, physical positions on the face of the detector. Before running distortion removal calibration, the following calibrations must be run: Preamp and Sumamp Offset, CFD Threshold, Coincidence Timing, and Automated Gain (final). After running distortion removal, the following calibrations must be run: Energy Correction, EC mock scans, Blank scans, and Normalization.

Energy Correction Time required: 15-30 minutes
Energy Correction allows the Global and Local Energy curves to be centered at a specified energy and curve shape to be made narrow about this center. Before running Energy Correction, the following calibrations must be run: Preamp and Sumamp Offset and CFD Threshold and Distortion removal (gdr). If the autoQC Energy Test gives warnings or fails, an Energy Correction will automatically be run.

Blank Scans Time required: 2 hours (2 scans)
Blank scans are transmission sinograms collected with nothing in the scanner field of view (the transmission source must be in the source holder). As part of the blank scan calibration, blank EC (emission contamination) sinograms are also collected. These blank EC sinograms are called leakage sinograms.

Time required: 8 hour collection (+1-2 hours to process and test)
Normalization is essentially independent of timing adjustments, as it is performed at a low enough rate to preclude randoms.

Gain changes affect normalization in two ways:
Spatial - Mispositioning of events moves counts from one part of the detector to another.
Sensitivity - As the gain of a PMT changes, the energy of events near that PMT also changes.

This can move counts out of the energy windows, leading to a local change in sensitivity. Energy correction should compensate for the sensitivity component of PMT gain changes.

一个TOF项目

http://www.nuclear.kth.se/diploma/PetProject/intro.html

Development of a Time-of-Flight and 3D Demonstration Set-up for Positron Emission Tomography

Mikael Steén and Per Uhlén

Abstract

This report describes the development of a time-of-flight and 3D demonstration set-up for positron emission tomography (PET). The detector system consists of 48 barium fluoride (BaF2) scintillators arranged in a ring.

In the initial phase of the development, measurements of the characteristics of two detector prototypes differing in the shapes of the scintillators, were performed. The measurements were then complemented by Monte-Carlo simulations of the entire set-up, in which the measured energy and time response of the detector prototypes were included. The data from the simulations where then processed by conventional image reconstruction algorithms. The results formed the basis for the overall design of the detector system.

The set-up will be used in the undergraduate as well as graduate student training program in Physics at the Royal Institute of Technology. Therefore, this report contains a general overview of the physics and instrumentation associated with positron emission tomography.

---

• 1 Introduction
• 1.1 Basic Physics in PET
• 1.1.1 Beta Decay
• 1.1.2 Annihilation Radiation
• 1.1.3 Interactions of Gamma Radiation with Matter
• 1.1.3.1 The Interaction Cross-Section
• 1.1.3.2 The photoelectric effect
• 1.1.3.3 Compton Scattering
• 1.1.3.4 The Stopping Power
• 1.1.3.5 The Concept of Resolution
• 1.1.3.6 Detector Efficiency
• 1.1.3.7 Scintillation Detectors
• 1.1.3.8 Photomultiplier Tubes
• 1.2 The Principles of PET and TOFPET
• 1.2.1 Basic Function
• 1.2.2 Coincidence Detection
• 1.2.3 PET versus TOFPET
• 1.2.4 Multi-Ring Scanner
• 1.2.5 Intrinsic Limitations
• 1.2.5.1 Positron Range
• 1.2.5.2 Noncollinearity
• 1.2.5.3 Statistical Limitations
• 1.2.5.4 Geometrical Limitations
• 1.3 Clinical PET Systems
• 2 Performance Measurements
• 2.1 Measurement Electronics
• 2.1.1 Constant Fraction Discriminator
• 2.1.2 Time to Pulse Height Converter
• 2.1.3 Multichannel Analyser
• 2.2 Results
• 2.2.1 Energy Resolution
• 2.2.2 Time Resolution
• 2.2.3 Efficiency
• 3 Simulations
• 3.1 GEANT
• 3.2 Simulations of the TOFPET System - General Approach
• 3.3 Simulation of Measured Efficiency in Singles Mode
• 3.4 Simulations of Ring Performance
• 3.4.1 Expected Spatial Resolution
• 3.5 The Steén Phantom
• 4 Image Processing
• 4.1 Raw Data
• 4.2 2D Image Reconstruction
• 4.3 Forward Projection
• 4.4 Reconstruction from Parallel Projections -The Central Slice Theorem
• 4.4.1 Back Projection
• 4.4.2 Filtered Back Projection
• 4.5 3D Image Reconstruction
• 4.6 Additional Notes on Image Reconstruction
• 4.6.1 Iterative Reconstruction Methods
• 4.6.2 Shannon's Sampling Theorem
• 4.7 Samples of Images
• 4.7.1 2D PET and TOFPET Images
• 4.7.2 3D TOFPET Images
• 5 Design
• 5.1 The Detectors
• 5.2 The Mechanical Support Structure
• 5.3 Electronics and Data Acquisition
• 6 Conclusions
• 7 Acknowledgements
• 8 References
• Appendix One MATLAB Code for Image Processing
• Appendix Two FORTRAN Code for GEANT Simulations

一个系统讲解PET Physics的网页

http://depts.washington.edu/nucmed/IRL/pet_intro/toc.html

Introduction to PET Physics

---

[Table of contents] [List of figures] [List of tables]
[References] [List of abbreviations] [Copyright notice]

 

Table of Contents

1. Introduction

2. The physical principles of PET

2.1 Introduction
2.2 Positron emission and annihilation
2.3 Coincidence detection and electronic collimation
2.4 Photon interactions in human tissue and correction for gamma-ray attenuation
2.5 Types of coincidence events

3. 2D mode and 3D mode

3.1 Principles of operation
3.2 Sensitivity to true coincidence events
3.3 Sensitivity to scattered events
3.4 Sensitivity to random events
3.5 Effect of camera geometry

4. Image reconstruction

4.1 Introduction
4.2 Notation and mathematical theorems used
4.3 Analytic image formation in 2D PET
4.4 Filtered Back-Projection in 3D and 3D-RP

5. Detection systems in PET

5.1 Introduction
5.2 Scintillators and scintillation detectors
5.3 Pulse processing
5.4 Coincidence processing
5.5 Dead-time
5.6 Block detectors
5.7 Camera configurations in PET

6. Corrections for quantitative PET in 2D and 3D mode

6.1 Introduction
6.2 Attenuation correction
6.3 Correction for random coincidences
6.4 Scatter correction
6.5 Detector normalisation
6.6 Dead-time correction

---

[top of page

Scintillators and scintillation detectors

The scintillation process involves the conversion of high-energy photons into visible light via interaction with a scintillating material, and consists of the following steps:

1) a photon incident on the scintillator creates an energetic electron, either by Compton scatter or by photoelectric absorption.

2) as the electron passes through the scintillator, it loses energy and excites other electrons in the process.

3) these excited electrons decay back to their ground state, giving off light as they do so.

In a scintillation detector, the scintillator is optically coupled to a photomultiplier tube (PMT) which then generates an electrical signal in response to light incident upon its face. There are several variations on this theme which are used in PET - for example, a 2D array of crystals may be coupled to 4 PMTs (the block-detector, Casey and Nutt, 1986), an array of PMTs may be coupled to a single planar crystal (the Anger camera,Anger, 1958), or an array of crystals may be coupled to a multi-channel PMT (Cherry et al 1997).



When real scintillation detectors are exposed to mono-energetic photons, the energy measured is not that of the electron generated by the initial interaction, but rather the total energy deposited by the photon in the detector. This distinction is important because photons initially interacting by Compton scatter may subsequently be involved in further interactions within the detector. In a sufficiently large detector, most Compton-scattered photons will eventually deposit all their energy, and most events will register in the photon energy peak. Under these circumstances, this feature of the energy distribution is better described by the term "full-energy peak" rather than "photo-peak".

http://depts.washington.edu/nucmed/IRL/pet_intro/intro_src/section5.html

一个TOFPET的讲稿

其中关于CFD(Constant Fraction Discriminator)一段讲得比较清楚.

The CFD is designed to trigger on a certain optimum fraction of the pulse height, thus making the performance of the CFD independent of pulse amplitude.

TOFPET2007.pdf