Timing calibration using internal radiation and external radiation source in time of flight positron emission tomography

Abstract

A method and system for providing improved timing calibration information for use with apparatuses performing Time of Flight Positron Emission Tomography scans. Relative timing offset, including timing walk, within a set of processing units in the scanner are obtained and corrected using a stationary limited extent positron-emitting source, and timing offset between the set of processing units is calibrated using an internal radiation source, for performing calibration.

Claims

1. A method of performing timing calibration in time of flight (TOF) positron emission tomography (PET), comprising: obtaining relative timing offset within each of a plurality of sets of detector units by placing a limited extent annihilation radiation source in a field of view (FOV) of a PET scanner, each set of the plurality of sets of detector units having more than two detector units; correcting the relative timing offset within each of the plurality of sets of detector units; calibrating a timing offset between the plurality of sets of detector units using an internal radiation; and determining a total timing offset as a sum of the corrected relative timing offset within each of the plurality of sets of detector units and the calibrated timing offset between the plurality of sets of detector units.

2. The method according to claim 1, wherein the step of correcting the relative timing offset includes correcting a timing walk.

3. The method according to claim 2, wherein the step of correcting the timing walk includes non-linear timing walk correction.

4. The method according to claim 1, wherein the internal radiation is radiation that results from decay of radioactive material that is part of a scintillator array of the PET scanner.

5. The method according to claim 4, wherein a decay process of the internal radiation includes at least two nearly simultaneous emissions from which coincidence events are formed.

6. The method according to claim 4, wherein a decay process of the internal radiation includes an emission from which coincidence events can be formed from Compton scattering in detectors caused by the emission.

7. The method according to claim 1, wherein the internal radiation is present in at least one of the scintillator, an adhesive holding a reflector in place, the reflector itself, and a detector housing.

8. The method according to claim 1, wherein the internal radiation is Lu-176 or Co-60.

9. The method according to claim 1, wherein the limited extent annihilation radiation source comprises a limited extent source with an extent so that each crystal of the scanner is coupled to many crystals in a particular set of detector units other than the plurality of sets of detector units.

10. The method according to claim 1, wherein the limited extent annihilation radiation source has a narrowest cross-sectional extent of less than 10 mm.

11. The method according to claim 1, wherein the limited extent annihilation radiation source is a line source.

12. The method according to claim 11, wherein the limited extent annihilation radiation source is a positron emitting source.

13. The method according to claim 1, wherein the relative timing offset within each of the plurality of sets of detector units is calculated using neural networks.

14. The method according to claim 1, wherein the limited extent annihilation radiation source is at least one of a Ge-68 line source, a F18-FDG line source or a Na-22 line source.

15. An imaging time of flight (TOF) positron emission tomography (PET) system, comprising: a limited extent annihilation radiation source arranged in an imaging region of the imaging system; a detector configured to detect coincident event pairs resulting from annihilation of positrons; and circuitry configured to perform timing calibration of the TOF PET system by obtaining relative timing offset within each of a plurality of set of detector units via the limited extent annihilation radiation source in a field of view (FOV) of the TOF PET scanner, each set of the plurality of sets of detector units having more than two detector units; correcting the relative timing offset within each of the plurality of sets of detector units; calibrating a timing offset between the plurality of sets of detector units using an internal radiation; and determining a total timing offset as a sum of the corrected relative timing offset within each of the plurality of sets of detector units and the calibrated timing offset between the plurality of sets of detector units.

16. The TOF PET system according to claim 15, wherein the correcting the relative timing offset includes correcting a timing walk, which includes a non-linear timing walk correction.

17. The TOF PET system according to claim 15, wherein the circuitry is further configured to obtain a portion of data from the internal radiation and the limited extent annihilation radiation source separately.

18. The TOF PET system according to claim 15, wherein the circuitry is further configured to obtain a portion of data from the internal radiation and the limited extent annihilation radiation source simultaneously.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The application will be better understood in light of the description which is given in a non-limiting manner, accompanied by the attached drawings in which:

(2) FIG. 1 shows a schematic of a PET scanner.

(3) FIG. 2A shows a schematic of an example position of an external source relative to opposite detectors.

(4) FIG. 2B shows a TOF difference histogram between the two detectors of FIG. 2A.

(5) FIG. 3 shows a schematic of another example position of a pair of external sources relative to opposite detectors.

(6) FIG. 4 shows a schematic of another example position of an external source relative to opposite detectors.

(7) FIG. 5 shows a schematic of another example position of an external source relative to opposite detectors.

(8) FIGS. 6A-6D show schematics of various embodiments of external positron sources within the detector ring.) FIGS. 7A-7B show schematics of various embodiments of internal sources within the detector ring.

(9) FIG. 8A shows an internal radiation timing distribution before applying the relative offset correction obtained from the external source.

(10) FIG. 8B shows the distribution for the same data after applying the timing corrections (including non-linear walk correction).

(11) FIG. 9 shows a schematic of an example layout of different stages of timing calibration.

(12) FIG. 10 shows a schematic of another example layout of different stages of timing calibration.

(13) FIG. 11 shows the leading-edge discriminator used in TOF PET systems.

(14) FIGS. 12A and 12B show a TOF difference histogram between DU 14 and DU 33 from Lutetium background radiation data. Timing center is found by applying parabola fitting to the peak region.

(15) FIG. 13 shows a TOF difference histogram between DU 14 and DU 33 from Lutetium background radiation data.

(16) FIG. 14 shows a TOF difference histogram between DU 14 and DU 33 from Ge line source and background radiation data.

(17) FIG. 15A shows timing offset between DU pairs before quick timing calibration.

(18) FIG. 15B shows timing offset between DU pairs after quick timing calibration.

(19) FIG. 16 shows the difference between the counts available when limiting counts by applying an energy window to avoid walk effects and when accepting all counts by performing walk correction.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

(20) Reference throughout this specification to “one embodiment” or “an embodiment” means that a particular feature, structure, material, or characteristic described in connection with the embodiment is included in at least one embodiment of the application, but do not denote that they are present in every embodiment.

(21) Thus, the appearances of the phrases “in one embodiment” or “in an embodiment” in various places throughout this specification are not necessarily referring to the same embodiment of the application. Furthermore, the particular features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments.

(22) A PET scanner in the present embodiments may have different electronics architectures. Non-limiting example layouts are shown in FIG. 9 and FIG. 10, where:

(23) PET scanner: a whole scanner, usually in the form of a ring.

(24) Region: a relatively large part of the scanner, such as a quadrant, which consists of advanced data processing, data transfer, clock control, signal processing, etc. A scanner might have several regions. Timing offset/drift due to clock distribution could be on a region basis.

(25) Detector unit (DU): a relatively isolated module, which consists of data transfer, clock control, signal processing, etc. A region might have 10-20 DUs. Timing offset/drift due to clock distribution could be on a DU basis.

(26) Board: an electronics board, which consists of signal processing circuitry for a number of channels. A DU might have 5-20 boards. Timing offset/drift due to power supply could be on a board basis.

(27) ASIC: the smallest signal processing unit, which usually consists of one timing processing channel and several energy processing channels. A board might have 1-10 application specific integrated circuits (ASICs). Timing offset/drift could be on an ASIC basis.

(28) Crystal: the smallest element in the scanner. An ASIC might perform signal processing for several tens of crystals.

(29) Timing calibration is usually done at different stages. The number of elements at each stage of the electronics architecture is scaled by approximately an order-of-magnitude. For the same statistical uncertainty, the acquisition time and analysis time for calibrating the timing offset at each stage varies significantly. For example, a DU might contain 500 to 1000 crystals, so calibrating the DU offset requires approximately √(500 to 1000) or 20-30 times shorter acquisition time than calibrating the crystal offset (and this ignores computation time).

(30) However, timing calibration might only need to be done at certain stages. After initial timing calibration, timing drift could happen at certain stages depending on the cause of the timing drift. Then, maintenance timing calibration only needs to be performed for stages where timing drift occurs, so it could be made much faster.

(31) Timing walk can be corrected by including an energy-dependent term in the offset correction. Correcting the timing walk results in better timing resolution (i.e. a narrower distribution of measured timing differences). Typically in PET, imaging is performed only using detected gamma rays in a narrow window around 511 keV. When only events in a narrow energy window are used, a linear walk correction (i.e. a walk-correction that only depends linearly on the energy) is usually sufficient. For example, an offset including a linear walk correction can be written as
t.sub.offset=t.sub.offset(E=511)+W.sub.1(E−511)
where W.sub.1 is the linear walk-correction coefficient. The walk-correction can be expanded to include non-linear terms, such as
t.sub.offset=t.sub.offset(E=511)+W.sub.1(E−511)+W.sub.2(E−511)+ . . . +W.sub.n(E−511).sup.n
where W.sub.1 through W.sub.n are walk correction coefficients.

(32) Here, since the method presented herein uses events across a very wide energy range for the “internal radiation” portion of the calibration (to reduce the total acquisition time to a practical range), a non-linear walk correction provides substantial improvement in performance. Since any function that would be used for walk correction can be represented by a Taylor series expansion, this is equivalent to toffset=f(E) where f is a function, which may be non-linear.

(33) During initial full timing calibration, the timing offset per crystal and timing walk per crystal are calibrated. The line source should be thick enough that each row of crystals is coupled to more than one row of crystals in the opposite DU. Data with positron-emitting line source and internal radiation could be acquired separately or simultaneously.

(34) In one embodiment, data with positron-emitting line source and internal radiation could be acquired separately. The positron-emitting source may be at least one of a Ge-68 line source, a F18-FDG line source or a Na-22 line source.

(35) In particular, during data acquisition with positron-emitting line source, the disclosed method: places the positron-emitting line source at the center of the scanner field of view (FOV); and acquires coincidence data with the positron-emitting line source. Standard clinical data acquisition FOV and coincidence timing window could be used. The number of coincidence events with positron-emitting line source needs to be enough to calibrate the peak position from the TOF difference histogram for each of the crystals.

(36) During data acquisition with internal radiation, the disclosed method: removes all radiation source from the scanner; and acquires coincidence data with internal radiation. Standard clinical data acquisition FOV could be used. The coincidence timing window should be large enough to allow radiation particles, such as gamma particles, to travel across the scanner. The number of coincidence events with internal radiation needs to be enough to calibrate the peak position from the TOF difference histogram for each of the DUs.

(37) In still another embodiment, data with positron-emitting line source and internal radiation could be acquired simultaneously.

(38) In particular, the disclosed method: acquires coincidence data with positron-emitting line source and internal radiation. Standard clinical data acquisition FOV could be used. The coincidence timing window should be large enough to allow radiation particles to travel across the scanner. The number of coincidence events with positron-emitting line source needs to be enough to calibrate the peak position from the TOF difference histogram for each of the crystals, and the number of coincidence events with internal radiation needs to be enough to calibrate the peak position from the TOF difference histogram for each of the DUs.

(39) During the data analysis, if data with positron-emitting line source events and internal radiation events are acquired simultaneously, they could be separated from TOF difference. FOV can also be used to separate line source events from internal radiation events, as line source events concentrate in narrow FOV while internal radiation events have a broader coverage.

(40) During the data analysis, timing correction is split to three different parts: non-energy-dependent relative timing offset per crystal within opposite DU pairs, timing walk correction coefficient per crystal, and non-energy-dependent timing offset between DUs. Non-energy-dependent relative timing offset per crystal within opposite DU pairs and timing walk correction coefficient per crystal are calculated from positron-emitting line source data, whereas non-energy-dependent timing offset between DUs are calculated from internal radiation data. In the description below, relative timing offset within opposite DU pairs and timing offset between DUs refer to non-energy-dependent terms.

(41) With regard to timing calibration within opposite DU pairs, the method splits coincidence data with positron-emitting line source to N/2 DU pairs, for a PET scanner with N DUs. The relative timing offset within opposite DU pairs and timing walk correction coefficients could be calibrated in parallel for different DU pairs. If the line source is not perfectly centered, the annihilation position correction is applied to the TOF difference of all the events. The relative timing offset per crystal within opposite DU pairs could be calculated iteratively, by: i) calculating the timing offset by finding the peak position of the timing histogram for each crystal ii) correcting the TOF difference for the timing offset per crystal calibrated above, then repeating step i) and step ii) until the sequence converges iii) the final timing offset per crystal within a DU pair is the sum over the timing offset per crystal calibrated in all iterations.

(42) After correcting for relative timing offset per crystal within opposite DU pairs, the timing walk correction coefficient per crystal could be calculated as the following: i) for each specific crystal, plotting the timing vs energy curve, the LORs connecting this specific crystal and any crystals on the other side are considered and ii) calculating the walk correction coefficient per crystal by applying appropriate fit (e.g., linear fit or exponential fit) to the timing-energy curve for that crystal.

(43) With regard to timing offset calibration between DUs, the method applies timing offset within DU pair correction and timing walk correction to the radiation coincidence data. Event position correction to the TOF difference is not necessary due to symmetry in DU pair TOF difference histograms. However, event position correction could also be applied to the TOF difference of all the events to achieve narrower timing histograms. The timing offset per DU could be calculated iteratively by: i) calculating the timing offset by finding the peak position of the timing histogram for each DU ii) correcting the TOF difference for the timing offset per DU calibrated above, then repeating step i) and step ii) until the sequence converges iii) the final timing offset per DU is the sum over the timing offset per DU calibrated in all iterations.

(44) In another embodiment, the timing offset per DU may also be calculated analytically. In particular, the TOF difference histogram for each DU pair covered by data acquisition FOV is calculated. The timing center for each DU pair is calculated by finding the peak position of the TOF difference histogram. A set of equations could be formed from the timing center per DU pair. The variables are the timing offset per DU. The rank of the coefficient matrix in the equations should be equal to the number of DUs. The timing offset per DU could be calculated by solving the above equations. The timing offset per DU could also be calculated using Neutral Network from timing center per DU pair.

(45) In still another embodiment, the timing offset per DU may also be calculated using Neural Networks. In particular, the input to the Neural Network could be an array where (for example) each column represents the timing histogram for a single DU. The output would be the offset for each DU. The Neural Network could be trained using target offset data that is generated by using any conventional timing offset calibration method. Training would require data from a large number of systems. Since only a few systems will have been built when the network must be trained, data augmentation (described below) may be used to generate a large number of additional training data sets.

(46) In particular, the data augmentation: acquires data from any existing systems (for example, 3 to 4); calibrates each system using conventional timing offset calibration; uses the calibration to generate corrected timing histograms for each DU pair; for many system realizations (hundreds or thousands), generates random timing offsets for each DU, and applies the random timing offsets to the corrected timing histograms to build augmented data sets for DU in each of the system realizations. For these augmented data sets, the target offset is known from the random timing offsets that we generated for each DU.

(47) The neural network design could be a convolutional Neural Network (to reduce the number of parameters required). In this case, the convolutional layers would be one-dimensional-only acting on the histogram from a single DU (such as the columns of the input matrix, if, as described above, each column represents the histogram from a single DU).

(48) The disclosed method may perform quick timing calibration with internal radiation.

(49) In particular, timing offset per processing unit is calculated. The processing unit here could be a DU, or it could also be an electronics processing unit within a DU. The data acquisition-with internal radiation removes all radiation sources from the scanner and acquires coincidence data with internal radiation. Standard clinical data acquisition FOV could be used. The coincidence timing window should be large enough to allow radiation particles to travel across the scanner. The number of coincidence events with internal radiation needs to be enough to calibrate the peak position from the TOF histogram for each of the processing units.

(50) In the data analysis, timing offset correction and timing walk correction from initial timing calibration is applied before performing quick timing calibration. The data analysis procedure to calibrate timing offset per processing unit in quick timing calibration is the same as the data analysis procedure to calibrate timing offset per DU in initial full timing calibration.

(51) In another embodiment, the disclosed method may perform quick timing calibration with internal radiation and positron-emitting line source. If timing offset for processing units smaller than DUs needs calibration, quick timing calibration could also be calculated similarly as initial timing calibration.

(52) The data acquisition is the same as the initial timing calibration, except that: the number of coincidence events with positron-emitting line source needs to be enough to calibrate the peak position from the TOF difference histogram for each of the processing units, and the number of coincidence events with internal radiation needs to be enough to calibrate the peak position from the TOF histogram for each of the processing units.

(53) Data analysis is similar as initial timing calibration.

(54) Timing offset correction and timing walk correction from initial timing calibration is applied before performing quick timing calibration. Timing offset per processing unit within opposite DU pairs is calculated using positron-emission line source at center data, similarly as in initial full timing calibration. Timing offset between DUs is calibrated using internal radiation, the same as in initial full timing calibration.

(55) FIG. 12A and FIG. 12B show example DU pair TOF difference histograms from lutetium background radiation data when calculating timing offset between DUs. TOF difference is calculated as (time stamp for the first hit−time stamp for the second hit).

(56) The timing center of the DU pair TOF difference histograms could be found from Gaussian fitting to the whole curve or parabola fitting to the peak region. Timing center of the DU pair TOF difference histograms could also be found using Neutral Network (NN).

(57) Equations for determining the timing offset per DU are as the following.
Tcenter.sub.14-33=Toffset.sub.14−Toffset.sub.33−Tdiff.sub.distance
Tcenter.sub.33-14=Toffset.sub.33−Toffset.sub.14−Tdiff.sub.distance
Toffset.sub.14−Toffset.sub.33=(Tcenter.sub.14-33−Tcenter.sub.33-14)/2

(58) Tdiff.sub.distance is event position correction to the TOF difference, which is cancelled out when calculating timing offset difference between DU 14 and DU 33.

(59) FIG. 13 shows another way of calculating DU pair TOF difference histograms from lutetium background radiation data when calculating timing offset between DUs. TOF difference is calculated as (time stamp for DU 14−time stamp for DU 33). Timing center of the DU pair TOF difference histograms could be found from Gaussian fitting to the whole curve or parabola fitting to the peak region. Timing center of the DU pair TOF difference histograms could also be found using Neutral Network (NN).

(60) Equations for timing offset per DU is as the following.
Tcenter.sub.left=Toffset.sub.14−Toffset.sub.33−Tdiff.sub.distance
Tcenter.sub.right=Toffset.sub.14−Toffset.sub.33+Tdiff.sub.distance
Toffset.sub.14−Toffset.sub.33=(Tcenter.sub.right+Tcenter.sub.left)2

(61) FIG. 14 shows example DU pair TOF difference histograms from lutetium background radiation data and Ge line source data acquired simultaneously. TOF difference is calculated as (time stamp for DU 14−time stamp for DU 33). The positron-emitting line source data and the lutetium internal radiation data could be separated from time of flight (TOF) difference.

(62) Timing offset between DU pairs is greatly reduced after quick timing calibration. For example, FIG. 15A shows timing offset between DU pairs before quick timing calibration, and FIG. 15B shows timing offset between DU pairs after quick timing calibration. According to one embodiment discussed herein, an accurate, convenient and fast method for timing calibration for TOF PET scanners is provided.

(63) Various embodiments discussed herein provide good timing resolution in order for a TOF PET scanner to effectively reduce the statistical noise in the reconstructed images to improve the image quality and may be used to maintain accurate timing correction during daily clinical use in order to achieve images for TOF PET scanners with a reduced number of artifacts.

(64) According to one embodiment, timing offset calibration is provided by coupling together all processing units to be calibrated by coincident events.

(65) According to another embodiment, timing offset calibration is provided by coupling together groups of overlapping crystals by coincident gamma photons until a sufficient number of coincident gamma photons between the groups provides sufficient timing offset calibration for all of the crystals.

(66) According to one embodiment, during initial full timing calibration, (1) timing offset and timing walk within DU pairs or DUs are calibrated by placing an limited extent positron-emitting source in the scanner FOV and (2) after correcting for timing offset and timing walk within DU pairs or DUs, timing offset between DU pairs or DUs are calibrated using an internal radiation (e.g., lutetium).

(67) According to one embodiment, in step (1), the limited extent source is preferably thick enough that crystal is coupled to many crystals in the other DU.

(68) According to another embodiment, during daily clinical use, (1) timing offset correction and timing walk correction from initial timing calibration is applied before performing timing calibration during daily clinical use, (2) timing offset per processing unit is calculated using an internal radiation (e.g., lutetium) and an limited extent positron-emitting source in the scanner FOV together, and/or (3) timing offset per processing unit is calibrated using an internal radiation (e.g., lutetium) while a scanner is not in use.

(69) According to two different implementations, (1) data with limited extent positron-emitting source and an internal radiation (e.g., lutetium) are acquired separately, and (2) data with limited extent positron-emitting source an internal radiation (e.g., lutetium) are acquired simultaneously.

(70) Advantageously, at least using a number of embodiments disclosed herein, (1) there is no need to move a radiation source or use large limited extent radiation source during initial full timing calibration; (2) quick timing calibration can be implemented without external radiation source during daily clinical use; (3) a simplified method is provided that does not require position dependent timing correction; and (4) calibration is relatively fast because of parallel processing and the simplified method.

(71) The method and system described herein can be implemented in a number of technologies but generally relate to processing circuitry for performing the calibration described herein. In one embodiment, the processing circuitry is implemented as one of or as a combination of: an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a generic array of logic (GAL), a programmable array of logic (PAL), circuitry for allowing one-time programmability of logic gates (e.g., using fuses) or reprogrammable logic gates. Furthermore, the processing circuitry can include a computer processor and having embedded and/or external non-volatile computer readable memory (e.g., RAM, SRAM, FRAM, PROM, EPROM, and/or EEPROM) that stores computer instructions (binary executable instructions and/or interpreted computer instructions) for controlling the computer processor to perform the processes described herein. The computer processor circuitry may implement a single processor or multiprocessors, each supporting a single thread or multiple threads and each having a single core or multiple cores. In an embodiment in which neural networks are used, the processing circuitry used to train the artificial neural network need not be the same as the processing circuitry used to implement the trained artificial neural network that performs the calibration described herein. For example, processor circuitry and memory may be used to produce a trained artificial neural network (e.g., as defined by its interconnections and weights), and an FPGA may be used to implement the trained artificial neural network. Moreover, the training and use of a trained artificial neural network may use a serial implementation or a parallel implementation for increased performance (e.g., by implementing the trained neural network on a parallel processor architecture such as a graphics processor architecture).

(72) Numerous modifications and variations of the embodiments presented herein are possible in light of the above teachings. It is therefore to be understood that within the scope of the claims, the disclosure may be practiced otherwise than as specifically described herein.