Patent Application
20120153165
The present invention relates to the imaging of a subject, by the detection of positron emission gamma rays and the subsequent analysis of data relating to the detection, for example by providing more accurate timing information and by weighting the use of data in an image construction algorithm.
Particular described embodiments use a pair of positron emission detectors having a scintillation layer formed of a material such as barium fluoride, an adjacent low pressure gas space, and an electrode grid sensor to detect the position of an electron burst travelling through the gas space.
Positron emission tomography (PET) is a well know technique in which a human, animal or other subject is given a dose of a tracer labelled with a positron-emitting radioisotope. A positron emitted from the radioisotope nucleus within the subject interacts with an atomic electron within a short distance of travel. The electron-positron pair annihilate to form two 511 keV gamma rays which travel away from the point of decay almost co-linearly. Gamma ray detectors disposed about the subject are used to detect these pairs of gamma rays in time coincidence, and the source of decay is assumed to be directly between the detected positions of the coincident gamma rays, along what is conventionally known as the line of response (LOR). An image of the biodistribution of the tracer within the subject is constructed using tomographic techniques from many such coincidences. Typical tomographic techniques used include filtered back-projection, and maximum likelihood expectation techniques.
A pair of gamma rays produced by an electron-positron pair annihilation may each travel to a separate gamma ray detector without any intervening scattering. In this case, for detector separations of the order of a metre, the difference in time between each gamma ray arriving at a detector will be less than about 3 ns. Generally, in order to associate pairs of detection events which are reasonably likely to originate from the same positron, only pairs of events which are closer together in time than a certain threshold, for example 12 ns, are considered coincidence events.
In Time-of-Flight (TOF) PET techniques the actual or relative arrival times at each detector are recorded, and the small time difference is used to estimate the difference between the distances travelled by each gamma ray along the LOR. This estimate of distance is then used in the image reconstruction stage, for example using confidence-weighted versions of the techniques mentioned above. The inaccuracies in the estimate of distance along the LOR are usually far larger than the inaccuracies in the LOR itself.
A significant proportion of gamma rays resulting from positron-emission annihilation in PET system operation are scattered before being detected, for example by nuclei within the subject, or from some part of the detector array or support structure. However, a scattered gamma ray may still reach a detector at a similar time to the other gamma ray of the pair, and within any coincidence time threshold for the system, to be interpreted wrongly as a line of response thereby increasing noise in the final image. Since a scattered gamma ray is likely to have an energy of rather less than the original 511 MeV, detected events having an energy below a threshold such as 400 MeV are typically discarded to mitigate this effect.
The invention seeks to provide improved timing data in respect of gamma rays detected in a positron emission system, and to provide improved image reconstruction from coincidence event data.
The invention relates to gamma ray detection for positron emission imaging, especially tomographic imaging of a subject, such as a human or animal subject for medical purposes.
In particular, a first aspect of the invention relates to a gamma ray detector with a scintillation layer, behind which is a low pressure gas space. An electron burst is generated in response to a gamma ray striking the scintillation layer, for example by conversion of ultraviolet photons from the scintillation layer causing photo-ionisation in the low pressure gas. The electron burst moves through the low pressure gas space to a locator element, such as a multi-wire proportional counter, which provides position information indicative of where on the scintillation layer the gamma ray struck. This position information may be used in imaging the subject. A timing signal for the incident gamma ray is received from a timing electrode plane within the low pressure gas.
According to the invention, the timing signal is adjusted to compensate for different possible travel times of the signal within the timing electrode plane by using position information originating from the locator element. This compensation may be carried out by a timing compensation element.
This compensation improves the accuracy of the timing information sufficiently for time-of-flight of the gamma rays to be taken into account in reconstructing an image of the subject.
Typically, the scintillation layer, the timing electrode plane and the locator element will be substantially parallel, defining a detector plane, with the position information indicating a position or coordinates within that detector plane.
The timing signal represents drift of the electron burst through the timing electrode plane. When using two opposing gamma ray detectors, a coincident timing signal from both detectors, for example coincident within a small window of perhaps 12 nanoseconds, may be taken as indicative of two collinear gamma rays originating from a single positron-electron annihilation event. This determination may be made by a coincidence detector. A trigger signal may then be sent to a gate electrode plane in each detector to permit the electron bursts to pass to the respective locator elements. This mechanism reduces the number of electron bursts reaching the locator element by perhaps two orders of magnitude. The locator element may be a multi-wire proportional counter using delay lines to establish signals carrying the position information.
The timing electrode plane may be formed of a plurality of coplanar and parallel wires, and the position information is then used to estimate position at which the electron burst drifted through or past one or more of the wires, for example as a distance along the wires from a terminus end at which the signal is received. The timing information is then compensated for the travel time of the timing signal along the wires.
Other aspects of the timing signal may also be compensated, for example using predetermined adjustments according to different travel times from the terminus of different parts of the timing electrode plane.
Separate compensated timing information may be generated for the or each detector, or difference compensated timing information, representing the time between two coincident gamma rays striking scintillation layers of two detectors, may be generated and output.
The invention provides one or more gamma ray detectors with appropriate control and data processing elements implementing the above, a system comprising at least two such gamma ray detectors, and a system further comprising data processing elements adapted to carry out construction of an image of the subject using the position and compensated timing information generated for the gamma ray coincidence events. The invention also provides corresponding methods, computer program elements, and computer readable media carrying such program elements.
A second aspect of the invention relates to construction of images of a subject using position and timing information generated for gamma ray coincidence events. This information may be generated by a gamma ray detection system as set out above and described in detail herein, or using other gamma ray detection systems such as a more conventional PET scanner. The compensated timing data may also advantageously be used within this aspect.
According to the second aspect a positron emission density image of a subject within a three dimensional subject space is constructed from detections of coincident positron emission gamma rays. Data defining a plurality of lines of response through said subject space is provided, the lines of response linking locations of said gamma ray detections. This data could be provided, for example, as coordinates of coincident gamma ray detections in the planes of each of two detectors, in association with the position and orientation of the detectors in the subject space.
For each line of response or coincidence event, an estimate of the positron emission location is provided, for example from absolute or difference timing information of the gamma ray detections of a coincidence event. Such an estimate is likely to be very approximate, with a 1 nanosecond uncertainty corresponding to about 200 to 300 mm in the subject space.
An envelope function is then provided within the subject space. The envelope function is intended to approximate the expected positron emission density image to be constructed, although this approximation could be very crude, for example a simple geometric form, or much more sophisticated.
A timing error function is then provided which is representative of the uncertainty of the positron emission location derived from the timing data. The timing error function may typically feature a peak at a best estimate of the emission location from the timing information, the breadth of the peak being representative of the uncertainty in the position of the emission location arising from errors in the timing information.
An evaluation of the envelope function along the line of response is then convolved with the timing error function evaluated along the same line of response and aligned according to the estimate of emission location. The result of the convolution is an emission event weight. An image of the subject is then constructed from each line of response weighted according to the emission event weight.
The envelope function could take a variety of forms. For example, geometric forms could be used such as a cylinder or sphere aligned and sized according to the expected subject image. The function could be two valued, with a larger value within the form and a smaller or zero value outside the form, or the function could be graduated with stepped or continuous values. More sophisticated predefined shapes, for example an approximate heart or kidney shape could be used.
The envelope function may be defined based on data derived from a scan of the subject, such as an X-ray CT scan or ultrasound scan.
The envelope function may also be defined iteratively. For example, a first image construction may be carried out without using an envelope function and timing error function convolution, or using an initial approximate envelope function, with a subsequent envelope function being generated from the image of the first image construction. Further iterations may be carried out to refine the envelope function and image.
The timing error function may take a variety of forms such as a Gaussian peak or a triangular peak, with the breadth of the peak representative of the uncertainty in the position of the estimated emission location due to errors and uncertainties in the timing information.
The invention also provides a data processing apparatus suitable for carrying out the above method, in particular a suitably programmed computer, and more extensively, a system adapted to establish the required coincidence event data such as a positron emission detection system in conjunction with such a data processing apparatus. The invention may also provide such a system incorporating a CT scan, ultrasound scan or other system for deriving an envelope function. Computer program code, and computer readable media carrying such code, the code being arranged to carry out the described methods, are also provided.
The coincidence event data may be stored in a database for use in the image reconstruction.
Embodiments of the invention will now be described, by way of example only, with reference to the accompanying drawings, of which:
Referring now to
The scintillation layer may be constructed using one or more arrays of adjacently stacked of BaF2 crystal rods each of which is aligned to extend in the plane of the scintillation layer, as discussed in copending application GB0709381.8, which is hereby incorporated by reference. Constructing the scintillation layer using rods aligned within the layer permits a thicker layer to be used while mitigating the loss of spatial resolution of the detector this would otherwise cause. In particular, the divisions between the rods reduces the lateral distance, in the width direction of the rods, over which uv photons generated within the layer can travel before entering the low pressure gas space. The provision of slots in each rod similarly limits the lateral distance in the length direction of the rods, over which uv photons can travel. The lateral travel of uv photons in directions coplanar with the scintillation layer is thereby reduced. However, the rods are still reasonably practical to handle and assemble into a frame or other structure to complete the scintillation layer. This is particularly important in larger area detectors, where even using this technique, hundreds of rods may be required.
Conductive wire 22 of 25 μm diameter is wound around each BaF2 crystal with a 250 μm pitch. A first wire plane 24 consisting of 50 μm diameter wire at a pitch of 500 μm is spaced 0.5 mm from the scintillation layer 20. A second plane 26 consisting of 100 μm wire at 1 mm pitch is spaced 3.0 mm from the first plane. A third plane 28, which acts as a timing electrode plane, also consists of 100 μm diameter wire at 1 mm pitch spaced 9.0 mm from the second plane. A gate electrode plane 30 comprising 100 μm wires at 1 mm pitch is positioned 20 mm from the timing electrode plane and has first and second metallic copper mesh screens 32, 34 positioned one on either side. The MWPC is spaced 13.2 mm beyond the gate and is consists of two cathode planes 36 formed of 50 μm wire at 2.0 mm pitch and an anode/cathode plane 38 of 20 μm anode wires perpendicular to 100 μm cathode wires at 4.0 mm pitch. Delay lines are used to read the magnitude and x/y coordinates of an electron burst from the anode/cathode plane.
Incident gamma radiation causes the BaF2 crystal of layer 20 to scintillate, generating ultra violet photons. Some of the UV photons convert in the low pressure gas space adjacent to the crystal, and the resulting electrons are avalanche amplified in the V1=300 V/mm electric field applied between the first and second planes and the lower V2=150 V/mm electric field applied between the second and third planes. A small reverse bias VR<100 Volts is applied to the mesh 22 to prevent build up of positive ions at the scintillation layer. The use of two separate acceleration regions, between the first and second, and second and third planes, permits sufficient electron cascade amplification without instabilities.
The gate electrode plane 30 is normally biased by ±30 V on alternate wires, which causes the plane to act as a barrier to passing electron bursts. If a passing electron burst is detected at the third plane 28, this first signal being represented in
The signal applied to the gate is of very high frequency, and the copper mesh screens 32, 34 positioned either side of the gate, which are held at a voltage consistent with drift of electrons past the gate and on towards the MWPC, act to shield this high frequency signal from the rest of the detector.
Although a gamma ray detector using TMAE gas as the photoionization medium has been described, other photoionization arrangements could be used, for example incorporating a more conventional low pressure gas such as Helium or Argon.
The time taken for a signal A1 to arrive at the coincidence detector also depends on the distance from a preamplifier, along a wire of the third plane, of the passing electron burst 55, illustrated in the figure as distance d. The velocity of the signal A1 along a wire of the plane may typically be about 200 mm per nanosecond, corresponding to a time variation of several nanoseconds dependent upon the position of the electron burst.
To reduce the effect of signal travel time along the third plane wires upon the detection of coincidences, the preamplifiers 52 and connections to the third plane of one of the gamma ray detectors 10 are disposed at the opposite end to those of the other, as illustrated in perspective in
The described diametric geometry is of decreasing benefit in terms of balancing the delays as a gamma ray event occurs further from a plane joining the centrelines of the third plane wires of both detectors. Consider off-centre positron decay event 60 which gives rise to a signal A1 which originates close to the preamplifiers in the third plane wires of one detector, and signal A2 which originates far from the preamplifiers 54 of the other detector. If the third plane wires are each about 600 mm in length then this gives rise to a difference of up to about 3 nanoseconds between the times of arrival of the signals A1, A2 at the coincidence detector. This time difference can be corrected by using knowledge of the spatial coordinates, for example the distance d, in each detector plane of the electron burst, for example from the MWPC signals, as described in more detail below.
When an electron burst passes both third planes 28 at about the same time, this coincidence is detected by the coincidence detector 38, for example by judging the arrival time of signals A1 and A2 to be within about 12 nanoseconds of each other at the coincidence detector. The detector 38 then sends a trigger signal T to the gates 30 to permit the electron bursts to pass on to the locator elements 40. Each locator element then generates a signal Z1, Z2 which represents the coordinates in the plane of each detector of the respective electron bursts. The signals Z1, Z2 are received by the spatial coordinate function which generates the coordinates x1, x2.
From the time of arrival at the coincidence detector 38 of the signals A1 and A2 from the respective third planes, the coincidence detector 38 is able to generate a raw time difference R. This is passed to the time compensation function 72. As described above in connection with
If the coordinates x1, x2 indicate distances d1, d2 of the electron bursts detected in the third plane from the ends of the wires in each plane at which the corresponding signal is collected and amplified, and the speed of travel of the signal along the third plane wires is v, then Δt may be determined by adjusting the raw time difference by an amount (d1−d2)/v to compensate for travel along the wires of the third plane, with the sign of the adjustment being determined as appropriate.
The pulse size data p1, p2 may be determined from the strength of the signals A1, A2, either by the coincidence detector 38, or as illustrated in
The control and signal handling function 14 may implement some filtering of the coincidence data in addition to the gating function described above in connection with
Whereas some aspects of the described control and signal handling function will be implemented using electronic circuitry, some aspects may be implemented in software, or may be incorporated instead in functions of the data processor 16 or elsewhere. For example, the compensation of the time difference for signal travel time in the third plane wires may be handled using suitable computer software based on an uncompensated version of the Δt in the coincidence event data 70, and the coordinates x1, x2 in the event data.
Separate compensated timing information may be output for each detector, as well as or instead of difference timing information.
Other corrections to the raw time difference signal may also be implemented in a similar manner to those described above. For example, adjustments to allow for different lengths of cables 54 from the preamplifiers 52 of different groups of third wire planes may be implemented by determining the appropriate wire group from the detected spatial coordinates, and applying an adjustment predetermined for that particular wire group.
Although in
An image of the subject 12, or more particularly, an image of a tracer distribution within the subject, is constructed using event data 70 or a subset thereof from a large number of coincidence events, using a tomographic technique such as filtered back-projection, and maximum likelihood expectation techniques. In these techniques, a weight can be attributed to each event, so that some events have a greater influence on the constructed image than others.
The spatial coordinates of each event are used to define a line of response LOR, somewhere upon which the two detected gamma rays are presumed to have originated at a positron emission event. Without any time difference data, nothing is known about the position of the event along the LOR. However, once some timing information is available, this can be used to influence the image construction, as long as the uncertainty in the timing information is small enough, for example, less than about 3 nanoseconds corresponding to roughly 900 mm of gamma ray travel in free space, or 600 mm in a human or animal subject. The uncertainty in the timing information can be reduced, for example, using the techniques described above for low pressure gas space gamma ray detectors.
A weight for a coincidence event which is based on the timing information can be derived using an estimate of the uncertainty in the timing information, and an estimate of the expected positron emission density which forms the subject of the constructed image.
The position of the positron emission on the line of response is not known accurately, because uncertainties in the timing information are large. However, a best estimate of the position is shown as 86. A timing error function 88 representative of the uncertainties in the timing information can be evaluated along the LOR, and will typically have a peak 90 at the best estimate of the emission position based on the timing information. In
The timing error function may also vary according to the data upon which the line of response is based, for example it may vary according to the position of the end points of the LOR, including angle of the LOR relative to the detector planes. It may depend upon the detected pulse sizes, for example having a narrower peak for larger pulse sizes which have better timing certainty.
An envelope function 92 is provided which is based on a prior estimate of the expected positron emission density to form the image to be constructed. The prior estimate may be very crude, such as a simple geometric shape such as a cylinder or sphere, with just two, many, or a continuous range of values within the space, and with sharply defined or more diffuse features and boundaries. More sophisticated envelope functions can include a form based on an X-ray scan CT image, ultrasound imagery, and prior models of organ forms such as a prior estimate of a kidney or heart shape.
The envelope function 92 may also be based on a previous construction of the positron emission density image, which itself used no envelope function, or used an initially estimated function such as a simple geometric form. The process of using an envelope function to generate an image, and using the image to generate a more refined envelope function may be iterated.
To derive a weight an evaluation of the envelope function along the LOR is convolved with the timing error function along the same LOR. For example, the two functions so evaluated may by multiplied together at regular intervals along the LOR, the products summed, and the result normalised as appropriate. The weight so derived is then used to provide a weighting for the use of the coincidence event data for that LOR in a construction of the subject image, for example using a filtered back projection technique.
The apparatus as described herein may be operated by accepting all events within a preset timing window and this allows events to be accepted even if they are obviously outside the field of view of the camera. All that is required is that a trigger is received by the third plane amplifiers of each detector within this timing window and that a positional readout is provided by each detector.
However, measurement of the spatial coordinates of the event can provide a time marker to about 25 ps for comparison with the timing difference, and this can be used to compare with the time of flight measurement for each event to accept true events and reject a large number of random coincidences or scattered photons.
In
Hence there is only about 1 ns difference between good events and scattered events independent of whether the events are directly across the detectors or obliquely into the corners. Without comparing timing differences with event coordinates the timing resolution may be about 3.5 ns FWHM with a roughly Gaussian distribution of events (standard deviation about 1.5 ns). The timing difference (dT) between the coincidence events measured using the timing discrimination circuits and that measured from event coordinates can be used to discriminate between true events and scattered events because scattered events will tend to have a higher probability of occurring later than the true time (+dT).
The ratio of true to scatter events (T/S) as a function of dT is shown in
Although particular embodiments have been described, variations and modifications will be apparent to the skilled person without departing from the scope of the invention as defined in the appended claims. For example, the described data weighting technique does not require the use of the gamma ray cameras described herein in detail, and other sources of coincidence data may be used.
| Number | Date | Country | Kind |
|---|---|---|---|
| 0808310.7 | May 2008 | GB | national |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/GB09/01159 | 5/7/2009 | WO | 00 | 11/5/2010 |
