WO2012101332A1

WO2012101332A1 - Method and device for correcting a pulse stack for a radiation detector

Info

Publication number: WO2012101332A1
Application number: PCT/FR2011/000055
Authority: WO
Inventors: Michel Tararine
Original assignee: Michel Tararine
Priority date: 2011-01-27
Filing date: 2011-01-27
Publication date: 2012-08-02
Also published as: RU2013139563A; EP2668521A1

Abstract

The patent relates to a method and device for integrating electronic pulse signals output by a gamma radiation detector, and for correcting pre-pulse and post-pulse stack effects. The device uses an architecture having a double integrator which operates in parallel, the first integrator partially integrating the portion of the signal that precedes the pulse, and the second integrator integrating the pulse signal itself. The first integrator (441) performs a partial integration of the signal before the pulse, i.e. a rapid measurement of any pulse tail of a preceding event. Then, an estimation of the value of the pre-pulse stack correction is calculated by extrapolation, from this measured value and in accordance with the value of the integration time used in the second integrator (401), of a model of the shape of the pulse, and of the value of the time separating the current event (during the integration in the second integrator) and the preceding event. The output of the second integrator, which integrates the signal of the current event over a time period that can be shortened when a post-pulse stack is generated, is corrected simultaneously with the pre-pulse stack effect and the variable integration time using the same pulse-shape model.

Description

METHOD AND DEVICE FOR PULSE STACK CORRECTION FOR A RADIATION DETECTOR

TECHNICAL AREA

The present invention relates to nuclear medical imaging systems, and more specifically to the correction of the stacking effects of the electronic pulse signals delivered by the gamma radiation detectors. STATE OF THE PRIOR ART

The present invention relates to a pulse signal detector occurring randomly over time.

The invention can be used for the detection of nuclear radiation using a scintillation detector, for example a gamma camera. A gamma camera includes means for locating scintillations occurring in the scintillator crystal under the effect of incident gamma radiation. The invention finds particular application

advantageous in the field of nuclear medicine in order to visualize in an organ the distribution of molecules labeled with a radioactive isotope and administered to a patient.

U.S. Patent No. 3,011,057, for example, describes a gamma camera in accordance with the preamble. This includes a collimator for selecting the direction of incident gamma photons, a scintillator crystal with which incident photons interact to give rise to light pulses or scintillations, and a network of photomultiplier tubes which in turn transform the received scintillations into pulses. electric. This network of photomultiplier tubes is part of the localization means which, from the electrical pulses provided by the tubes, deliver, in a known manner, X and Y coordinate signals from the place where the scintillation has occurred, as well as a signal E proportional to the energy of the photon that produced the scintillation. Signal E is used to differentiate the different types of incident gamma photons and to identify noise, direct radiation, and scattered radiation.

The scintillation produces a sheaf of photons emitted in all directions isotropically within the scintillator crystal, it is simultaneously seen by several tubes photomultipliers. The determination of the location of this scintillation on the crystal, itself representative of the emission site of the incident gamma photon, is obtained by calculating the location of the center of gravity, that is to say a weighted sum of electrical pulses. delivered by all the photomultiplier tubes excited by the scintillation considered. Similarly, the energy of the incident photon is obtained by unweighted summation of the electric pulses delivered by all the photomultiplier tubes.

These electrical pulses generally have a known and constant average shape for a given scintillator material and a given light detection and signal processing electronics. This shape, or amplitude of the signal as a function of time, represents the average flux of scintillation photons emitted over time. For most scintillator materials, this pulse can be with a good approximation described by the following bi-exponential model:

when t> 0

when t <0 where xi is the decay time constant of the pulse

τ ₂ is the pulse rise time constant

For a thallium-activated sodium iodide crystal (Nal (Tl)), which is of conventional use in gamma cameras, τ is typically of the order of 250 nanoseconds and τ ₂ typically of the order of 80 nanoseconds. However, since the emission of scintillation photons is a random phenomenon in time and space, each electrical pulse is subject to statistical fluctuations that modify its shape and amplitude unpredictably. These fluctuations are directly responsible for the inaccuracy of the calculated values (X, Y) of the coordinates of the scintillation site and the calculated E value of the total energy of the scintillation, which determine the spatial resolution and the energy resolution. of the instrument.

In order to minimize the influence of these fluctuations on the accuracy of the calculation, these pulses are integrated over a certain time before being used for the calculation of X, Y and E. This Integration operation physically corresponds to counting the total number of scintillation photons emitted throughout the duration of the pulse. The integration time is typically of the order of several times the decay time of the pulse, so that the amount of residual scintillation photons after integration is negligible compared to the amount of photons that have been integrated. Thus, for example for a crystal of Nal (Tl), this integration time is typically of the order of 1 microsecond. The integration of the signals makes it possible to overcome to a certain extent the instantaneous statistical fluctuations in the flow of scintillation photons and to obtain values proportional to the total quantity of photons emitted during the scintillation, which are therefore more precise than the instantaneous value of the signal.

The disadvantage of integrating the signals is that the quality of the measurement of the values of X, Y and E (and therefore the performance of the detector) becomes very dependent on the counting rate to which the detector is exposed. that is, the number of scintillations that occur per second. Indeed, scintillations occurring randomly in time according to the statistical law of Poisson, it happens that pulses occur very closely in time and that as a result the electronic signals overlap partially, causing a known phenomenon under the name of stacking. Uncorrected, stacking effects produce severe distortions when calculating (X, Y) and E and are likely to lead to unacceptable degradation of image quality.

The permissible count rate limitation of such imaging devices is

particularly problematic in many clinical situations requiring high counting capacity, such as imaging of patients who have received therapeutic and therefore high doses of radioisotopes, dynamic first-pass cardiac imaging, the use of tracers with very short lifetimes, acquisition of transmission data simultaneously with data acquisition

of emission, the use of dual-detector gamma-cameras operating without collimator in coincidence mode in imaging of positron emitters, etc.

The stacking phenomenon can be of two types: the upstream stack (or prepulse), and the downstream stack (or postpulse). Upstream or pre-impulse stacking is due to events occurring before the beginning of the integration of the current event. The pulse signal of the current event is thus superimposed on the tail of the previous event. The downstream or post-pulse stack is due to events occurring after the beginning of the integration of the current event, that is, while the current event is being integrated. Different methods of correcting these two types of stacking have been described in the literature.

Some gamma cameras of the prior art use a method based on simple rejection of events. Indeed, as the stacking phenomenon has the effect of artificially increasing the integral of the signals, the selection of events is done on the basis of the comparison of their energy with a given acceptance window. Events whose energy is outside this acceptance window are rejected. This method avoids taking into account events whose values (X, Y, E) are distorted or distorted, but suffers from severe losses of events induced by high rejection rates occurring at high counting rates.

US Pat. No. 5,276,615 describes a method for shortening pulse duration by deconvolution or signal filtering, so as to reduce the integration time and thus reduce both losses due to dead time and the probability of stacking. . Although the signal after deconvolution can theoretically be shortened at will without loss of information, a practical limit is imposed by the noise component in the signal. Indeed, the shortening of the signal is obtained by amplification of the high frequencies (in the Fourier space), which carry more and more noise and less and less signal as the frequency increases. For very fast filters, that is to say producing very short pulses, the noise level can reach such values as the losses in precision, that is to say the losses in spatial resolution and in energy, can become unacceptable. In addition, since filtering is a stationary operator, these impairments are introduced independently of the count rate, which makes the system optimized for only one count rate at a time.

U.S. Patent No. 5,210,423 discloses another method where signal integration

impulse of an event is stopped when another event is detected, so as to avoid downstream stacking. Then this shortened integral, in association with the

knowledge of a model of the general pace of pulses, is used to calculate by extrapolation an estimate of the missing information, ie the integral of the remainder of the signal. This estimation of the integral of the remainder of the signal is subsequently used firstly to correct the current event of the shortening of the integration time due to the downstream stack by adding the integral of the remainder of the signal, and on the other hand to correct the integral of the following event, which caused the downstream stacking of the upstream stack by subtracting the same integral from the rest of the signal, corresponding to the tail of the previous event.

However, this method has the disadvantage of requiring that the two events, current and next, are necessarily detected and integrated, fully or partially, even in the case where one of these events is unnecessary and could have been eliminated at one time. early stage of treatment, for example by selection of the amplitude of events or another method. Such superfluous treatment of unnecessary events (capable of increasing the total number of processed events by an order of magnitude) ultimately requires additional computational capacity which inevitably has an impact on both the downtime of the system and the cost of the equipment.

In addition, since all events must be detected in order to perform upstream stacking correction, the effectiveness of such a method is very dependent on the event detection principle. The presence or absence of an event is determined by a processing and a given analysis of the pace and amplitude of the signal, usually performed by a device called "trigger". Such triggers, besides the fact that they have a certain dead time, are therefore sensitive to the pace of the signal and noise levels. In addition, triggers of any type have fixed trigger parameters, such as a low threshold, which makes it possible to avoid considering fluctuations in the level of rest due to noise as possible events. Low energy events are never detected. At high counting rates, stacking phenomena, increased noise levels, and dead-time losses make detection of random events and triggers less dramatic, which correct upstream stacking correction.

Another additional disadvantage of this method is that there is no upstream stacking correction for events that are separated by a time greater than the integration time, while the remaining part of the signal still carries a small portion but not insignificant information, of the order of a few percent. This systematic underestimation of tails of residual events is also the cause of bias and distortions that contribute to the general degradation of image quality. US Pat. No. 6,291,825 partially solves the first problem cited above by calculating an estimate of the tail integral of the previous event from the instantaneous value of the signal, thus not requiring knowledge of the arrival times of the successive events.

However, this approach suffers from the high level of noise associated with an individual sample of the pulse and contributes to degrading the system's spatial and energy resolutions, the improvement of which is in fact the intrinsic reason why the signals are integrated.

In addition, this method, as well as the method described above, does not properly process bursts of three or more events close enough to be stacked on top of each other. Indeed, as downstream and upstream stacking corrections are performed independently, it is not taken into account a possible

shortening the integration of the second event with a third close event. When such a situation occurs, the second event is corrected from the upstream stack by the tail of the previous event as if this event queue were integrally integrated, i.e. integrated over a time that may be longer than the actual integration time of the second event, which overestimates the upstream stacking correction of the second event. This non-correlation of upstream and downstream stacking corrections is the reason why such a system performs a false correction, the average value of which can be partially compensated by any phenomenological model (for example as a function of the average counting rate or other ), but can not correct exact stacks for a given sequence of multiple events.

A third disadvantage of this method is that the fact of estimating the tail of a pulse from the instantaneous value of the signal assumes that the pace of the pulse is constant, for example purely exponential or linear, as well as described in the aforementioned patent. This approximation becomes invalid for more complex pulse patterns such as the bi-exponential model described in the preamble, and it would be necessary to take into account not only the instantaneous amplitude of the signal, but also the time at which this amplitude is measured. relative to the beginning of the impulse.

US Patent No. 7,439,515 only bypasses the second problem mentioned above, by executing the upstream and downstream stacking corrections in a correlated manner, in order to compensate for the upstream stacking correction of the downstream stack or the shortening of the stack. signal integration time. However, as in the mentioned patent previously, the efficiency of the method relies on the effective detection by the trigger of all the events independently of their characteristics, in order to detect the appearance of stacking conditions and to be able to process the events

correctly. Thus, in practice, not only does the stacking correction fail at a high counting rate because it is highly dependent on the effectiveness of the trigger, but it also requires significant calculation means so as to be able to continuously deal with a high flow of data. much of which is unnecessary, which in any case increases the loss of events associated with the system timeout. The present invention describes a new and improved method and apparatus that do not have these disadvantages as well as others of the prior art.

The object of the present invention is to provide, in particular for radiation detection devices delivering electronic pulse signals in response to events arising from the interaction of quanta radiation incident with the detector, a method for performing a correction of upstream and downstream stacking effects, the method of detecting the occurrence of an incident event, integrating a portion of the signal preceding the impulse of the event over a predetermined time interval, integrating the pulse of the event on a predetermined time interval or until a new event is detected, applying an upstream or prepulse stack correction based on the integral of the signal portion preceding the pulse of the event and integration time of the event pulse, application of a downstream or post-pulse stack correction based on the time of the event integration of the impulse of the event.

According to another aspect of the invention, the method may further comprise a step of measuring the time that has elapsed between the current event (being integrated) and the previous event, so that the upstream stacking correction is based not only on the integral of the signal portion preceding the event pulse and the integration time of the event pulse, but also on the time that has elapsed between the current event and the previous event. This makes it possible to take into account the shape factor of the pulse and to accurately calculate the recovery integrals corresponding to complex pulse shapes.

The invention also relates to devices that make it possible in particular to implement the method described above. The subject of the invention is a signal processing device for a radiation detector delivering electronic pulse signals in response to events arising from the interaction of incident radiation quanta with the detector, which device comprises means (a) for detecting the occurrence of an incident event, means (b) for integrating a signal portion preceding the pulse of the incident event over a predetermined time interval, means (c) for integrating the pulse of the incident event incident on a predetermined time interval or until a new event is detected, means (d) for performing an upstream stack correction based on the integral of the signal portion preceding the pulse of the event and the time

event pulse integrating means (e) for performing a downstream stack correction based on the event pulse integration time.

Advantageously, the means (d) for performing the upstream stacking correction furthermore include a counter intended to measure the time which has elapsed between the current event (being integrated) and the preceding event, so as to the upstream stacking correction is based on the integral of the signal portion preceding the pulse of the current event, the integration time of the pulse of the current event, and the time that 'is passed between the current event and the previous event.

Preferably, the integration (b) of the signal portion preceding the pulse of the event and the integration (c) of the event pulse are performed by two separate integrators operating in parallel, the signal signals input applied to the integrators being delayed by two different delays before being applied to the integrators. This architecture makes it possible to perform both integration operations simultaneously, then to execute the upstream and downstream stacking corrections in the same treatment cycle independently of the arrival time of the successive events.

Preferably, the means for performing the upstream stacking correction (d)

comprise a multiplier, a memory and a subtractor, said correction being effected by multiplying by means of said multiplier the integral of the signal portion preceding the pulse of the event by the output of said memory whose address is given by the integration time of the event pulse, then subtracting the result of this operation from the integral of the event pulse. When the means for performing the upstream stack correction further comprise a counter measuring the time elapsed between the current event (being integrated) and the previous event, the address of the memory below. above mentioned is obtained by a combination of the integration time of the pulse of the current event and the value of said counter.

Preferably, the means for performing the downstream stacking correction (e) comprise a multiplier and a memory, the said correction being performed by multiplying with the aid of the said multiplier the integral of the pulse of the corrected event of the stack upstream by the output of said memory addressed by the integration time of the pulse of the event.

The present invention has the advantage of making accurate and reliable upstream and downstream stacking corrections, which improve the spatial and energy resolutions of the instruments, particularly at high count rates.

Another advantage of the present invention is that the improvement of the spatial and energy resolutions depends to a lesser extent on the efficiency of the triggering system.

The present invention may be advantageously used in nuclear medical imaging systems such as gamma cameras or positron emission tomographs.

SUMMARY DESCRIPTION OF THE FIGURES

The present invention is described with the help of the following figures, given by way of illustration and not limitation:

FIG. 1 shows by way of example the signal associated with a burst of three events approximated in time, the time intervals used to measure the different integrals of the pulse signals of the events, and the integrals themselves without stacking correction.

FIG. 2 shows the signal associated with a burst of three time-close events and the integral of the pulse tails of the previous events to be subtracted from the raw integrals shown in FIG. 1 so as to perform upstream stacking correction.

FIG. Figure 3 shows the signal associated with a burst of three time-lapse events and the time intervals used to measure the different integrals of the pulse signals of the events that will be used to calculate the integral of the pulse tails of the previous events indicated in FIG. . 2.

FIG. 4 shows a preferential way of producing an integrator comprising upstream and downstream stacking corrections. DETAILED DESCRIPTION OF THE INVENTION

Although the invention is described as implemented in a gamma camera

conventional Anger type, the invention is not limited to gamma cameras and can be broadly applicable to any type of particle detector delivering electronic pulse signals in response to incident events. In particular, in the implementation mentioned, it can be applied to the individual photomultiplier tubes as well as to any signal obtained by summing the output signals of the photomultipliers.

It should be noted that for practical implementations of the method described below, the analog signals delivered by the detector are converted into digital signals using Analog-Digital Converters (ADCs) which convert the signal continuously and at a constant rate. frequency high enough to satisfy the Nyquist criterion. As a result, all operations described in the following method are performed on digital signals.

The event trigger referred to below is a device delivering a logic signal when an event is detected and is used to trigger the sequence of integration and stack correction of the pulse signal of an event.

FIG. 1 illustrates the stacking phenomena occurring when a burst of three events close together in time is detected.

The beginning of the signal, indicated in dotted line, corresponds to a moment when one does not have any information on the presence or the absence of a possible previous event. This corresponds to situations where the event trigger did not detect any event in a significantly long time interval before the first burst event. This significantly long time interval corresponds to the value of the time from which the energy contained in the remaining portion of a pulse falls below the energy of an individual scintillation photon, i.e. it corresponds to the extinction time or time at which the scintillation actually ends. As an indication, for a scintillating crystal of Nal (Tl) detecting gamma rays of 100 keV, this value is of the order of several microseconds, that is to say that in practice it can be significantly greater than the time typical integration of an event.

It is important to note that the fact that the trigger did not detect any events does not necessarily mean that there were no events. The trigger may have detected no events during this time for various reasons, such as: trigger makes a selection of events with peak amplitude values that must be included in a predefined acceptance window, or an event has occurred while the trigger is idle because the dead time following the detection of a previous event has not yet elapsed, or the counting rate is so high that the stacking phenomena make events difficult to discern, or the signal undergoes statistical fluctuations such that the trigger is unable to identify valid pulses of events.

The signal portion corresponding to the first event is integrated from to to ti, giving a value integral Ii. At the end, the integration is stopped by the detection by the trigger of a second event.

The signal portion corresponding to the second event is integrated from t ₁ to t ₂ , giving an integral of value I ₂ . At t ₂ the integration is stopped due to the detection by the trigger of a third event.

The signal portion corresponding to the third event is integrated from t ₂ to t ₃ , giving an integral of value I ₃ . At t ₃ the integration is stopped after the nominal integration time because no event is detected before t ₃ , and therefore (t ₃ -t ₂ ) is equal to the nominal integration time.

FIG. 2 indicates the portions of the pulse tails of the previous events which are integrated with the integrals I ₁ , I ₂ and I ₃ and which must be subtracted from them in order to carry out the upstream stacking correction. The integral contains a portion of a possible residual tail of a previous event that has not been detected. This residual tail integral is denoted ΙΌ.

Similarly, the integral I ₂ also contains a portion of the residual tail of the previous event, noted ΙΊ and whose value also depends on fo-ti), which shows how the upstream stacking correction is in close correlation with the integration time and therefore can not be performed independently of the correction

stacking downstream.

The integral I ₃ also contains a portion of the residual tails of the previous events, denoted Γ ₂ , which itself includes a portion of the residual tail of the first event, denoted

FIG. 3 describes, on the same salvo of three close events, the different operations to be performed in order to make precise corrections of upstream and downstream stacking. In order to perform upstream stacking correction on the first event, the signal portion preceding the event is integrated on an Atso time from tso = to-Atso until time to, giving a value integral. Co. This value Co is of particular importance because it corresponds to the measurement of a possible pulse tail of an event that would have occurred before the first burst event, and is used to extrapolate an estimate of I'o, which is used later to correct the value of Ii.

The signal portion corresponding to the first event is integrated from t0 to _t5 giving a value integral Ii. At the time the integration is stopped due to the detection by the trigger of a second event.

In order to perform the upstream stack correction on the second event, the signal portion preceding the second event, i.e., a signal portion corresponding to the first event, is integrated over a time At _S i from tsi = ti-Atsi to time ti, giving an integral of value Ci. This value is used to calculate by extrapolation an estimate of ΙΊ, which is used later to correct the value of I ₂ . The signal portion corresponding to the second event is integrated from t ₁ to t ₂ , giving an integral of value I ₂ . At time t ₂ the integration is stopped due to the detection by the trigger of a third event.

In order to perform upstream stacking correction on the third event, the signal portion preceding the third event, i.e., a signal portion corresponding to the second event, is integrated over a time Ats ₂ to from ts ₂ = t ₂ -Ats ₂ to time t ₂ , giving an integral of value C ₂ . This value is used to extrapolate an estimate of I ' ₂ , which is used later to correct the value of I ₃ . The signal portion corresponding to the third event is integrated from t ₂ to t ₃ , giving an integral of value I ₃ .

Accordingly, for the three-event burst described above, the measurement and correction method performs in parallel the measurements of the integrals of the event impulse signals, namely L, I ₂ and I ₃ , and the measurements of the integral of the events. signals preceding events Co, Ci and C ₂ . Then it corrects L, I ₂ and I ₃ upstream stacking corrections obtained from Co, Ci and C ₂ to obtain corrected signals of the upstream stack Ji, and J ₃ , and finally it corrects the values of the signals. corrected the upstream stack Ji, J ₂ and J ₃ to obtain final values K ₁ , K ₂ and K ₃ corrected for both upstream and downstream stacking effects. The entire correction process is described in the following sequence of operations:

Operation Result

First event detected by the trigger

Measure of the integral Co

Beginning of the accumulation of the integral Ii

Waiting for the detection of a second event

defined

(or flow of integration time)

At this moment, end of the accumulation of the integral

At this moment, beginning of the accumulation of the integral I ₂

Measure of the Ci integral

Calculation from C ₀ of the value of ΙΌ I _Q = C _Q X f (At _S0 , t _l - t _Q , i _max

Correction of 1 \ upstream stacking

(subtraction of the tail of the previous event)

Ji correction of downstream stacking

(compensation of the tail of the missing signal) K, = J, x gfa -

Waiting for the detection of a third event

t ₂ defined

(or flow of integration time)

At this moment, end of the accumulation of the integral I ₂ = Σs, At this moment, beginning of the accumulation of the integral I ₃

Measure of the integral C ₂

Calculation from Ci of the value of ΙΊ

Correction of I ₂ of upstream stacking

(subtraction of the tail of the previous event) J ° 2 = 1 ¹ 2 - Γ ¹ \

J ₂ correction of the downstream stack

(compensation of the tail of the missing signal) K ₂ = J ₂ ^ g {t ₂ - t _x )

Waiting for the flow of integration time

t ₃ defined

(or waiting for a new event)

At this moment, end of the accumulation of the integral I ₃ = ± s,

Calculation from C ₂ of the value of Γ ₂

)

Correction of I ₃ of the upstream stack j _^ _ j _ jt

(subtracting the tail of the previous event) 3 3 2

J ₃ correction of the downstream stack K _~ JX g (t - t

(compensation of the tail of the missing signal) ³ 3 ⁰ V 3 2 /

In the sequence described above, the function ^ Ats, ti, PE) represents the correction fraction due to upstream stacking, ie the proportion of the tail of the previous event which is integrated in same time as (i.e., stacked on) the pulse signal of the current event, where:

Ats is the time of integration of the signal before the pulse of the current event

ti is the integration time of the pulse of the current event

tpE is the time that elapsed between the current event and the previous event.

For a purely exponentially decreasing momentum model, the integrals of the signal over two successive and distinct time intervals are linked by a simple linear relation that depends only on the respective positions of the bounds of the integrals. In this case, the time separating the current event from the previous event plays no role, and function / only depends on Ats and ti, removing the need to have t _PE in the input argument list of function /

However, assuming a more realistic and complex pulse model, such as the bi-exponential model described in the preamble, the integrals of the signal over two successive and distinct time intervals depend not only on the respective positions of the bounds of the integrals, but also the time at which these limits are positioned with respect to the temporal distribution of the event, that is to say with respect to the time at which the event occurred. Thus, when the signal integrals are positioned far from the top of the pulse, the purely exponentially decreasing model can be considered a good approximation, whereas when the integrals are near the vertex or even include the peak of the pulse, this approximation ceases to be valid. These considerations emphasize the need to have the parameter tpe in the list of input arguments of the function /, that is to say to take into account the temporal proximity of

the previous event to correct pulse form factor.

In the complex situation described in FIG. 2 where the pulse signal of the third event is superimposed on the tails of the two previous events which occurred at different times, consider only the parameter tpE, the time interval between the second and the third event, introduces only one second-order approximation on the value of ΓΊ which has only a minimal influence on the accuracy of the correction.

It is important to note that measurements of C ₀ , C ₁ and C ₂ values and their use in the correction process are performed regardless of whether a previous event has been detected or not, so that a correction of Effective upstream stacking is performed for each event detected by the trigger, even in cases of trigger failure to detect a possible previous event. This technique makes the upstream stacking correction still operational even under metering rate conditions severe enough to produce a high percentage of trigger failures.

The accuracy of the correction is also independent of the integration time of the current event, ie it always takes into account the exact contribution of the remaining queues of the impulses of the previous events (detected or not) to integration, regardless of the value of the integration time. In the case where there is a close previous event that would not have been detected by the trigger, the value of ÎPE (time separating current and previous events) becomes very high, and everything happens as if it were considered that the The preceding event occurred well before the current event, and thus the pace of the tail of the impulse of the preceding event becomes close to the purely exponentially decreasing pattern.

In the case where there is no previous event close before the first event, the value of the associated upstream stacking correction Co will be equal to zero and the correction, made in all cases, will have no effect. The values of the Atsi integration times used to measure the Co, Ci and C ₂ values, namely Atso, Atsi and Ats ₂ , can be selected on the fly for each event to optimize the final accuracy on the Ko calculation. , Ki and K ₂ , depending on different relevant parameters, such as the time between the current event of the previous event and the average count rate.

The low limit for the Atsi integration time values is to consider only one sample, but as previously stated, this introduces a high error due to statistical fluctuations. Integrating several samples decreases the error due to statistical fluctuations and the upper limit is to take for the Atsi values equal to the integration time of the previous events, that is, the upstream stacking correction uses the value of the previous impulse integrated since the beginning of the impulse.

However, using values close to or equal to the upper limit for Atsi leads to systematically favoring high values, but this does not improve and can even significantly degrade the accuracy of upstream stacking correction, especially when the previous event will not have been detected due to a trigger failure. In general, to be precise, the extrapolation process must use a portion of the signal as close as possible to the point from which the extrapolation is performed. The further the signal is considered from the current event, the less significant it will be for its extrapolation to the current event, and the greater the probability of reaching high values, which are consequently affected by a high level of statistical and error fluctuations, or inconsistent values.

In practice, the values of the integration times Atsi must be carefully optimized so that the signal-to-noise ratio in the measurement of C is maximal, which usually causes the Ats, to be relatively constant (whatever the counting rate) and of the same order of magnitude as the material's decay time constant scintillator, that is to say typically three. or four times less than the nominal signal integration time of an event. This also represents a significant improvement in terms of accuracy and final position and energy resolutions compared to the state of the prior art.

The function g (ti) represents the correction fraction due to the downstream stack, that is to say the correction for the missing part of the tail of the pulse, where:

ti is the integration time of the pulse.

This correction for the missing part of the tail of the pulse is obtained by multiplying the value J by a tabulated coefficient which depends on the integration time ¾. Indeed, considering a constant shape for all pulses regardless of their amplitude, which is the case for most scintillation detectors, one skilled in the art can determine the remaining part of the signal knowing the value of its integral. and the integration time, and the multiplicative coefficient is thus inversely proportional to the collection efficiencies associated with the different integration times.

It should be noted that this method makes no assumptions about the shape of the pulse, ie any form can be considered.

FIG. 4 is a block diagram of a preferred embodiment of the invention. It is a device comprising two integrators performing the integration and correction functions of the upstream and downstream stacks, and having a sliced structure so as to perform the sequence of operations described above regardless of the number of events in the salvo and their temporal distribution.

The signals indicated in FIG. 4 are described as follows:

If Samples of the input signal

Δΐ + Delay applied to the input signal before integrating the pulse of the current event by the integrator 401

Ai- Delay applied to the input signal before integration of the signal portion preceding the current event pulse by the integrator 411

ti integration time of integrator 401

Ats Integration time of integrator 411

tpE Time elapsed between the current event and the previous event (clock

"Previous Event")

I Integral of the pulse of the current event, uncorrected Integral portion of the signal preceding the impulse of the current event, uncorrected

Integral of the tail of the impulse of the previous event integrated with the pulse of the current event, corrected for the incomplete integration and the pulse shape factor related to the temporal proximity of the previous event Integral the pulse of the current event, corrected by the upstream stack of the current event, corrected for upstream and downstream stacks. As shown in FIG. 4, the digital samples of the pulse signal are simultaneously received by the FIFO delay 400 having a delay time of value At + and the FIFO delay 410 having a value delay time At- The output of the FIFO delay 400 is applied to the integrator 401, and the output of the FIFO delay 410 is applied to the integrator 411. The value Δί + of the FIFO delay 400 is adjusted so that when an event is detected by the trigger, the integrator 401 starts the integration. the pulse signal of the event, while the value Δί- of the FIFO delay 410 is adjusted so that the integrator 411 simultaneously starts in parallel the integration of the signal portion preceding the pulse signal of the event, that is to say, corresponding to the tail of a possible previous event (At- is therefore always greater than Δΐ +). The multiplier 420 receives the output of the integrator 411 and the output of a memory 403 which contains the tabulated values of the function / The input address of the memory 403 is obtained by concatenating the time ti of the integrator 401 (Current event pulse integration time), integrator time 411, and EPI time elapsed between the current event and the previous event. The PEI value is provided by a clock counting the time between events and is reset at each event. A time limiting device makes it possible to avoid an overflow of the counter when the time separating two successive events becomes too high, so that tpE is always in the interval [1, tMAx], where ÎMAX is a fixed value dependent on the capacity of the memory 403. The arithmetic unit 430 subtracts the output of the multiplier 420 from the output of the integrator 401. The multiplier 440 receives the output of the arithmetic unit 430 and the output of a memory 402 which contains the tabulated values of function g. The input address of the memory 402 is obtained by the integration time ¾ of the integrator 401.

The sequencing of the various elements (beginning and end of the integrations, clocks, resets) is performed using control signals provided by an external logic comprising the event trigger and other elements which for simplicity are not shown in FIG. 4, but which are known to those skilled in the art.

In another embodiment of the invention, if the integration time Ats signal before. the pulse of the current event is taken as constant, it disappears from the input address of the memory 403. In this case, the input address is obtained by concatenation of the only integration times t i. pulse of the current event and the time PEI elapsed between the current event and the previous event.

In another embodiment of the invention, if, for example, the chosen pulse shape model is a purely exponential decreasing function, it becomes unnecessary to take into account the form factor and the function / cease to depend on the elapsed PEI time. between current and previous events. In this case, tpE disappears from the input address of the memory 403 and this input address is obtained by the integration time ¾ of the integrator 401.

Thus, when a salvo of three events such as that described in FIG. 1 occurs, the following sequence of operations is executed:

Action Sequence

At the beginning of the integration of the event 1 detected by the trigger,. , ". , _A. , ^

r start of Co integration

end of nominal integration for

integrator 411 or end of Co integration and sending of Co downstream event 2 detected by the trigger

end of integration of Ii and sending of Ii downstream integrator reset 401 and start integration I ₂ event 2 detected by the trigger

integrator reset 411 and start integration Ci calculation of Ki

end of nominal integration for

integrator 411 or end of integration of Ci and sending of Ci downstream event 3 detected by the trigger

end of integration of I ₂ and sending of I ₂ downstream integrator reset 401 and early integration I ₃ event 3 detected by the trigger

reset integrator 411 and start integration C ₂ end of nominal integration for

end of integration of C ₂ and sending C ₂ downstream integrator 411 end of nominal integration for end of integration of ¾ and sending of I ₃ downstream integrator 401 K3 calculation

It should be noted that although the invention has been described with reference to radiation detectors based on scintillator materials associated with photomultipliers, the inventive concept can be applied to any type of apparatus delivering electronic pulse signals independently of their temporal characteristics and their form.

Thus, a method and apparatus for performing accurate stack correction in a radiation detector has been described.

Although the invention has been described in detail with reference to specific ways of carrying it out as an example, it is clear that various changes and modifications can be made without departing from the spirit and the field of application of the invention as set forth in the claims.

Claims

1. A method of correcting upstream and downstream stacking effects for a detector of

radiation delivering impulse signals in response to events from the incident radiation quanta interaction with the detector, the method comprising:

a) detecting the occurrence of an incident event;

b) integrating a signal portion preceding the pulse of the event over a predetermined time interval;

c) integrating the event pulse over a predetermined time interval or until a new event is detected;

d) applying a pre-pulse stack correction based on the integral of the signal portion preceding the event pulse and the integration time of the event pulse;

e) the application of a post-pulse stack correction based on the integration time of the event pulse.

The method of claim 1, wherein the pre-pulse stack correction

further comprises a step of measuring the time elapsed between the current event and the previous event, said correction being based on the integral of the signal portion preceding the event pulse, the integration time of the impulse of the event, and the time elapsed between the current event and the previous event.

A signal processing apparatus for a radiation detector delivering electronic pulse signals in response to events from the radiation quanta interaction incident with the detector, comprising:

a) means for detecting the occurrence of an incident event;

b) means for integrating a portion of the signal preceding the pulse of

the incident event on a predetermined time interval;

c) means for integrating the pulse of the incident event over a predetermined time interval or until a new event is detected;

d) means for performing pre-pulse stack correction based on the integral of the signal portion preceding the event pulse and the integration time of the event pulse;

e) means for performing a post-pulse stack correction based on the integration time of the event pulse.

The processing device according to claim 3, wherein said means for performing a pre-pulse stacking correction (d) further comprises a counter for measuring the time elapsed between the current event in progress. processing and the preceding event, said pre-pulse stack correction being based on the integral of the signal portion preceding the current event pulse, the event pulse integration time current, and the time elapsed between

the current event and the previous event.

5. Treatment device according to claim 3, wherein said integration means (b) and (c) comprise two distinct delays and two separate integrators operating in parallel, the first integrator integrating the signal portion preceding the first one. pulse of the event and receiving the input signals delayed by the first delay, the second integrator performing the integration of the pulse of the event and receiving the input signals delayed by the second delay.

6. Treatment device according to claim 3, wherein said means for performing the pre-pulse stack correction (d) comprise a multiplier, a memory and a subtractor, said correction being performed by multiplying with said multiplier. the integral of the signal portion preceding the pulse of the current event by the output of said memory whose address is given by the integration time of the pulse of the current event, then subtracting from using this subtractor the result of this operation of the integral of the pulse of the current event.

7. Treatment device according to claim 4, wherein said means for performing the pre-pulse stack correction (d) comprise a multiplier, a memory and a subtractor, said correction being performed by multiplying with said multiplier. the integral of the signal portion preceding the pulse of the current event by the output of said memory whose address is given by a combination of the integration time of the pulse of the current event and the time having elapsed between the current event and the previous event, then subtracting the result of this operation from the integral of the pulse of the current event by means of said subtractor.

8. Treatment device according to claim 3, wherein said means for performing the post-pulse stack correction (e) comprise a multiplier and a memory, said correction being performed by multiplying with said multiplier the integral the impulse of the event, corrected by the pre-pulse stack, by the output of said memory addressed by the integration time of the pulse of the event.