Pulse recognition method based on three-channel cross-checking

Abstract

The present application belongs to the technical field of photon counting imaging, and particularly relates to a pulse recognition method based on three-channel cross verification. The method comprises the following steps: S1: obtaining time sequence synchronized signal sequences of three channels output by a charge sensitive amplifier, and pre-processing the signal sequences of the channels to obtain respective deconvolution signals of the three channels; S2: designing a threshold based on the respective deconvolution signals of the three channels, and performing three-channel cross recognition on each pulse event by using the threshold to obtain a pile-up discrimination group and an amplitude extraction group; and S3: recording a pulse event with an amplitude to be extracted into a non-pile-up amplitude extraction group by using the relationship among a time interval of each pulse event, a pulse shaping width and an amplitude extraction width, and obtaining three-channel pulse amplitudes of each pulse event contained in the non-pile-up amplitude extraction group. The present application saves the time cost of debugging, and reduces the instability of results caused by human factors.

Description

Technical Field

[0001] This invention belongs to the field of photon counting imaging technology, and particularly relates to a pulse recognition method based on three-channel cross-validation. Background Technology

[0002] Photon counting imaging is a technique for single-photon imaging of targets with weak light intensity. A photon counting imaging system mainly includes a microchannel plate (MCP), a position-sensitive anode, a charge-sensitive amplifier circuit, a pulse shaper, pulse peak discrimination, and image inversion. The principle of photon counting imaging is as follows: after single-photon photoelectric conversion, the photoelectron cluster is multiplied by the MCP and falls onto the position-sensitive anode. The detector outputs a three-channel pulse signal through a charge-sensitive amplifier (specifically, a pulse signal with low noise and accurate amplitude is obtained after preprocessing methods such as charge-sensitive amplifier and pulse shaping filtering). The pulse amplitudes of the three channels collectively carry the information of the photoelectron cluster's centroid position. Then, pulse recognition and amplitude extraction methods are used to accurately extract the pulse signal. Finally, a specific method is used to calculate the position where the single photon falls, thus achieving single-photon imaging.

[0003] Currently, pulse recognition is usually achieved by setting a threshold and making comparisons. When a signal exceeds the set threshold, it will be recognized as a pulse.

[0004] The shortcomings of existing technologies are that the pulse-threshold comparison method cannot meet the needs of pulse recognition in environments with strong baseline fluctuations, pulse accumulation caused by high count rates, and high noise. Summary of the Invention

[0005] In view of this, the present invention aims to provide a pulse identification method based on three-channel cross-validation to solve the shortcomings of existing technologies that cannot provide accurate pulse identification results. The present invention has the functions of optimal threshold calculation, pulse identification and stacking identification, saving debugging time and reducing the instability of results caused by human error.

[0006] To achieve the above objectives, the technical solution created by this invention is implemented as follows: A pulse recognition method based on three-channel cross-validation specifically includes the following steps: S1: Obtain the signal sequences of the three channels of the charge-sensitive amplifier output with time synchronization, and preprocess the signal sequences of each channel to obtain the deconvolution signals of the three channels respectively. Each pulse event corresponds to a sampling frame of the deconvolution signal of the three channels respectively. S2: Based on the deconvolution signals of the three channels of time synchronization, design thresholds and use the thresholds to perform three-channel cross-identification of each pulse event to obtain the stacking discrimination group and amplitude extraction group. S3: Utilizing the relationship between the time interval of each pulse event, the pulse shaping width, and the amplitude extraction width, the pulse events whose amplitudes are to be extracted are recorded in the non-stacking amplitude extraction group, and the three-channel pulse amplitudes of each pulse event included in the non-stacking amplitude extraction group are obtained.

[0007] Furthermore, in step S1, if the deconvolution signals of the three channels participate in the pulse shaping experiment, and the decay time constants of the signal sequences of the three channels are known, then the deconvolution operation is performed using the decay time constants of the signal sequences of the three channels to obtain the deconvolution signals of the three channels. If the deconvolution signals of the three channels are used in the pulse shaping experiment, or if the pulse undershoot amplitude of the signal sequence of the three channels is greater than 1 / 10 of the pulse amplitude, or if the maximum value of the pulse decay time constant of the signal sequence of the three channels is at least three times the minimum value, then the signal sequences of the three channels are first subjected to pole-zero cancellation processing, and then the signals after pole-zero cancellation processing are subjected to deconvolution operation to obtain the deconvolution signals of the three channels.

[0008] Furthermore, in step S1, the deconvolution signal includes a white noise signal and an impulse signal. The white noise signal follows a Gaussian distribution, and the impulse signal follows a Poisson distribution.

[0009] Furthermore, in step S2, the threshold designed based on the deconvolution signals of the three channels with time synchronization is either a single threshold or a double threshold.

[0010] Furthermore, the distribution curves of the deconvolution signals of the three channels are obtained, and in the three distribution curves, the first minimum peak along the positive X-axis that is between the peaks of the white noise signal and the impulse signal is taken as the single threshold of the three channels.

[0011] Furthermore, the distribution curves of the deconvolution signals of the three channels are obtained, and in the three distribution curves, the first minimum peak along the positive X-axis that is between the peaks of the white noise signal and the impulse signal is taken as the high threshold of the three channels respectively. The distribution curve of the deconvolution signal is folded with the Y-axis as the axis of symmetry. The intersection point between the white noise signal in the positive X-axis direction and the impulse signal folded to the positive X-axis direction is taken as the low threshold of each of the three channels.

[0012] Furthermore, the specific process of using a single threshold to perform three-channel cross-identification of each pulse event is as follows: To keep the timing of the deconvolution signals of the three channels synchronized, and to ensure that the impulse signal of the same pulse event occurs in the same sampling frame of the three channels; For pulse events in the same sampling frame, the threshold state of the deconvolution signal that is higher than the corresponding channel single threshold is set to 1, and the threshold state of the deconvolution signal that is lower than the corresponding channel single threshold is set to 0, so as to obtain the threshold state of the current pulse event. Repeat the above operation until the threshold state of all pulse events is obtained; Pulse events with threshold states containing at least two 1s are classified into amplitude extraction groups, and pulse events with threshold states containing 1s are classified into stacking recognition groups. The location points of each pulse event included in the stacking recognition group are recorded.

[0013] Furthermore, the specific process of using dual thresholds for three-channel cross-identification of pulse events is as follows: To keep the timing of the deconvolution signals of the three channels synchronized, and to ensure that the impulse signal of the same pulse event occurs in the same sampling frame of the three channels; For pulse events in the same sampling frame, the threshold state of deconvolution signals that are higher than the corresponding channel high threshold is set to 2; the threshold state of deconvolution signals that are lower than the corresponding channel high threshold but higher than the corresponding channel low threshold is set to 1; the threshold state of deconvolution signals that are lower than the corresponding channel low threshold is set to 0; and the threshold state of the current pulse event is obtained. Repeat the above operation until the threshold state of all pulse events is obtained; The threshold state numbers marked on the three channels and pulse events less than 2 are classified into the amplitude extraction group; the threshold state numbers marked on the three channels and pulse events greater than or equal to 2 are classified into the stacking identification group, and the location points of each pulse event included in the stacking identification group are recorded.

[0014] Furthermore, the method for obtaining the location point of a pulse event is: the X-axis coordinate value of the pulse event on the corresponding deconvolution signal map.

[0015] Furthermore, step S3 specifically includes the following steps: S31: Set a Gaussian window with a fixed pulse shaping width of L and shaping parameter α; perform convolution operations between the Gaussian window and the three-channel deconvolution signals corresponding to each pulse event in the stacking recognition group to obtain the shaping pulse signals of each pulse event for each of the three channels: , ; ; in, Let n be a Gaussian window, and n be the index of the Gaussian window. Here, k represents the forming pulse signal for Gaussian window shaping, and k is the index symbol for the convolution operation. The deconvolution signal currently being used in the calculation; S32: Set a single pulse event to generate a shaping pulse signal in each of the three channels, and set the amplitude extraction width of the shaping pulse signal to M; S33: If at least one of the time intervals between the current pulse event and the adjacent pulse event is less than (M+L) / 2, then the current pulse event is marked as a stacked pulse. If the time intervals between the current pulse event and the adjacent pulse event are both greater than (M+L) / 2, and the current pulse event belongs to the amplitude extraction group, then the current pulse event is marked as a non-stacked pulse. S34: Repeat step S33 until all non-stacked pulses in the stacking identification group are obtained, and use the set of non-stacked pulses as the non-stacked pulse amplitude extraction group. S35: Based on the location point D of each pulse event in the non-stacking pulse amplitude extraction group, extract M acquisition points from D+(L-2-M) / 2+1 to D+(L-2-M) / 2+M in the shaping pulse signal of each of the three channels corresponding to each pulse event. The M value of each of the three channels is the same, and the average value of the M acquisition points of each of the three channels corresponding to each pulse event is taken as the three-channel pulse amplitude of each pulse event.

[0016] Compared with the prior art, the present invention can achieve the following beneficial effects: (1) This invention achieves automated selection of pulse recognition threshold. Compared with traditional threshold adjustment methods, this invention has higher stability in terms of issues such as false recognition and missed recognition. This invention also saves the time cost of threshold adjustment and reduces instability caused by human error.

[0017] (2) The present invention realizes the three-channel cross pulse recognition for a specific detector. Compared with the independent recognition method of each channel, the present invention subdivides the pulse into amplitude extraction group and stacking discrimination group by setting different recognition states, thereby improving the pulse recognition capability and reducing the probability of misidentification caused by high baseline fluctuation and high noise.

[0018] (3) The pulse recognition method based on deconvolution described in this invention can accurately locate the pulse time, is suitable for high count rate environments, and realizes pulse stacking discrimination by comparing with the forming width, thereby reducing the probability of misidentification caused by pulse stacking. Attached Figure Description

[0019] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A schematic flowchart of the pulse recognition method based on three-channel cross-validation described in the embodiments of the present invention; Figure 2A comparison diagram of the detector output signal and the deconvolution signal described in the embodiments of the present invention; Figure 3 The deconvolution signal distribution diagram described in the embodiments of the present invention; Figure 4 The deconvolution signal distribution folding diagram described in the embodiments of the present invention; Figure 5 The single threshold discrimination diagram described in the embodiments of the present invention; Figure 6 The dual threshold discrimination diagram described in the embodiments of the present invention; Figure 7 The dual threshold generation diagram of the S-channel described in the embodiment of the present invention; Figure 8 The dual threshold generation diagram of the W channel described in the embodiment of the present invention; Figure 9 The dual threshold generation map of the Z channel described in the embodiment of the present invention; Figure 10 The three-channel threshold discrimination diagram described in the embodiments of the present invention; Figure 11 The stacking discrimination effect diagram of the forming pulse signal and the deconvolution signal described in the embodiment of the present invention is shown. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof.

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0022] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0023] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0024] The invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0025] like Figure 1 As shown, a pulse recognition method based on three-channel cross-validation includes the following steps: S1: Obtain the time-synchronized signal sequences of the three channels output by the charge-sensitive amplifier, and preprocess the signal sequences of each channel to obtain the deconvolution signals of the three channels respectively. Each pulse event corresponds to a sampling frame of the deconvolution signals of the three channels respectively; S2: Design a threshold based on the deconvolution signals of the three channels respectively, and use the threshold to perform three-channel cross-validation on each pulse event to obtain a stacking discrimination group and an amplitude extraction group; S3: Utilize the relationship between the time interval of each pulse event, the pulse shaping width, and the amplitude extraction width, record the pulse events whose amplitudes are to be extracted into the non-stacked amplitude extraction group, and obtain the three-channel pulse amplitudes of each pulse event included in the non-stacked amplitude extraction group.

[0026] This invention proposes a pulse identification method suitable for environments with high baseline fluctuations, high count rates, and high noise levels. It is primarily designed for the three-channel output signals of a specific location-sensitive anode detector and a charge-sensitive amplifier, but can be extended to more channels (not limited to three). This invention combines deconvolution technology with a three-channel cross-identification method. It utilizes the high-pass characteristic of deconvolution to reduce the low-frequency impact of baseline fluctuations, leverages the characteristics of the impulse signal obtained from deconvolution to distinguish accumulated pulses in high count rate environments, and employs the three-channel cross-identification method to effectively distinguish between true and false pulse events, improve pulse identification rate, and reduce pulse false identification rate.

[0027] In summary, this invention mainly includes three steps: Step 1, obtaining the preprocessed deconvolution signal for each channel; Step 2, generating a corresponding threshold, which is used to identify pulse events, and performing three-channel cross-validation using the deconvolution signal to distinguish pulse events into a "stacking discrimination group" and an "amplitude extraction group"; Step 3, stacking judgment and amplitude extraction, using the relationship between the time interval of pulse events, the pulse shaping width, and the amplitude extraction width, recording the pulse events requiring amplitude extraction as the "non-stacking amplitude extraction group," and performing amplitude extraction.

[0028] In some embodiments, in step S1, if the deconvolution signals of the three channels participate in the pulse shaping experiment and the decay time constants of the signal sequences of the three channels are known, then the deconvolution operation is performed using the decay time constants of the signal sequences of the three channels to obtain the deconvolution signals of the three channels. If the deconvolution signals of the three channels are used in the pulse shaping experiment, or if the pulse undershoot amplitude of the signal sequences of the three channels (the ideal exponential pulse falls slowly to the baseline; when the falling edge falls below the baseline, it is called undershoot, and the degree to which it falls below the baseline is called undershoot amplitude) is greater than 1 / 10 of the pulse amplitude, or if the maximum value of the pulse decay time constant of the signal sequences of the three channels is at least three times the minimum value, then the signal sequences of the three channels are first subjected to pole-zero cancellation processing, and then the signals after pole-zero cancellation processing are subjected to deconvolution operation to obtain the deconvolution signals of the three channels.

[0029] It should be noted that, based on prior knowledge, the original voltage signal output by the detector and the charge-sensitive amplifier is an exponentially decaying signal, which requires preprocessing. The preprocessing includes: 1. Baseline correction using the signal during the event-free period; 2. Eliminating pulse undershoot through pole-zero phase cancellation; 3. Obtaining the deconvolution signal through differential processing or deconvolution processing.

[0030] The deconvolution signal consists of an impulse signal and a noise signal, such as Figure 2As shown, compared to the original voltage signal output by the charge-sensitive amplifier, the low frequencies of the deconvolution signal are suppressed, while high-frequency noise is amplified. The rising edge of a pulse event in the original voltage signal output by the charge-sensitive amplifier corresponds to an impulse signal in the deconvolution signal. The width of a single impulse signal is one sampling point, and the intensity of a single impulse signal is highly positively correlated with the rising edge of the corresponding pulse event in the original voltage signal. The impulse signal intensity is generally higher than that of the noise signal, but some low-amplitude impulse signals are overwhelmed by the noise signal and difficult to distinguish. Therefore, it is necessary to set an appropriate judgment threshold to distinguish impulse signals from noise signals. The choice of threshold is influenced by human experience. A high threshold can effectively improve the recognition accuracy of pulse events and reduce the false recognition rate. However, at the same time, the impulse signals of a small number of pulse events are low, and a high threshold cannot recognize them, which leads to an increased false recognition rate. In addition to reducing the pulse event counting rate, a high false recognition rate is also not conducive to the discrimination of pulse accumulation. Two pulse events that are likely to accumulate may be judged as non-accumulated because one of them is not recognized, causing the amplitude extraction of subsequent pulse events to be affected by the accumulated pulses, thus reducing the amplitude extraction accuracy. Similarly, thresholds can reduce the missed recognition rate of impulse events, but they can also identify some noise as impulse events, thus increasing the false recognition rate.

[0031] Based on the principle of single-photon pulse generation events, the deconvolution signal is considered as a combination of two signals: a "white noise signal" that follows a Gaussian distribution and an "impulse signal" that follows a Poisson distribution. The distribution curve of the combined two signals contains both a Gaussian peak and a Poisson peak, and the negative pole between the two peaks can be used as a recognition threshold. This pole has the following characteristics: the area above the pole contains a small amount of noise signal and most of the impulse signal, making it suitable as a recognition threshold.

[0032] Therefore, threshold calibration can be performed using a portion of samples from the actual testing environment before formal data collection.

[0033] There are three methods to obtain the deconvolution signal: Method 1, differential processing; Method 2, deconvolution using the time decay constant of the exponential pulse; Method 3, first performing pole-zero cancellation to reduce the undershoot effect, and then performing deconvolution using the time decay constant after pole-zero cancellation.

[0034] The choice of which deconvolution method to use depends on the design of the photon counting imaging system and the characteristics of the actual detector output signal. When the deconvolution signal does not participate in subsequent shaping, method 1 can be used directly; When the deconvolution signal participates in subsequent shaping and the decay time constant is known, the signal undershoot phenomenon is not obvious, which is an ideal deconvolution signal, and method 2 can be used; When the deconvolution signal is involved in subsequent pulse shaping experiments, or the pulse decay rate is unstable, or the pulse undershoot amplitude reaches more than 1 / 10 of the pulse amplitude, it is necessary to perform pole-zero cancellation first, and then perform deconvolution, i.e., use method 3.

[0035] In some embodiments, in step S2, the threshold designed based on the deconvolution signals of the three channels of time synchronization is a single threshold or a double threshold.

[0036] In some embodiments, the distribution curve of the deconvolution signal of any one of the three channels is obtained, and the first minimum peak along the positive X-axis, which is between the peaks of the white noise signal and the impulse signal, is taken as a single threshold. Repeat the above steps until you obtain the individual threshold values ​​for each of the three channels.

[0037] It should be noted that, as Figure 3 As shown, in the single threshold discrimination process, the distribution of the deconvolution signal is first analyzed. The symmetrical Gaussian peak that is symmetrical about x=0 is the noise peak, and the shorter bulge on the right is the impulse signal. The first minimum point between the two peaks (along the positive direction of the X-axis) is taken as the single threshold.

[0038] In some embodiments, the distribution curve of the deconvolution signal of any one of the three channels is obtained, and the first minimum peak along the positive X-axis, located between the peaks of the white noise signal and the impulse signal, is taken as the high threshold. ; like Figure 4 As shown, the distribution curve of the deconvolution signal is folded with the Y-axis as the axis of symmetry. The intersection point between the white noise signal in the positive X-axis direction and the impulse signal folded to the positive X-axis direction is taken as the low threshold. ; Repeat the above steps to obtain the high and low thresholds for each of the three channels.

[0039] It should be noted that the dual threshold includes a high threshold and a low threshold. The dual threshold discrimination method first analyzes the distribution of the deconvolution signal, and following the single threshold setting method described above, the first minimum point between the two peaks is recorded as the high threshold point. Fold the negative half (x<0) on the left side of the distribution curve about the x=0 axis to the positive half (x>0), and record the overlap between the folded curve and the original curve in the positive half as the low threshold point. .

[0040] Furthermore, if the impulse signal is below the low threshold point If the impulse signal is higher than the low threshold point, it can be considered an interference signal. Below the high threshold point If the impulse signal is higher than the high threshold point, it is considered a potentially valid event. If so, it is highly likely to be a valid event.

[0041] This invention targets a specific detector that, along with a charge-sensitive amplifier, outputs three channels of exponentially decaying signals. The amplitude of the pulse signal contains the position information of the electron cluster. Each pulse event generates its own exponentially decaying pulse in each of the three channels simultaneously. By utilizing the deconvolution signal and a discrimination threshold, the pulse events in each channel can be identified and located. The basic principle is that when the deconvolution signal of a certain channel exceeds the discrimination threshold of that channel, it is recorded as a pulse event, and the time series of the impulse signal is located.

[0042] However, excessive noise can cause photon counting imaging systems to misjudge and misidentify noise as pulse events. Therefore, this invention utilizes the comparison between three channels to reduce the probability of misidentification.

[0043] In this invention, it is assumed that the noise signal is independent in all three channels. Therefore, the probability of simultaneous misidentification across all three channels is lower than the probability of misidentification through a single channel. Based on this, the logic for cross-identification across the three channels is set. This invention designs an optimal threshold calculation method for the "deconvolution localization pulse method," which can support multiple deconvolution methods. Compared to threshold errors caused by different deconvolution methods and human factors, this invention can automatically generate the optimal threshold using the deconvolution signal. Furthermore, this threshold has the characteristic of automatically calculating the corresponding optimal threshold for different noise environments and different count rate environments. Addressing the characteristic that noise signals affect pulse recognition, this invention designs a dual threshold selection method to distinguish between accurate pulse events, suspected pulse events, and noise interference.

[0044] The following uses single threshold and double threshold to perform three-channel cross recognition.

[0045] In some embodiments, the specific process of using a single threshold to perform three-channel cross-identification of each pulse event is as follows: To keep the timing of the deconvolution signals of the three channels synchronized, and to ensure that the impulse signal of the same pulse event occurs in the same sampling frame of the three channels; For pulse events in the same sampling frame, the threshold state of the deconvolution signal that is higher than the corresponding channel single threshold is set to 1, and the threshold state of the deconvolution signal that is lower than the corresponding channel single threshold is set to 0, so as to obtain the threshold state of the current pulse event. Repeat the above operation until the threshold state of all pulse events is obtained; Pulse events with threshold states containing at least two 1s are classified into amplitude extraction groups, and pulse events with threshold states containing 1s are classified into stacking recognition groups. The location points of each pulse event included in the stacking recognition group are recorded.

[0046] In some embodiments, the specific process of using dual thresholds to perform three-channel cross-identification of pulse events is as follows: To keep the timing of the deconvolution signals of the three channels synchronized, and to ensure that the impulse signal of the same pulse event occurs in the same sampling frame of the three channels; For pulse events in the same sampling frame, the threshold state of deconvolution signals that are higher than the corresponding channel high threshold is set to 2; the threshold state of deconvolution signals that are lower than the corresponding channel high threshold but higher than the corresponding channel low threshold is set to 1; the threshold state of deconvolution signals that are lower than the corresponding channel low threshold is set to 0; and the threshold state of the current pulse event is obtained. Repeat the above operation until the threshold state of all pulse events is obtained; The threshold state numbers marked on the three channels and pulse events less than 2 are classified into the amplitude extraction group; the threshold state numbers marked on the three channels and pulse events greater than or equal to 2 are classified into the stacking identification group, and the location points of each pulse event included in the stacking identification group are recorded.

[0047] To gain a detailed understanding of the single-threshold three-channel cross-recognition method and the double-threshold three-channel cross-recognition method, the two three-channel recognition methods will be described in detail below.

[0048] First, we introduce a single-threshold three-channel cross-identification method: the three-channel deconvolutioned signals are time-synchronized to ensure that the impulse signals of valid events occur at the same time. The deconvolutioned signals are observed on an oscilloscope, and manual calibration is performed. After calibration, the impulse signals caused by the same pulse event occur in the same sampling frame.

[0049] Then, threshold discrimination is performed within the same sampling frame. If the deconvolution signal of each channel is higher than the single threshold T of the corresponding channel, it is recorded as 1; if it is lower than the single threshold, it is recorded as 0. After threshold discrimination, there are a total of 4 states: 000 is the state with no event, 001 is the state with one channel recognizing the event, 011 is the state with two channels recognizing the event, and 111 is the state with all three channels recognizing the event.

[0050] like Figure 5 As shown, states 001, 011, and 111 are recorded in the stacking identification group, while states 011 and 111 are recorded in the amplitude extraction group. The stacking identification group includes events from the amplitude extraction group, making the stacking determination of pulse signals more stringent and effectively reducing stacking interference. In other words, by utilizing the mutual comparison of the three channels, noise interference and the probability of false identification are effectively reduced.

[0051] If some valid events are filtered out due to insufficient impulse signals and thus fail to participate in imaging, the number of valid events can be compensated for by increasing the exposure time of the light source, based on the principle of single-photon imaging. Therefore, ensuring signal accuracy should be prioritized over signal quantity.

[0052] The following describes a three-channel cross-identification method with dual thresholds: The discrimination result of a single channel is divided into three states, where the threshold value is lower than the lower threshold. Recorded as state 0, above the low threshold And below the high threshold This is recorded as state 1, which is above the high threshold. This is denoted as state 2. The three-channel threshold recognition output has a total of 10 states: state 000 represents the no-event state, state 222 represents the confirmed pulse event state, and states 001, 002, 011, 111, 012, 112, 022, and 122 are set as quasi-confirmed states with progressively increasing certainty. Different state groups are set as accumulation recognition groups and amplitude extraction groups, and it must be ensured that the accumulation recognition group includes the amplitude extraction group.

[0053] like Figure 6 As shown, this invention recommends using states 002, 011, 111, 012, 112, 022, 122 and 222 as a stacking identification group, and states 122 and 222 as an amplitude extraction group.

[0054] In practical applications, the ranges of the stacking identification group and amplitude extraction group can be appropriately adjusted based on the noise intensity and pulse count rate of the test environment to further optimize the identification logic and balance the relationship between pulse recognition rate and false recognition rate. For example, when the noise intensity is less than 1 / 10 of the average pulse amplitude, the amplitude extraction group can be expanded to states 222, 122, 022, and 112. When the pulse count rate is less than 1 / 20 of the sampling frequency divided by the pulse width, it is a low count rate environment, and the stacking identification group can be limited to states 222, 122, 022, 112, 012, 111, and 011. Conversely, when the pulse count rate is greater than 1 / 10 of the sampling frequency divided by the pulse width, the stacking identification group can be expanded to states 222, 122, 022, 112, 012, 111, 011, 002, and 001.

[0055] In some embodiments, the method for obtaining the location point of a pulse event is to obtain the X-axis coordinate value of the pulse event on the corresponding deconvolution signal map.

[0056] In some embodiments, step S3 specifically includes the following steps: Step S3 specifically includes the following steps: S31: Set a Gaussian window with a fixed pulse shaping width of L and shaping parameter α; perform convolution operations between the Gaussian window and the three-channel deconvolution signals corresponding to each pulse event in the stacking recognition group to obtain the shaping pulse signals of each pulse event for each of the three channels: , ; ; in, Let n be a Gaussian window, and n be the index of the Gaussian window. Here, k represents the forming pulse signal for Gaussian window shaping, and k is the index symbol for the convolution operation. The deconvolution signal currently being used in the calculation; S32: Set a single pulse event to generate a shaping pulse signal in each of the three channels, and set the amplitude extraction width of the shaping pulse signal to M; S33: If at least one of the time intervals between the current pulse event and the adjacent pulse event is less than (M+L) / 2, then the current pulse event is marked as a stacked pulse. If the time intervals between the current pulse event and the adjacent pulse event are both greater than (M+L) / 2, and the current pulse event belongs to the amplitude extraction group, then the current pulse event is marked as a non-stacked pulse. S34: Repeat step S33 until all non-stacked pulses in the stacking identification group are obtained, and use the set of non-stacked pulses as the non-stacked pulse amplitude extraction group. S35: Based on the location point D of each pulse event in the non-stacking pulse amplitude extraction group, extract M acquisition points from D+(L-2-M) / 2+1 to D+(L-2-M) / 2+M in the shaping pulse signal of each of the three channels corresponding to each pulse event. The M value of each of the three channels is the same, and the average value of the M acquisition points of each of the three channels corresponding to each pulse event is taken as the three-channel pulse amplitude of each pulse event.

[0057] The stacking discrimination of the present invention is mainly aimed at the forming method with fixed pulse width, that is, the pulse forming width of each pulse after forming is a preset fixed value L, where L is a positive integer.

[0058] The amplitude extraction method of this invention involves weighted summation of the central region at the top of the formed pulse. The amplitude extraction width is the number of sampling points at the center of the formed pulse participating in amplitude extraction. A common method is to use the average amplitude of the central region at the top of the pulse as the amplitude, and the central region at the top is the amplitude extraction width, which is M. Wherein, if L is odd, then M is odd; if L is even, then M is even.

[0059] Stacking determination: Analyze the pulse timing of the stacking determination group in step two. If at least one of the timing intervals between the current pulse and the two adjacent pulses is less than (M+L) / 2 (rounded up), then the current pulse is marked as a stacked pulse. If the timing intervals between the current pulse and the two adjacent pulses are both greater than (M+L) / 2 (rounded up), then the current pulse is marked as a non-stacked pulse. This non-stacked pulse (all pulse events that need amplitude extraction) belongs to the non-stacked amplitude extraction group and is used for subsequent amplitude extraction.

[0060] Amplitude extraction: Based on the stacking discrimination, the positioning point D of the impulse signal corresponding to the pulse event of the non-stacked amplitude extraction group is obtained. M acquisition points from D+(L-2-M) / 2+1 to D+(L-2-M) / 2+M in each channel shaping signal are extracted, and the average value of the M acquisition points is used as the amplitude of the pulse event. The three-channel pulse amplitude of the same event is obtained. After obtaining and recording the three-channel pulse amplitude of all non-stacked amplitude extraction groups, subsequent image restoration and other processing procedures are performed.

[0061] Example 1 A photon counting imaging system was constructed, comprising a light source, a detector, a charge-sensitive amplifier, a data acquisition card, and a computer, arranged sequentially. The workflow is as follows: The light source emits weak single photons that strike a microchannel plate. After photomultiplication, photoelectron clusters are output. These clusters land on the anode of a two-dimensional position-sensitive detector, which generates pulsed electrical signals in three channels. Each channel's pulsed electrical signal carries the position information of the photoelectron cluster. These pulsed electrical signals are then amplified by the charge-sensitive amplifier and output to the data acquisition card. The acquired data from the three channels is then transferred to the computer for processing.

[0062] sampling frequency of the acquisition card The sampling period is The acquisition time was 0.2s, and the signal sequence was... Based on the three channels, they are respectively denoted as , and Here, n is the index within a single channel signal sequence. For ease of description, the three channels can be referred to as the S channel, W channel, and Z channel, respectively. First, the signal sequences of each channel undergo pole-zero cancellation processing: ; in, This is a zero-pole cancellation signal. The first time constant, The second time constant is, and the S channel is... , The W channel is , The Z channel is , ,Will , and Substitute the formulas sequentially This allows us to obtain the zero signals of each channel.

[0063] The deconvolution operation is performed on the pole-zero cancellation signals of each channel, and the recursive formula is as follows:

[0064] ; ; in, For deconvolution signals, parameters and parameters The exponential decay time constant of the zero-pole destructive signal for each channel obtained from the fitting. and exponential decay time constant We obtain, where the S channel is... , ; W channel , Z-channel , .

[0065] By controlling the size of the aperture between the light source and the detector, the amount of light supplied to the detector can be controlled, allowing a single channel to have a roughly estimated light intensity within 0.2 seconds. ~ The number of pulse events. Then deconvolve the signals of each of the three channels. As samples, perform distribution plot analysis separately. For example... Figures 7-9 As shown, the distribution along the left negative axis of the distribution chart is folded to the right positive axis (blue line), and the intersection of the two is recorded as the low threshold of that channel. The first pole of the positive semi-axis is denoted as the high threshold of the channel. .

[0066] The threshold values ​​for each channel are as follows: , , , , , .

[0067] The above operations yield high and low thresholds for each channel, which are then used to discriminate the deconvolution signals of each channel. These high and low thresholds can also be applied to new signal segments under the same operating conditions. This invention uses the original sample's signal segment as input to the threshold discrimination system, obtains the discriminated state timing of the three channels within a 0.2s signal segment, and analyzes the state.

[0068] Figure 10 This is a threshold discrimination process for instance samples with 3000 sampling points. In event a, the deconvolution signal of the S channel falls between the high and low thresholds, the deconvolution signal of the W channel falls above the high threshold, and the deconvolution signal of the Z channel falls above the high threshold. Therefore, the state at this sampling moment is 122, and it is recorded in both the stacking discrimination group and the amplitude extraction group. Similarly, the state at the sampling moment corresponding to event b is 112, and it is recorded in the stacking discrimination group. The state at the sampling moment corresponding to event c is 002, and it is recorded in the stacking discrimination group.

[0069] The results show that there are 74,940 stacking discrimination groups and 64,749 amplitude extraction groups. The timing of each event in the stacking discrimination group is also recorded.

[0070] This invention uses a fixed pulse forming width L=75, forming parameters The method for forming a Gaussian window. A Gaussian window is: , n is the index of the Gaussian window. The Gaussian window is convolved with the deconvolution signal to obtain the shaping pulse signal. .

[0071] Set amplitude extraction width That is, only the peak center vertex is used as the amplitude, and its stacking discrimination effect is as follows: Figure 11 As shown, a, b, c, d, e, f, and g represent different pulse events. Event a has a state of 222 and belongs to both the amplitude extraction group and the stacking discrimination group. The location coordinates of event a are... =64, location coordinates of event b =204, the sampling interval between the impulse signals of events a and b. 140, greater than (M+L) / 2=38, is determined to be a non-stacking pulse and participates in the extraction of the actual amplitude. The extracted amplitude is the mean of M acquisition points from the time series D+(L-2-M) / 2+1 to D+(L-2-M) / 2+M of the shaped pulse signal, i.e. Figure 11 In the figure, the value of the forming pulse signal at the horizontal coordinate 101 is such that the amplitudes of the three channels of event a are 0.103101, 0.225218, and 0.157315 respectively.

[0072] Similarly, the three-channel amplitudes of event b can be obtained. The location coordinates of events c and d are respectively... 248, 258, with a sampling interval of 10, is less than 38, and is therefore classified as a stacking event, meaning events c and d do not participate in amplitude extraction. Event f has a state of 211, does not participate in actual amplitude extraction, but participates in stacking detection. The interval between event e and event f is 49, which is greater than the interval of 38, so event e participates in amplitude extraction. Ultimately, the number of events participating in amplitude extraction is 21034, with each event having amplitude values ​​from three channels.

[0073] This invention utilizes a single or double threshold method to identify pulse events, dividing pulse events into "accumulation discrimination group," "amplitude extraction group," and "non-accumulation amplitude extraction group," thereby achieving effective identification of accumulation events and reducing amplitude extraction errors caused by accumulation.

[0074] The stacking discrimination logic of the present invention uses the width of a single sampling point of the deconvolution signal to realize pulse positioning, obtain the pulse positioning value D, and then determines the range of amplitude extraction by using the amplitude extraction width M and the pulse shaping width L. It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0075] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A pulse recognition method based on three-channel cross-validation, characterized in that: Specifically, the steps include the following: S1: Obtain the signal sequences of the three channels of the charge-sensitive amplifier output with time synchronization, and preprocess the signal sequences of each channel to obtain the deconvolution signals of the three channels respectively. Each pulse event corresponds to a sampling frame of the deconvolution signal of the three channels respectively. In step S1, if the deconvolution signals of the three channels participate in the pulse shaping experiment and the decay time constants of the signal sequences of the three channels are known, then the deconvolution operation is performed using the decay time constants of the signal sequences of the three channels to obtain the deconvolution signals of the three channels. If the deconvolution signals of the three channels participate in the pulse shaping experiment, or the pulse undershoot amplitude of the signal sequence of the three channels is greater than 1 / 10 of the pulse amplitude, or the maximum value of the pulse decay time constant of the signal sequence of the three channels is at least three times the minimum value, then the signal sequence of the three channels is first subjected to pole-zero cancellation processing, and then the signal after pole-zero cancellation processing is subjected to deconvolution operation to obtain the deconvolution signals of the three channels. S2: Design thresholds based on the deconvolution signals of the three channels of time synchronization, and use the thresholds to perform three-channel cross-identification of each pulse event to obtain the stacking discrimination group and amplitude extraction group; S3: Utilizing the relationship between the time interval of each pulse event, the pulse shaping width, and the amplitude extraction width, the pulse events whose amplitudes are to be extracted are recorded in the non-stacking amplitude extraction group, and the three-channel pulse amplitudes of each pulse event included in the non-stacking amplitude extraction group are obtained.

2. The pulse recognition method based on three-channel cross-validation according to claim 1, characterized in that: In step S1, the deconvolution signal includes a white noise signal and an impulse signal, wherein the white noise signal follows a Gaussian distribution and the impulse signal follows a Poisson distribution.

3. The pulse recognition method based on three-channel cross-validation according to claim 2, characterized in that: In step S2, the threshold designed based on the deconvolution signals of the three channels with time synchronization is either a single threshold or a double threshold.

4. The pulse recognition method based on three-channel cross-validation according to claim 3, characterized in that: Obtain the distribution curves of the deconvolution signals for each of the three channels. In the three distribution curves, the first minimum peak along the positive X-axis that is between the peaks of the white noise signal and the impulse signal is taken as the single threshold for each of the three channels.

5. The pulse recognition method based on three-channel cross-validation according to claim 3, characterized in that: Obtain the distribution curves of the deconvolution signals of the three channels respectively, and in the three distribution curves, take the first minimum peak along the positive X-axis that is between the peaks of the white noise signal and the impulse signal as the high threshold of the three channels respectively. The distribution curve of the deconvolution signal is folded with the Y-axis as the axis of symmetry. The intersection point between the white noise signal in the positive X-axis direction and the impulse signal folded to the positive X-axis direction is taken as the low threshold of each of the three channels.

6. The pulse recognition method based on three-channel cross-validation according to claim 4, characterized in that: The specific process of using a single threshold to perform three-channel cross-identification of each pulse event is as follows: To keep the timing of the deconvolution signals of the three channels synchronized, and to ensure that the impulse signal of the same pulse event occurs in the same sampling frame of the three channels; For pulse events in the same sampling frame, the threshold state of the deconvolution signal that is higher than the corresponding channel single threshold is set to 1, and the threshold state of the deconvolution signal that is lower than the corresponding channel single threshold is set to 0, so as to obtain the threshold state of the current pulse event. Repeat the above operation until the threshold state of all pulse events is obtained; Pulse events with threshold states containing at least two 1s are classified into amplitude extraction groups, and pulse events with threshold states containing 1s are classified into stacking recognition groups. The location points of each pulse event included in the stacking recognition group are recorded.

7. The pulse recognition method based on three-channel cross-validation according to claim 5, characterized in that: The specific process of using dual thresholds to perform three-channel cross-identification of pulse events is as follows: To keep the timing of the deconvolution signals of the three channels synchronized, and to ensure that the impulse signal of the same pulse event occurs in the same sampling frame of the three channels; For pulse events in the same sampling frame, the threshold state of deconvolution signals that are higher than the corresponding channel high threshold is set to 2; the threshold state of deconvolution signals that are lower than the corresponding channel high threshold but higher than the corresponding channel low threshold is set to 1; the threshold state of deconvolution signals that are lower than the corresponding channel low threshold is set to 0; and the threshold state of the current pulse event is obtained. Repeat the above operation until the threshold state of all pulse events is obtained; The threshold state numbers marked on the three channels and the pulse events less than 2 are divided into amplitude extraction groups; The threshold state numbers of the three channels and pulse events greater than or equal to 2 are divided into stacking identification groups, and the location points of each pulse event included in the stacking identification group are recorded.

8. The pulse recognition method based on three-channel cross-validation according to claim 6 or 7, characterized in that: The method for obtaining the location point of a pulse event is: the X-axis coordinate value of the pulse event on the corresponding deconvolution signal graph.

9. The pulse recognition method based on three-channel cross-validation according to claim 4, characterized in that: Step S3 specifically includes the following steps: S31: Set a Gaussian window with a fixed pulse shaping width of L and shaping parameter α; perform convolution operations between the Gaussian window and the three-channel deconvolution signals corresponding to each pulse event in the stacking recognition group to obtain the shaping pulse signals of each pulse event for each of the three channels: ， ； ； in, Let n be a Gaussian window, and n be the index of the Gaussian window. Here, k represents the forming pulse signal for Gaussian window shaping, and k is the index symbol for the convolution operation. The deconvolution signal currently being used in the calculation; S32: Set a single pulse event to generate a shaping pulse signal in each of the three channels, and set the amplitude extraction width of the shaping pulse signal to M; S33: If at least one of the time intervals between the current pulse event and the adjacent pulse event is less than (M+L) / 2, then the current pulse event is marked as a stacked pulse. If the time intervals between the current pulse event and the adjacent pulse event are both greater than (M+L) / 2, and the current pulse event belongs to the amplitude extraction group, then the current pulse event is marked as a non-stacked pulse. S34: Repeat step S33 until all non-stacked pulses in the stacking identification group are obtained, and use the set of non-stacked pulses as the non-stacked pulse amplitude extraction group. S35: Based on the location point D of each pulse event in the non-stacking pulse amplitude extraction group, extract M acquisition points from D+(L-2-M) / 2+1 to D+(L-2-M) / 2+M in the shaping pulse signal of each of the three channels corresponding to each pulse event. The M value of each of the three channels is the same, and the average value of the M acquisition points of each of the three channels corresponding to each pulse event is taken as the three-channel pulse amplitude of each pulse event.

Images