PET imaging methods, electronic devices and storage media for multi-field particle therapy

Abstract

This application discloses a PET imaging method, electronic device, and storage medium for multi-field particle therapy. The PET imaging method includes: identifying one or more time intervals for each field based on the temporal distribution characteristics of acquired single-event data, with each time interval corresponding to a particle beam delivery state; performing conformation processing on the single-event data to generate conformation event data; and reconstructing images from the conformation event data in the identified time intervals of each field to obtain reconstructed images corresponding to each field. Through intelligent identification of temporal distribution characteristics and a partitioned reconstruction strategy, the method of this application solves the problem of severe aliasing of PET signals in multi-field delivery modes, achieving accurate extraction and image reconstruction of the independent dose distribution of each field.

Description

Technical Field

[0001] This application relates to the field of medical imaging, specifically to PET imaging methods, electronic devices, and storage media for multi-field particle therapy. Background Technology

[0002] Rays can travel through space and matter in the form of waves or particles, releasing energy in the process. In medical applications, this property is used to kill tumor cells. Currently, particle radiotherapy, with proton therapy as a typical example, has been successfully applied in the field of cancer treatment and has become one of the most advanced radiotherapy methods. In clinical practice, it is often figuratively referred to as a "proton knife."

[0003] Proton therapy, as a precision radiotherapy technique, possesses unique dose distribution characteristics. When a proton beam travels through matter, its energy deposition follows a curve that initially decreases and then increases with depth, ultimately reaching a peak at a specific depth—a phenomenon known as the "Bragg peak" effect. Utilizing this physical property, a proton beam can release most of its energy at a specific depth and then rapidly decay, thereby achieving precise irradiation of tumor lesions while minimizing damage to surrounding healthy tissues. Thanks to this advantage, proton therapy is widely recognized as a key means of achieving precision medicine in the field of radiotherapy.

[0004] However, the clinical application of proton therapy faces the challenge of range uncertainty. Factors such as patient anatomical differences, tissue density heterogeneity, and physiological movements during treatment (e.g., respiration, organ displacement) can cause the actual dose distribution to deviate from the treatment plan. Under-delivery leads to incomplete tumor killing, while over-delivery exposes critical tissues and organs to additional radiation, threatening the safety and effectiveness of treatment. To address this challenge, real-time monitoring technologies have become a research focus. Among these, positron emission tomography (PET) is considered a key method for in-beam proton therapy monitoring because it can non-invasively locate the distribution of positron-emitting nuclides induced by the proton beam. In-beam PET, using a simultaneous treatment beam and imaging system, can dynamically capture transient positron-emitting nuclides (such as...) generated by the interaction between the proton beam and tissue. 15 O, ¹¹C), thus providing three-dimensional information with high spatiotemporal resolution for range verification.

[0005] In clinical practice, a multi-field delivery mode has been proposed, which involves irradiating the tumor from different angles using multiple proton beams to cover different areas of the tumor. This complex delivery mode leads to severe overlap of PET signals—positrons in adjacent energy layers within the field overlap spatially, and the cross-irradiation of multiple fields further exacerbates the signal aliasing.

[0006] However, existing PET in-beam monitoring techniques are limited by traditional image reconstruction algorithms, which cannot effectively address the signal aliasing problem caused by multi-field delivery modes. These algorithms struggle to effectively distinguish the independent contributions of each field and energy layer to the overall signal. This limitation not only hinders the independent verification of dose distribution in each irradiation field but also restricts the dynamic adjustment and adaptive optimization of treatment plans.

[0007] Therefore, there are significant shortcomings in the current methods for processing PET monitoring signals in multi-field delivery modes of particle therapy.

[0008] The background description is provided solely for the purpose of understanding the relevant technologies in this field and is not intended as an admission of prior art. Summary of the Invention

[0009] Therefore, this application intends to provide a PET imaging method, electronic device, and storage medium for multi-field particle therapy, which can at least partially solve the signal aliasing problem caused by the above-mentioned multi-field cross-irradiation.

[0010] In a first aspect, a PET imaging method for multi-field particle therapy is provided, the PET imaging method comprising:

[0011] Based on the temporal distribution characteristics of the collected single-event data, one or more time intervals for each firing field are identified, and each time interval corresponds to a particle beam delivery state.

[0012] The single event data is processed to generate conforming event data;

[0013] Image reconstruction is performed on the matching event data in the identified time interval of each shooting field to obtain the reconstructed image corresponding to each shooting field.

[0014] In some embodiments, identifying one or more time intervals for each firing field based on the temporal distribution characteristics of the collected single-event data includes:

[0015] The distribution of single-event data for each shooting field over time is statistically analyzed, and a single-event statistical histogram for each shooting field is generated.

[0016] Determine the intersection point between the set counting threshold and the single-event statistical histogram of each shooting field;

[0017] The intersection points determine one or more time intervals for each firing field.

[0018] In some embodiments, determining the one or more time intervals for each firing field based on the intersection point includes:

[0019] The time interval between the initial intersection point and the final intersection point is defined as the in-beam interval for each field.

[0020] In some embodiments, determining the one or more time intervals for each firing field based on the intersection point includes:

[0021] The time interval between the intersection of odd-numbered indices and the subsequent intersection of even-numbered indices is defined as the high-count-value time interval of the bundle interval.

[0022] In some embodiments, determining the one or more time intervals for each firing field based on the intersection point includes:

[0023] The time period after the final intersection point in each field is defined as the beam separation interval of each field.

[0024] In some embodiments, the identified time interval is included in the bundle interval;

[0025] The process of reconstructing images from the matching event data within the identified time interval of each shooting field to obtain the reconstructed image corresponding to each shooting field includes:

[0026] Image reconstruction is performed on the coincidence event data in the in-bundle interval of each field to obtain the in-bundle reconstructed image corresponding to each field.

[0027] In some embodiments, the bundle interval includes multiple periods of high count values;

[0028] The step of reconstructing the coincidence event data in the in-beam interval of each firing field to obtain the in-beam reconstructed image corresponding to each firing field includes:

[0029] Extract coincidence event data from multiple high count periods within the in-beam interval of each field and merge them to obtain the first coincidence event set for each field;

[0030] Image reconstruction is performed using the first coincidence event set of each field to obtain the in-beam reconstructed image of each field.

[0031] In some embodiments, the identified time interval further includes a beam-off interval;

[0032] The process of reconstructing images from the matching event data within the identified time interval of each shooting field to obtain the reconstructed image corresponding to each shooting field includes:

[0033] Decay correction processing is performed on the time-distributed count data of coincidence events in the off-beam interval of each field to obtain the decay correction factor.

[0034] Image reconstruction is performed using coincidence event data and corresponding decay correction factors in the off-beam intervals of each field to obtain the off-beam reconstructed image for each field.

[0035] In some embodiments, the attenuation correction processing of coincidence event data in the off-beam intervals of each beam field includes:

[0036] Based on prior information, the various nuclide categories corresponding to the coincidence event data in the beam-off interval of each field are obtained;

[0037] Based on the time-distributed count data of coincidence events in each beam-off interval of the beam field and the various nuclide categories, a multi-exponential fitting was performed to obtain the sum of the initial activities of all nuclides and the sum of the decay activities of all nuclides.

[0038] The decay correction factor corresponding to each coincidence event in the off-beam interval of each field is determined based on the sum of the initial activities of all nuclides and the sum of the decay activities of all nuclides.

[0039] In some embodiments, the step of performing multi-exponential fitting based on the time-distributed count data of coincidence events in each beam-off interval and the nuclide category to obtain the sum of the initial activities of all nuclides and the sum of the decay activities of all nuclides includes:

[0040] A multi-exponential decay model corresponding to the various nuclide categories is established, wherein the parameters of the multi-exponential decay model include the initial activity coefficient and decay coefficient of each nuclide category;

[0041] Using the time-distributed count data of the coincident events, a multi-exponential fit is performed based on the multi-exponential decay model to determine at least the initial activity coefficient of each nuclide class at the reference time.

[0042] Calculate the sum of the initial activities of all the nuclides based on the initial activity coefficients of each nuclide class at the reference time;

[0043] Using a multi-exponential decay model that includes the determined initial activity coefficient, the sum of the decay activities of all nuclides at any given time is calculated for each coincidence event.

[0044] In some embodiments, the decay coefficient of each nuclide class is determined based on the half-life of that nuclide class;

[0045] Using the time-distributed count data of the coincident events, a multi-exponential fit is performed based on the multi-exponential decay model to determine at least the initial activity coefficients of each nuclide class at the reference time, including:

[0046] Using the time-distributed count data of the matching events as input, the linear least squares method is used for multi-exponential fitting to estimate the optimal square approximation to obtain the initial activity coefficient of each nuclide class at the reference time.

[0047] In some embodiments, using time-distributed count data of the coincident events, a multi-exponential fit is performed based on the multi-exponential decay model to at least determine the initial activity coefficients of each nuclide class at a reference time, including:

[0048] Using the time-distributed count data of the coincident events as input, a nonlinear least squares method is used for multi-exponential fitting to estimate the optimal square approximation to obtain the initial activity coefficient and decay coefficient of each nuclide class at the reference time.

[0049] In some embodiments, using time-distributed count data of the coincident events as input, a nonlinear least squares method is employed for multi-exponential fitting to estimate the optimal square approximation to obtain the initial activity coefficient and decay coefficient of each nuclide class at the reference time, including:

[0050] When performing multi-index fitting, the decay coefficients of each nuclide category are restricted to a preset constraint range.

[0051] In some embodiments, the step of reconstructing the image from the matching event data in the identified time interval of each shooting field to obtain the reconstructed image corresponding to each shooting field further includes:

[0052] Before performing decay correction, random correction and / or background correction are performed on the coincidence event data of each beam off-beam interval.

[0053] In some embodiments, the random correction of the coincidence event data of each beamout interval includes: performing statistical processing on the coincidence event data of at least each beamout interval to form a first event count set based on time distribution; performing delay processing and statistical processing on the coincidence event data of at least each beamout interval to form a second event count set based on time distribution; and performing random correction based on the difference between the first event count set and the second event count set.

[0054] In some embodiments, the step of statistically processing the coincidence event data of each beamout interval to form a first event count set based on time distribution includes: at least binning the coincidence event data of each beamout interval at equal time intervals according to a preset duration to obtain a first bin set including multiple first bins, wherein each first bin includes coincidence event counts falling within the time range of its respective first bin; the step of delaying and statistically processing the coincidence event data of each beamout interval to form a second event count set based on time distribution includes: at least delaying the coincidence event data of each beamout interval to obtain delayed coincidence event data; binning the delayed coincidence event data at equal time intervals according to the preset duration to obtain a second bin set including multiple second bins, wherein each second bin includes delayed coincidence event counts falling within the time range of its respective second bin, the number of first bins is equal to the number of second bins and the binning times correspond one-to-one.

[0055] In some embodiments, the random correction based on the difference between the first event count set and the second event count set includes: subtracting the delayed compliance event count of the corresponding second box in the second box set from the compliance event count of each first box in the first box set to obtain the first difference count of each third box that is randomly corrected.

[0056] In some embodiments, the background correction of the coincidence event data of each beamout interval includes: performing statistical processing on the coincidence event data of at least each beamout interval to form a first event count set based on time distribution; performing delay processing and statistical processing on the coincidence event data of at least each beamout interval to form a second event count set based on time distribution; and performing background correction based on the difference between the first event count set and the second event count set.

[0057] In some embodiments, the step of statistically processing the coincidence event data of each beamout interval to form a first event count set based on time distribution includes: at least binning the coincidence event data of each beamout interval at equal time intervals according to a preset duration to obtain a first bin set including multiple first bins, wherein each first bin includes coincidence event counts falling within the time range of its respective first bin; the step of delaying and statistically processing the coincidence event data of each beamout interval to form a second event count set based on time distribution includes: at least delaying the coincidence event data of each beamout interval to obtain delayed coincidence event data; binning the delayed coincidence event data at equal time intervals according to the preset duration to obtain a second bin set including multiple second bins, wherein each second bin includes delayed coincidence event counts falling within the time range of its respective second bin, the number of first bins is equal to the number of second bins and the binning times correspond one-to-one.

[0058] In some embodiments, the background correction based on the difference between the first event count set and the second event count set includes: subtracting the delayed compliance event count of the corresponding second bin from the compliance event count of each first bin in the first bin set to obtain a first difference count for each third bin; determining the minimum value among the first difference counts of each third bin as the background level; and subtracting the background level from the compliance event count of each first bin in the first bin set to obtain a second difference count for each fourth bin that has been background corrected.

[0059] In some embodiments, the background correction based on the difference between the first event count set and the second event count set includes: removing the bins before the start time of each beamout interval and using the remaining bin set for subsequent decay correction processing.

[0060] In some embodiments, the background correction based on the difference between the first event count set and the second event count set further includes: converting the start time of each bin in the remaining bin set into a relative start time relative to the start time of each beam separation interval of the field, forming a relative time vector; counting the bins in the remaining bin set after random correction and / or background correction, forming an observation vector, wherein the relative time vector and the observation vector are used for subsequent multi-exponential fitting.

[0061] In some embodiments, the step of using coincidence event data and corresponding decay correction factors in the off-beam interval of each radiation field to perform image reconstruction and obtain the off-beam reconstructed image corresponding to each radiation field includes:

[0062] Obtain the decay correction factor at the start time of the off-beam interval of each field, use the corresponding decay correction factor to correct the activity of each nuclide corresponding to the coincidence event in the off-beam interval of each field, and reconstruct the first off-beam reconstruction image based on the coincidence event of the off-beam interval of each field after correction.

[0063] In some embodiments, the step of using coincidence event data and corresponding decay correction factors in the off-beam interval of each radiation field to perform image reconstruction and obtain the off-beam reconstructed image corresponding to each radiation field includes:

[0064] For the first field of fire, the decay correction factor at the start of the off-beam interval of the first field of fire is obtained. The decay correction factor is used to correct the activity of each nuclide corresponding to the coincidence event in the off-beam interval of the first field of fire. Based on the corrected coincidence event, the second off-beam reconstruction image of the first field of fire is reconstructed.

[0065] For any subsequent field s, the decay correction factor at the start time of the beamout interval of the s-th field is obtained. The decay correction factor is used to correct the activity of each nuclide corresponding to the coincidence event in the beamout interval of the s-th field. Based on the corrected coincidence event in the beamout interval of the s-th field, the first intermediate reconstructed image of the s-th field is reconstructed. The decay correction factor is used to correct the activity of each nuclide corresponding to the coincidence event in the beamout interval of the (s-1)-th field. Based on the corrected coincidence event in the beamout interval of the (s-1)-th field, the second intermediate reconstructed image of the (s-1)-th field is reconstructed. The second beamout reconstructed image of the s-th field is obtained according to the difference between the first intermediate reconstructed image and the second intermediate image.

[0066] In some embodiments, the PET imaging method further includes: acquiring single-event data generated by a PET detector during multi-field particle therapy, the single-event data including the time, energy, and location information of the event.

[0067] In a second aspect, an electronic device is also provided, which may include: a processor and a memory storing a computer program, the processor being configured to implement the methods described in the embodiments of this application when running the computer program.

[0068] In a third aspect, a storage medium is also provided, the storage medium storing a computer program configured to be executed to implement the methods described in any embodiment of the present application.

[0069] The PET imaging method for multi-field particle therapy provided in this application identifies one or more time intervals for each field based on the temporal distribution characteristics of the acquired single-event data, where each time interval corresponds to a particle beam delivery state; performs conformance processing on the single-event data to generate conformance event data; and reconstructs the image of the conformance event data in the identified time intervals of each field to obtain the reconstructed image corresponding to each field. Thus, by identifying the temporal distribution characteristics, this application can accurately divide different delivery states of each field, such as the in-beam interval, and adopt corresponding reconstruction strategies for the identified time intervals. This technical solution solves the problem of severe aliasing of PET signals in multi-field delivery mode, thereby achieving accurate extraction of the dose distribution of a single field, providing the possibility for dynamic adjustment and adaptive optimization of treatment plans, and thus improving the accuracy of particle, such as proton therapy, range verification.

[0070] A further embodiment of this application employs a decay correction technique based on multi-exponential fitting to further process the off-beam interval. Specifically, a multi-exponential decay model corresponding to multiple nuclide categories is established, and the time distribution count data of coincidence events in the off-beam interval are used for fitting to determine the initial activity coefficient of each nuclide category at the reference time, and then the decay correction factor is calculated. This decay correction factor is used to reconstruct the coincidence events in the off-beam interval of each radiation field, thereby compensating for the decrease in count values ​​caused by the decay of radionuclides over time. This allows the low activity signal in the off-beam interval to be effectively used for reconstruction, expanding the data range available for radiation field separation, and thus significantly improving the accuracy of particle, such as proton therapy, range verification. Furthermore, for subsequent radiation fields, the residual signal of the previous radiation field can be corrected to the reference time of the current radiation field, and signal interference between radiation fields can be eliminated through differential processing. This technical solution solves the signal aliasing problem caused by the accumulation of positron-emitting nuclide residues in multi-field delivery modes, achieving accurate separation and reconstruction of the independent dose distribution of each radiation field, and further significantly improving the accuracy of particle, such as proton therapy, range verification.

[0071] Optional features and other effects of the embodiments of this application are described in part below, and in part will be apparent from reading this document. Attached Figure Description

[0072] The embodiments of this application will be described in detail with reference to the accompanying drawings. The elements shown are not limited to the scale shown in the drawings, and the same or similar reference numerals in the drawings denote the same or similar elements, wherein:

[0073] Figure 1 The image shows a PET imaging aliasing map under multiple energy layer delivery in a two-field configuration, according to an example.

[0074] Figure 2 A first exemplary flowchart of a PET imaging method according to an embodiment of this application is shown;

[0075] Figure 3 A second exemplary flowchart of a PET imaging method according to an embodiment of this application is shown;

[0076] Figure 4 A single-event count-time statistics histogram according to a specific embodiment of this application is shown;

[0077] Figure 5 A third exemplary flowchart of a PET imaging method according to an embodiment of this application is shown;

[0078] Figure 6 A fourth exemplary flowchart of a PET imaging method according to an embodiment of this application is shown;

[0079] Figure 7 A fifth exemplary flowchart of a PET imaging method according to an embodiment of this application is shown;

[0080] Figure 8 A sixth exemplary flowchart of a PET imaging method according to an embodiment of this application is shown;

[0081] Figure 9 An eighth exemplary flowchart of a PET imaging method according to an embodiment of this application is shown;

[0082] Figure 10 This illustrates a practical conformity binning statistics histogram according to a specific embodiment of this application;

[0083] Figure 11 A ninth exemplary flowchart of a PET imaging method according to an embodiment of this application is shown;

[0084] Figure 12 A tenth exemplary flowchart of a PET imaging method according to an embodiment of this application is shown;

[0085] Figure 13 A schematic diagram of inter-field correction according to a specific embodiment of this application is shown; and

[0086] Figure 14A schematic diagram of an electronic device for a PET imaging method according to an embodiment of this application is shown. Detailed Implementation

[0087] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0088] The term "comprising" and its variations as used herein signify open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "at least partially based on". The terms "one example embodiment" and "one embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc., may refer to different or the same objects.

[0089] In this application embodiment, the term "field" refers to a particle beam delivery unit within a treatment fraction, determined by incident geometry (e.g., gantry angle / bed position), scan pattern (point scan / line scan / area scan), and energy layer sequence and timing. A field typically contains multiple energy layers to cover lesions at different depths. In this application embodiment, the term "multi-field" refers to delivering two or more fields sequentially or alternately from different incident directions during a treatment fraction to improve the conformity of dose distribution and dose constraint on organs at risk.

[0090] In the embodiments of this application, the term "particle therapy" refers to the general term for radiotherapy of tumors using charged particle beams.

[0091] As mentioned earlier, in current proton therapy techniques, PET signals from different fields in multi-field delivery modes exhibit severe temporal and spatial aliasing, which traditional in-beam monitoring techniques cannot effectively address. Specifically, while traditional in-beam monitoring can capture the distribution of positron-emitting nuclides generated by the interaction between the proton beam and tissue in real time, providing important information for range verification, when multiple fields irradiate the tumor from different angles, positron-emitting nuclides generated by each field accumulate within the tissue. Traditional in-beam monitoring techniques still focus on accurate localization of the distal range and theoretical innovations in signal processing. Even though traditional image reconstruction methods attempt to filter out transient radiation events to improve image quality through accelerator radio frequency phase filtering, they still cannot effectively distinguish the independent contributions of each field. Furthermore, since each field also includes the delivery of multiple energy layers, the signal overlap between adjacent energy layers further exacerbates the difficulty of separation.

[0092] refer to Figure 1 The image illustrates aliasing in PET imaging under multiple energy layers delivered across two radiation fields. Left image A shows the clinically acquired PET means, while right image B shows the Monte Carlo simulation (MC PET) results reflecting only the physical process under the same delivery parameters. In right image B, two clearly defined bands corresponding to the two radiation fields are visible, converging in an approximately "T" shape in the central region. In contrast, in left image A, due to the simultaneous action of multiple radiation fields and energy layers, adjacent energy layers within a single radiation field merge into a single band, and the bands from different radiation fields further overlap in the intersection area, forming significant temporal-spatial aliasing. Therefore, traditional in-beam monitoring and conventional image reconstruction methods struggle to attribute aliasing to radiation fields or energy layers (distinguishing their respective contributions), making it difficult to accurately extract range / dose information for a single radiation field. This limits independent verification of the actual delivered dose for each radiation field and dynamic adjustment and optimization based on real-time monitoring results. Therefore, due to the lack of effective means to separate aliased signals in multi-field delivery modes, the dose information of a single field cannot be accurately extracted, which limits the ability to independently verify the actual dose delivered by each irradiation field. This limitation not only affects the accuracy of range verification but also restricts the possibility of dynamically adjusting treatment plans based on real-time monitoring results, ultimately impacting the precision and safety of proton therapy. Furthermore, current proton beam PET systems are typically based only on three-dimensional space, which particularly limits the implementation of signal separation techniques between fields.

[0093] In response, this application provides a PET imaging method, storage medium, and electronic device for multi-field particle therapy, aiming to achieve PET signal separation and reconstruction by field under multi-field and multi-energy layer delivery conditions, reduce the impact of range uncertainty on verification, improve the accuracy and availability of range / dose information, and thus improve the precision and safety of treatment plans.

[0094] In some embodiments of this application, the particles include protons, and the particle therapy includes proton therapy (PT). As an explanation and not a limitation, utilizing the characteristic that the energy deposition of a proton beam in tissue initially decreases with depth and then increases, reaching a peak at a specific depth (the Bragg peak), proton therapy can significantly reduce the dose to surrounding healthy tissue. However, it is conceivable that the methods of the embodiments of this application can also be reasonably combined with other particle therapies, such as heavy ion therapy, where heavy ions include, but are not limited to, carbon ions, helium ions, neon, and oxygen, etc., and this application does not impose any limitations on this.

[0095] The PET imaging method of this application embodiment can be implemented through different software, hardware, or a combination of both. In some embodiments, the PET imaging method of this application embodiment can be partially or wholly stored in a storage device (such as the built-in storage module of the detection device or an external storage device) in the form of a program or instructions. When the program or instructions are executed, they can implement the PET imaging method according to the embodiment of this application. Meanwhile, the hardware used to implement the PET imaging method of this application embodiment can be a device with a large amount of computing resources (e.g., a computer, server, cloud computing, etc.) or a device with limited computing resources (e.g., FPGA (Field Programmable Gate Array) chip board, ASIC (Application-Specific Integrated Circuit) chip board, etc.). This application does not impose any restrictions on this.

[0096] refer to Figure 2 This paper illustrates a first exemplary flowchart of a multi-field particle therapy PET imaging method according to some embodiments of the present application, which may include the following steps S210 to S240.

[0097] S210: Acquires single-event data generated by the PET detector during multi-field particle therapy.

[0098] In different embodiments of this application, step S210 is optional. In some embodiments, single-event data may have already been obtained through a previous acquisition process and stored in a data storage system. In this case, step S210 can be omitted, and the existing single-event data can be directly read from the data storage system for subsequent processing and analysis. In other embodiments, especially when the method of this application is applied to real-time monitoring application scenarios, the single-event data acquisition described in step S210 can be performed synchronously during particle beam delivery to achieve real-time data acquisition and processing.

[0099] In some embodiments, single-event data may include time, energy, and location information of the event. In this embodiment, time information may include, for example, the precise moment when the recorded gamma photon was captured by the detector, with timestamp accuracy down to the nanosecond or sub-nanosecond level; energy information may include, for example, the energy value of the gamma photon deposited in the scintillation crystal, typically in keV, which can be used for energy coincidence window filtering, excluding scattering events and random events; location information may include, for example, the three-dimensional spatial coordinates of the event in the detection system, such as a PET system, which may be the crystal number (e.g., IP address encoding), detector module index, or Cartesian coordinates after location decoding, etc. This location information is used to determine the spatial location of the Line of Response (LOR).

[0100] In some embodiments, the single-event count value per unit time (e.g., 1 second) can be continuously recorded during data acquisition, i.e., the number of single events acquired per unit time. Subsequently, all single-event data can be categorized and saved / cached into multiple files according to the detector channel. In this embodiment, the particle beams of each field are delivered sequentially, for example, according to a preset treatment plan. For example, for a three-field treatment plan, the delivery order is Field 1 → Field 2 → Field 3. Each delivery in a field corresponds to an independent scanning session: Scan 1: containing the acquisition of Field 1, Scan 2: containing the acquisition of Field 2, ... Scan S: containing the acquisition of Field S. Unless otherwise specified, in the embodiments described below, 's' will refer to the acquisition containing the 's'th field, where 1 ≤ s ≤ S.

[0101] With the development of digital PET, digital PET detectors, with their modular and fully digital characteristics, can acquire a large amount of digital threshold-voltage sampling information during particle therapy and conveniently, quickly, and accurately form single-event data for subsequent processing. In optional embodiments, buffering and transmission mechanisms can be used to efficiently manage the real-time data stream acquired by the PET detector, such as double-buffering / multi-buffering technology that alternates between acquisition and processing threads on a frame-by-frame basis. This application does not impose any limitations on this approach.

[0102] S220: Identify one or more time intervals for each shooting field based on the temporal distribution characteristics of the collected single-event data.

[0103] In some embodiments of this application, the collected single-event data can exhibit different distribution characteristics in the time dimension based on changes in the particle beam delivery state during particle therapy. For illustrative purposes, and not as a limitation, taking proton therapy as an example, the delivery of the proton beam during treatment exhibits significant temporal characteristics: when the proton beam is on and irradiates the tissue, a large number of positron-emitting nuclides are instantaneously generated, resulting in a high density of single events recorded by the PET detector; while during beamoff or energy layer switching, only the previously generated nuclides continue to decay, and the single-event count decreases significantly. Therefore, one or more time intervals for each radiation field can be identified based on this distribution characteristic. In this embodiment, each time interval corresponds, for example, to a delivery state of a specific particle beam, which includes, for example, delivery not started, delivery in progress, and delivery ended.

[0104] In some embodiments of this application, the act of identifying time intervals should be interpreted broadly, including but not limited to: identifying (main) time intervals, such as background intervals (e.g., before any particle beams have been delivered), beam-on intervals (e.g., during particle beam delivery), beam-off intervals (e.g., after particle beam delivery has ended), and directly or additionally identifying sub-intervals within the (main) time interval, such as high count periods and low count periods. For explanation, during the time period of multi-energy-layer particle beam delivery, i.e., the beam-on interval, the single-event count will increase significantly, manifesting as a distinct high count peak in the statistical histogram, where the delivery of each energy layer can correspond to an independent count peak, the height of which is related, for example, to the delivery dose and duration of that energy layer. Correspondingly, during the intervals between particle beams (e.g., due to the switching time between different energy layers), since no new positron-emitting nuclides are generated, the count value will decrease accordingly, forming a low count valley region in the histogram. Thus, alternating peaks of high count values ​​and valleys of low count values ​​can be formed within the beam on interval. In some embodiments, periods of high count values ​​and periods of low count values ​​can also be identified, as described below.

[0105] refer to Figure 3 S220 may include the following sub-steps S221, S222 and S223.

[0106] S221: Statistically analyze the distribution of single-event data for each shooting field over time, and generate a single-event statistical histogram for each shooting field.

[0107] In some embodiments, the single-event data acquired in each field can be preprocessed. In some embodiments, after the acquisition of each field is completed, the single-event data of each channel is already arranged in ascending chronological order within the detector channel. At this point, sorting can be performed to obtain a globally time-ordered set of single events. In other embodiments, the raw data that is not arranged in chronological order can also be sorted, for example, by sorting based on a timestamp field. It is understood that the above sorting operations may include merge sort or other sorting algorithms, which will not be elaborated upon here.

[0108] In some embodiments, the generation of a single-event statistical histogram may optionally include a time binning step. As an explanation, and not a limitation, a single-event statistical histogram can be generated by binning the time axis, that is, dividing the collection period into a series of continuous and non-overlapping sub-intervals according to a preset or adaptive time parameter (e.g., duration). The count value corresponding to each sub-interval (represented by the height of the bars in the single-event statistical histogram) can be obtained by counting the number of single events within each sub-interval. In this embodiment, the time parameter can be selected empirically, and this application does not impose any restrictions on it. In optional embodiments, the time parameter used in S221 can also be reasonably reused in the steps of subsequent embodiments of this application, and this application does not impose any restrictions on it.

[0109] S222: Determine the intersection of the set counting threshold and the single-event statistical histogram of each field.

[0110] In some embodiments, the counting threshold can be determined based on the numerical distribution of the entire statistical histogram. For example, the maximum count value in the statistical histogram can be identified, and the counting threshold can be set as the maximum value multiplied by a preset percentage. In this embodiment, the preset percentage is, for example, 30% to 50%, and the specific value can be adjusted according to experience or actual signal characteristics and noise levels. This application does not impose any limitations on this.

[0111] In some embodiments, all intersections between the counting threshold (horizontal line) and the statistical histogram can be identified and recorded. In this embodiment, the identification of intersections is achieved, for example, by traversing the data of the statistical histogram. Specifically, for example, starting from the beginning position (0) of the time axis, the relationship between the single event count value and the counting threshold at each adjacent time point can be checked. When it is detected that the single event count value at a certain time point is greater than or equal to the counting threshold, an intersection point can be obtained based on that time point and the counting threshold. In this embodiment, for example, the time point of each intersection between the counting threshold (horizontal line) and the histogram can be recorded:

[0112]

[0113] Where s refers to the s-th field of fire, and s = 1, 2, ..., S; In the s-th field, the threshold intersects with the histogram. n refers to the number of intervals of the statistical histogram corresponding to high count values ​​greater than the counting threshold. The interval corresponding to the same high count value has two intersections with the counting threshold at its endpoint.

[0114] In a specific embodiment of this application, reference is made to Figure 4 The example shows a single-event count-time statistics histogram for a certain field of fire, where the count threshold is set to 200,000. This count threshold is shown in the single-event statistics histogram as a black dashed line extending horizontally at the vertical axis height of 200,000, and the black dashed line has multiple intersections with the bars in the histogram, each bar representing the count value of a single event within a different time period.

[0115] In this embodiment, two of the multiple intersection points are exemplarily identified: as well as .

[0116] S223: Determine one or more time intervals for each shooting field based on the intersection points.

[0117] In some embodiments of this application, after determining the intersection of a set counting threshold with the single-event statistical histogram of each field, one or more time intervals can be determined (separated) based on these intersections.

[0118] In some embodiments, step S223 may include: determining the time period prior to the initial intersection point as the background interval for each field. In this embodiment, the initial intersection point refers to the first intersection point where the count threshold and the histogram bars intersect in chronological order. The time interval prior to this initial intersection point can be considered as a period before any particle beams have been delivered.

[0119] In some embodiments, among the multiple intersections of the count threshold (horizontal line) recorded above and the histogram, the time period corresponding to the background interval is, for example: .

[0120] In some embodiments, step S223 may include: determining the time period after the final intersection point as the beam-off interval for each field. In this embodiment, the final intersection point refers to the intersection point where the counting threshold and the histogram bar intersect for the last time in chronological order. The time interval after the final intersection point can be regarded as the end of the field particle beam delivery process, and no further particle beam delivery will be made.

[0121] In some embodiments, among the multiple intersections of the count threshold (dashed line) recorded above and the histogram, the time period corresponding to the beam-off interval is, for example:

[0122] ;(in (This is the end time of the scan for the s-th field).

[0123] In some embodiments, step S223 may further include: determining the time period between the initial intersection point and the final intersection point as the beam-on interval of each field.

[0124] In some embodiments, among the multiple intersections of the count threshold (dashed line) recorded above and the histogram, the time period corresponding to the beam on interval is, for example: .

[0125] In some embodiments, step S223 may further include: determining the time period between the intersection of odd-numbered indices and the subsequent adjacent intersection of even-numbered indices as the high count value period of the bundle interval.

[0126] In some optional embodiments, step S223 may further include: determining the time period between the even-numbered intersection point and the subsequent adjacent odd-numbered intersection point as the low count value period of the bundle interval.

[0127] In this embodiment, each high count period typically corresponds to a specific energy layer, the duration of which depends on the prescribed dose and beam intensity of that energy layer. As explained earlier, during the time period of multi-energy layer particle beam delivery, i.e., within the beam-on interval, alternating high count peaks and low count valleys can form. High count periods, for example, correspond to the actual delivery (irradiation) time window of the particle beam, while low count periods, for example, correspond to short pauses during energy layer switching. Therefore, in the single-event statistical histogram, due to the physical characteristics of proton beam delivery, high count intervals (during beam delivery) and low count intervals (during beam intervals) always alternate, and there will be two consecutive intersections between the count threshold and the bar representing the delivery of a single energy layer (i.e., the aforementioned "interval corresponding to the same high count value has two intersections with the count threshold at its endpoint"). In some embodiments, the initial intersection point can be set to 1, so that any odd-numbered intersection point in the beam on interval corresponds to the delivery of a particle beam of a certain energy layer between the subsequent adjacent even-numbered intersection points.

[0128] In some embodiments, following the previous example, the time period corresponding to the high count value period is, for example:

[0129] ;

[0130] In an optional embodiment, step S223 may further include: determining the time period between the intersection of even-numbered indices and the subsequent adjacent intersection of odd-numbered indices as the low count value time period of the bundle interval.

[0131] In some embodiments, following the previous example, the time period corresponding to the low count value period is, for example:

[0132] .

[0133] In this embodiment, the aforementioned high count period can be derived in reverse from the determined low count period, which falls within the protection scope of this application.

[0134] S230: Perform conformation processing on single event data to generate conformation event data.

[0135] In some embodiments of this application, the collected single-event data can be subjected to conformation processing, wherein the single-event data is, for example, the single-event data described in step S220 above.

[0136] In some embodiments of this application, the matching process may include, but is not limited to, using time matching, energy matching, or a combination of time matching and energy matching alone. This application does not impose any limitations on this. In one example, energy matching may be performed first, i.e., determining whether the energy in a single event data is within a given energy range (e.g., 350~650 keV). If not, the single event is discarded; otherwise, it is retained. Subsequently, time matching may be performed, i.e., for each single event that has undergone energy matching, a predetermined time window is opened, and it is checked whether the events within the time window meet the matching strategy. Events that meet the strategy and the main event of the current time window form matching event pairs and are recorded. More specific details regarding the matching process can be found in the prior art, and will not be repeated here.

[0137] In an optional embodiment, the matching process can employ a "take-all-goods" strategy, where all events that meet the basic conditions found within the time window are paired with the main event and recorded to obtain a set of matching event pairs arranged in ascending order of the main event's time. However, it is conceivable that other strategies may be employed in other embodiments, and this application does not impose any limitations on them.

[0138] In some embodiments, the conformity processing of each field in multiple firing fields can be performed independently or uniformly and then classified according to the time stamp. This application does not impose any restrictions on this.

[0139] In one example, the processing completes to obtain a list of matching events for the s-th field. For example, including There are 1 matching events, where each matching event... Record as

[0140]

[0141] Where P represents the location coordinates, E represents the energy value, t represents the timestamp, and subscripts 1 and 2 correspond to the two single events in the pair.

[0142] S240: Perform image reconstruction on the matching event data in the identified time interval of each shooting field to obtain the reconstructed image corresponding to each shooting field.

[0143] In some embodiments of this application, image reconstruction can be performed on each field based on the obtained coincidence event data of each field, thereby obtaining the reconstructed image corresponding to each field.

[0144] In some embodiments, the coincidence event data for each field includes coincidence event data within the identified time interval of each field. In some embodiments, the identified time interval particularly includes one or more of the aforementioned time intervals that have been separated.

[0145] In some embodiments, the identified time interval may include (identified and separated) beam-on intervals. It is understood that this application is not limited to using coincident event data within beam-on intervals. As a supplementary alternative, coincident event data within other time intervals may also be used for image reconstruction in alternative embodiments, which will be further described below.

[0146] In this embodiment, step S240 may include: step S241 (unidentified): performing image reconstruction on the coincidence event data in the in-beam interval of each field to obtain the in-beam reconstruction image corresponding to each field.

[0147] In some embodiments, image reconstruction can be performed based on all coincidence event data within the in-beam interval of each field. In some embodiments, the image reconstruction is based, for example, on the OSEM (Ordered Subsets Expectation Maximization) algorithm. In optional embodiments, a Time-of-Flight (ToF) algorithm can be further incorporated to utilize the time difference information of the arrival of the two photons generated by positron annihilation at different detectors to estimate the positional distribution of the annihilation event on the response line, thereby obtaining a higher quality reconstructed image.

[0148] In some embodiments, instead of using all coincidence event data in the in-beam interval of the firing field for image reconstruction, in step S241 above, a portion of the coincidence event data in the in-beam interval of each firing field may be selectively selected for image reconstruction.

[0149] In some embodiments, the beam-on interval in step 241 may include multiple high-count-value time periods. In this embodiment, image reconstruction may be selectively performed using coincidence event data from the high-count-value time periods identified based on the beam-on interval.

[0150] In some embodiments, reference Figure 5 Step S241 may include the following steps S2411 and S2412.

[0151] S2411: Extract coincidence event data from multiple high count periods within the in-beam interval of each field and merge them to obtain the first coincidence event set for each field.

[0152] In some embodiments, all coincidence events in each field can be traversed to identify coincidence events that belong to the high count period.

[0153] In one example, for the s-th field, for instance, all matching events belonging to the following high-count time period can be extracted:

[0154] .

[0155] Subsequently, after extracting the subsets of coincident events for all high-count periods, multiple subsets of coincident events can be obtained: (where n is the total number of high count periods), where each subset corresponds to a high count period. Subsequently, the data of the above subsets can be merged, keeping the complete information of each matching event unchanged, and the merged dataset can be re-sorted according to the time order, thereby determining the first matching event set C-merged for the s-th field.

[0156] S2412: Use the first coincidence event set of each field to perform image reconstruction to obtain the in-beam reconstructed image of each field.

[0157] In some embodiments, the OSEM-ToF algorithm can be used to reconstruct the image from the first coincidence event set C-merged obtained in step S2411. Specifically, this algorithm improves upon the traditional OSEM framework by introducing Time-of-Flight (ToF) information. In one example, the image reconstruction process includes, for instance: firstly, dividing the first coincidence event set C-merged into several ordered subsets according to the subset partitioning strategy of the OSEM algorithm; for each coincidence event in the subset, the traditional OSEM algorithm only uses its geometric response line (LOR) information for forward and backward projection operations, while the OSEM-ToF algorithm further utilizes the time-of-flight measurement value of each coincidence event, constraining the position probability distribution of annihilation events to a specific region of the response line through the ToF kernel function; in the forward projection stage, the expected projection value is calculated by combining ToF weights; in the backward projection stage, the update factor is weighted and allocated to the corresponding image voxels on the response line according to the ToF information, thereby achieving more accurate image updates; each subset is processed sequentially, iterating until the image converges or reaches a preset stopping criterion, and finally the in-beam reconstructed image of the field is obtained.

[0158] As a supplement or alternative to the aforementioned in-beam interval reconstruction, embodiments of this application can further utilize coincidence event data from the beam-off interval for image reconstruction, thereby enhancing the field separation effect. Explained but not limited, during proton therapy, after the proton beam delivery of each field stops, numerous single events may still be captured and coincidence events formed through coincidence processing; these events still contain important dose distribution information. However, due to the decrease in coincidence event counts caused by radioactive decay, directly using coincidence event data from the beam-off interval for reconstruction may yield poor results. Furthermore, for multi-field delivery modes, the nuclides produced by different fields will superimpose over time, and the residual activity of the previous field will affect the signal of subsequent fields, potentially further impacting the reconstruction effect. To address this, embodiments of this application combine appropriate correction processing, especially decay correction processing, with beam-off interval reconstruction processing to obtain high-quality beam-off reconstruction images, thereby more effectively separating the independent contributions of each field.

[0159] In this embodiment of the application, step S240 may include correcting the coincidence event data in the beam-off interval of each field, including optional random correction and background correction, obtaining decay correction factors through multi-exponential fitting, and using the corrected data to perform image reconstruction to obtain the beam-off reconstructed image corresponding to each field.

[0160] As mentioned above, in some embodiments, the identification of the beam-off interval is based on the time interval division determined in step S223, for example, the final intersection point can be used. Then until the field scan is completed. The time period is defined as the beam-off interval of the s-th field.

[0161] In some embodiments, reference Figure 6 Step S240 may include steps S242 to S244:

[0162] S242: Perform random correction and / or background correction on the coincidence event data of each beamout interval.

[0163] In the embodiments of this application, step S242, namely, the random correction and / or background correction processing, can be performed or at least partially performed before the decay correction processing described below, and the time-distributed count data after random correction and / or background correction can be used for subsequent decay correction processing, as further described below. However, it is also conceivable that step S242 is an optional step.

[0164] For illustrative purposes, and not as a limitation, random coincidence events refer to the situation where two gamma photons from different annihilation events are detected by chance within a time window. Such events do not carry effective spatial positioning information and reduce image contrast. Background events include non-target signals such as ambient background radiation and detector dark noise. By performing random correction and / or background correction to correct these interfering factors, the accuracy of subsequent image reconstruction can be significantly improved.

[0165] However, it is conceivable that in some embodiments, step S242 is optional, for example, in different cases, one or both of random correction and background correction may not be performed as needed.

[0166] refer to Figure 7 Step S242 may include steps S2421 to S2423:

[0167] S2421: At least the coincidence event data of each beamout interval of each field are statistically processed to form a first event count set based on time distribution.

[0168] In a preferred embodiment, the statistical processing is not only applied to the off-beam interval, but also to multiple time intervals of each field, preferably the entire time period, including the background interval, the in-beam interval, and the off-beam interval.

[0169] In a specific embodiment, the coincidence event data of each firing field can be binned at equal time intervals according to a preset duration to obtain a first bin set including multiple first bins. Each first bin includes a count of coincidence events falling within its respective first bin's time range.

[0170] In a specific example, for any field of fire s, the total number of... A list pattern set that matches the events This set Perform equal-interval first-time bin sorting. In a preferred example, as described above, the first bin sorting in step S2421 can use the same time parameters (such as duration) as the bin sorting described in the embodiment of step S220.

[0171] In some embodiments, in step S2421 above, for the s-th firing field, by selecting an appropriate bin size It can count the bins that match the events. :

[0172]

[0173] in, Let be the start time of the j-th sub-box in the s-th shooting field. End time:

[0174]

[0175] in, The number of single events in the j-th sub-box of the s-th field.

[0176] In an optional embodiment, when performing statistical binning, statistics can be based on the average time of the matching events; and preferably, a statistical histogram can be drawn according to the average time of each matching event.

[0177] Specifically, the average time of the i-th coincidence event in the s-th field. The definition is as follows:

[0178]

[0179] Therefore, we can base our decision on the average time of the i-th matching event. And container splitting time information, such as the start time of container splitting. To determine which bin the i-th matching event falls into, the bin count is incremented accordingly.

[0180] However, it is also conceivable that in other embodiments, other time information that matches the event can be used for binning statistics, which falls within the scope of this application.

[0181] S2422: At least the coincidence event data of each beamout interval of the firing field are delayed and statistically processed to form a second event count set based on time distribution.

[0182] In this embodiment, the time range of the statistical processing in step S2422 is consistent with that in step S2421. Accordingly, in a preferred embodiment, the statistical processing is not only applied to the off-beam interval, but also to multiple time intervals of each field, preferably the entire time period, including the background interval, the in-beam interval, and the off-beam interval.

[0183] In a specific embodiment, such as Figure 8 As shown, step S2422 may include:

[0184] S24221: Delay the coincidence event data of each beamout interval to obtain delayed coincidence event data.

[0185] In some embodiments, delay processing generates delayed list pattern matching data by delaying the timestamps of individual events by a set delay time. In some specific embodiments, the timestamp of each event in the original single-event data stream can be incremented by a fixed delay time τ_delay, for example, 100-500 nanoseconds, much larger than the matching time window (typically 5-10 nanoseconds). By way of explanation and not limitation, after such delay processing, the actual matching event pairs are broken up, while the statistical properties of random matching remain unchanged.

[0186] In some embodiments, the above-described delay processing can be performed using a PET device that supports delay compliance. In a preferred embodiment, when performing delay processing, delayed list-mode compliance data (Delayed List-mode Coincidence, or simply Delay Compliance) can be generated. Its definition can be consistent with data Similarly, and also arranged in chronological order of the first single event:

[0187]

[0188] S24222: Delayed matching event data is binned at equal intervals according to a preset duration to obtain a second bin set including multiple second bins.

[0189] Each second sub-box includes a count of delayed compliance events falling within its respective second sub-box time range.

[0190] In a further embodiment, the delayed coincident event data can be binned at equal intervals according to the preset duration, resulting in a second bin set including multiple second bins. In this embodiment, the number of first bins is equal to the number of second bins, and the binning times correspond one-to-one. This correspondence ensures that subsequent differential operations can be performed bin by bin.

[0191] In some embodiments, in step S24222 above, for the s-th firing field, using and For the same bin size, Also, bins are statistically analyzed based on the average time for each matching event:

[0192]

[0193] S2423: Perform random correction and / or background correction based on the difference between the first event count set and the second event count set.

[0194] In some embodiments, random correction based on the difference between the first event count set and the second event count set may include: step S24230 (not shown): subtracting the delayed coincidence event count of the corresponding second box in the second box set from the coincidence event count of each first box in the first box set to obtain the first difference count of each third box that has been randomly corrected.

[0195] In these embodiments, random correction is achieved through a simple differencing operation: the count of coincidence events in each first bin in the first bin set is subtracted from the count of delayed coincidence events in the corresponding second bin in the second bin set to obtain the first difference count of random correction for each third bin. By way of explanation and not limitation, this differencing operation effectively eliminates the contribution of random coincidences because the statistical distribution of random coincidence events is the same in both the original and delayed data, while true coincidence events are corrupted in the delayed data.

[0196] As a supplementary or parallel embodiment to the above embodiments, refer to Figure 9 Background correction based on the difference between the first event count set and the second event count set may include: Step S24231: Subtracting the delayed coincidence event count of the corresponding second bin from the coincidence event count of each first bin in the first bin set to obtain the first difference count of each third bin; Step S24232: Determining the minimum value among the first difference counts of each third bin as the background level; Step S24233: Subtracting the background level from the coincidence event count of each first bin in the first bin set to obtain the second difference count of each fourth bin that has been background corrected.

[0197] In these embodiments, background correction employs a minimum value estimation method. First, the coincidence event count of each first bin in the first bin set is subtracted from the delayed coincidence event count of the corresponding second bin in the second bin set to obtain the first difference count of each third bin that is randomly corrected. Then, the minimum value is determined from the first difference count of the third bin as the background level. Finally, the coincidence event count of each first bin in the first bin set is subtracted from the background level to obtain the second difference count of each fourth bin that is background corrected.

[0198] In other embodiments, random correction and background correction can be performed simultaneously. In these embodiments, if random correction and background correction are performed simultaneously, steps S24230 and S24231 above can be the same step, that is, first remove random events by using delayed coincidence technique, and then subtract the background by using minimum value method to obtain net true coincidence count.

[0199] The following describes specific embodiments of performing random correction and background correction simultaneously. However, as mentioned above, random correction and background correction can be performed selectively or simultaneously, which falls within the scope of this invention.

[0200] In some embodiments, in the above steps S24230 (S24231), for the s-th firing field, the... First box count minus The second bin count yields the actual binning result (third bin):

[0201]

[0202] Optionally, a statistical histogram can be plotted for each actual bin, as shown in the reference. Figure 10 The diagram illustrates a practical conformity binning statistical histogram according to a specific embodiment.

[0203] In some embodiments, in step S24232 above, for the s-th field, the minimum value of the actual binning result is selected as the background level:

[0204]

[0205] Furthermore, in step S24233 above, for the s-th field of fire, the background level can be subtracted from the binning results:

[0206]

[0207] Thus, the second difference count of the fourth bin, which is background corrected, can be obtained (in this specific embodiment, it is also randomly corrected).

[0208] In some embodiments, step S2423 may include: step A1 (not shown): removing the bins before the start time of each beam separation interval and using the remaining bin set for subsequent decay correction processing.

[0209] In these embodiments, for example, when performing event binning for multiple time intervals of each field in steps S2421 and S2422, preferably for the entire time period, the bins before the start time of each field's beam separation interval in the bins that have been randomly corrected and / or background corrected can be removed, and the remaining bin set can be used for subsequent decay correction processing.

[0210] In some embodiments, in step A1 above, for the s-th field, the start time of the beam-off interval is selected. All subsequent binning data yields beam-off binning with random and background corrections:

[0211]

[0212] In some other embodiments, step S2423 may further include:

[0213] Step A2 (not shown): Convert the start time of each bin in the remaining bin set into a relative start time relative to the start time of each beam separation interval, forming a relative time vector;

[0214] Step A3 (not shown): Count the remaining bins in the bin set after random correction and / or background correction to form an observation vector.

[0215] The relative time vector and the observed value vector are used for fitting the subsequent decay correction model, and more specifically, for multi-exponential fitting.

[0216] In some embodiments, in steps A2 and A3, for the s-th field of fire, beamoff binning data with random and background correction is used. As input, let the number of beam-off intervals be... If there are several sub-boxes (e.g., from the k-th sub-box to the n-th sub-box), then:

[0217] Binning time vector:

[0218]

[0219] In this process, the time point of each bin is converted into a time difference relative to the start time of beam-off, forming a time vector. This allows for a unified time starting point, facilitating the fitting of the exponential decay model.

[0220] Observation vector:

[0221]

[0222] Therefore, for the count value of each bin, the corresponding delayed coincidence count (random correction) and the background count (background correction) can be subtracted to obtain the net true coincidence count, thereby obtaining the fitted target observation value, which reflects the true intensity of the radioactive decay signal.

[0223] S243: Perform decay correction processing on the time-distributed count data of coincidence events in the off-beam interval of each field to obtain the decay correction factor.

[0224] In some embodiments, decay correction processing can be used as a key technical step to reduce quantization errors in decay signals. Thus, the contribution of each nuclide can be separated using a multi-exponential fitting method, and a precise decay correction factor can be calculated based on the fitting results to compensate for the signal intensity decrease caused by radioactive decay.

[0225] In the embodiments of this application, the input data for decay correction processing is time-distributed count data, which is preferably time-binned data that has undergone random correction and / or background correction, as described above.

[0226] In this embodiment, the decay correction process may include several steps: first, determining the types of nuclides involved in decay; second, establishing a mathematical model describing the decay process of multiple nuclides; then, estimating the model parameters using a fitting algorithm; and finally, calculating the decay correction factor based on the fitting parameters. In this embodiment, the output of the decay correction process is the decay correction factor DCF(t) corresponding to each coincidence event, which is used to adjust the weight of each coincidence event in subsequent image reconstruction.

[0227] refer to Figure 11 Step S243 may include steps S2431 to S2433:

[0228] S2431: Based on prior information, obtain multiple nuclide categories corresponding to the coincidence event data in the beam-off interval of each field.

[0229] In the embodiments of this application, the prior information is obtained from previous experiments or practical applications. Based on the application scenario or the particles acting on the object, the types of nuclides that induce the generation of the target can be summarized. For example, the prior information can be a table including the application scenario, particle type and / or nuclide type, or the prior information can be a database obtained by simulating the application scenario, particle type and / or nuclide type in the simulation system.

[0230] For example, in a proton therapy scenario, the main components of the positron-emitting nuclides generated by proton-induced target formation are... 15 O、 11 C and 10 C. For example, in heavy ion therapy, a heavy ion beam (such as carbon ions) is used... 12 C) The main components of the positron-emitting nuclides induced by the target are: 12 C 16 O and 14 N. For example, in the scenario of oil exploration, the main components of the positron-excited nuclides generated are... 16 O、12 C 28 Si、 27 Al.

[0231] S2432: Based on the time-distributed count data of coincidence events in each beam-off interval of the beam field and multiple nuclide categories, multi-exponential fitting is performed to obtain the sum of the initial activities of all nuclides and the sum of the decay activities of all nuclides.

[0232] In some embodiments, for any radiation field s, it is assumed that the change in total radioactivity in the image over time can be expressed as the sum of the decay activities of M nuclides, i.e.:

[0233]

[0234]

[0235] Where M represents the number of nuclide species. m = 1...M, , representing the time difference relative to the start time of Beam off. The decay constant is given by the known half-life of each nuclide. Given, Let be the initial activity coefficient of the m-th nuclide in the s-th field at the reference time. In one example, for field s, the reference time can be the start time of the beam-off interval of the corresponding field. However, it can also include other times. For example, when performing cross-field correction, the reference time can be the start time of the beam interval of the field to which the correction is being performed. For instance, assuming that the conformity data of the beam interval of field s-1 is to be corrected to field s, the reference time is the start time of the beam interval of field s. . Let be the decay constant of the m-th nuclide. In some embodiments, the same combination of nuclides is used for fitting each radiation field.

[0236] In some embodiments, as described above, for the field of fire s, the data can be binned using beam-off data. As input, for example, using Beam-off binned data that has been randomized and background-corrected. As input. In a further preferred embodiment, as previously described, the relative time vector obtained through time transformation in steps A2 and A3 can be used. and observation vector As input.

[0237] In a specific embodiment, refer to Figure 12 Step S2432 may include steps S24321 to S24324:

[0238] S24321: Establish a multi-exponential decay model corresponding to multiple nuclide categories.

[0239] The parameters of the multi-exponential decay model include the initial activity coefficient and decay coefficient of each nuclide class.

[0240] S24322: Using time-distributed count data consistent with events, perform multi-exponential fitting based on a multi-exponential decay model to determine at least the initial activity coefficients of each nuclide class at the reference time.

[0241] S24323: Calculate the sum of the initial activities of all nuclides based on the initial activity coefficients of each nuclide class at the reference time.

[0242] In some embodiments, the multi-exponential fitting in step S2432 above can employ fixed half-life fitting. Accordingly, in these embodiments, the decay coefficient of each nuclide class described in step S24321 above can be determined based on the half-life of that nuclide class. Step S24322 above may include: Step B1: Using the time-distributed count data (preferably time-binned data) of the coincident event as input, multi-exponential fitting is performed using linear least squares to estimate the initial activity coefficient of each nuclide class at the reference time.

[0243] In a specific embodiment, when using a fixed half-life fitting, the model function for the field s can be expressed as:

[0244]

[0245] Here, the unknown quantity is only This transforms the original nonlinear fitting problem into a linear fitting problem, resulting in higher computational efficiency.

[0246] Accordingly, parameter vector It can be represented as: Taking the least squares method as an example, the objective function is:

[0247]

[0248] Where χ² represents the sum of squared residuals between the observed values ​​and the model's predicted values. Minimizing this function allows us to find the parameter values ​​that best fit the observed data to the model. The j-th component of the measurement value vector described in step A2 above represents the number of single events in the j-th interval (after random correction and / or background correction). The j-th component of the relative time vector described in step A3 above represents the time of the j-th interval after time point conversion (i.e., the time difference compared to the start of the beam departure).

[0249] The linearized problem can be written in matrix form:

[0250]

[0251] This is the matrix representation of linear regression, where The design matrix contains the exponential decay values ​​at various time points. Let be the parameter vector to be determined. Given a vector of observations, this form facilitates solving the problem using matrix operations.

[0252] Among them, the design matrix for:

[0253]

[0254] Design Matrix Each row corresponds to a time point, and each column corresponds to the decay function value of a nuclide. Matrix element X jm Let represent the decay factor of the m-th nuclide at time j. The matrix constructed in this way allows for linear combination. It can represent the superposition decay signal of multiple nuclides.

[0255] The least squares solution is:

[0256]

[0257] As an explanation and not a restriction, this is an analytical solution to the linear least squares problem, from which the optimal parameter estimates are obtained directly through matrix operations. This is called the Moore-Penrose pseudo-inverse, when When it is reversible, this solution is the only optimal solution.

[0258] Here, when determining the initial activity coefficient, the sum of the initial activities of all nuclides can be calculated, which can be used to subsequently determine the decay factor.

[0259] In a specific example, for the radiation field s, taking the triple-exponential fitting function as an example, in a proton therapy scenario, the main component of the positron-emitting nuclide generated by the proton-induced target is... 15 O、 11 C and 10 C. Construct the following three-exponential model, assuming that the total radioactivity is the superposition of the independent decays of the three nuclides, then: .

[0260] in, 15 The half-life of O is 2.04 minutes. 11 The half-life of C is 20.334 minutes. 10 The half-life of C is 19.29 seconds, and its decay constants are as follows:

[0261]

[0262]

[0263]

[0264] Therefore, in this scenario, The variables are only , and .

[0265] Therefore, by using the time-binned data of coincidence events in the off-beam interval as input, and employing the linear least squares method for multi-exponential fitting, the initial activity coefficients of each nuclide class at the reference time can be estimated through the best square approximation. , and It can also calculate the sum of the initial activities of all nuclides. .

[0266] In the specific embodiments described above in this application, the coefficients of each exponential term in the three-exponential model are expressed as a non-explicit multiplication by the decay constant, i.e., the coefficients of the exponential terms. , and The decay constant of the corresponding nuclide is already included. Therefore, after completing the multi-exponential fitting, if it is necessary to calculate the contribution ratio of each nuclide, the initial activity coefficient can be obtained by dividing each initial activity coefficient by its corresponding decay constant. For example, it can be obtained by... Divide by The contribution percentage of a particular nuclide.

[0267] In other embodiments of this application, the coefficients of each exponential term in the three-exponential model can also be expressed by explicitly multiplying by the decay constant. In this case, the decay constant is displayed as an explicit factor in the model, and the decay constant and another fitting coefficient together serve as the coefficients of the exponential terms. Thus, after completing the multi-exponential fitting, the aforementioned other fitting coefficient can be directly used to calculate the contribution ratio of the nuclide, which falls within the protection scope of this application.

[0268] In some embodiments, the multi-exponential fitting in step S2432 above can employ variable half-life fitting. Accordingly, in these embodiments, step S24322 above may include: Step B2: Using the time-distributed count data (preferably time-binned data) of the coincident events as input, a nonlinear least squares method is used to perform multi-exponential fitting to estimate the initial activity coefficient of each nuclide class at the reference time and the decay coefficient of each nuclide class. In a further preferred embodiment, step B2 above may include: Step B21: When performing multi-exponential fitting, the decay coefficient of each nuclide class is restricted to a preset constraint range.

[0269] In the embodiments of this application, the nonlinear least squares method described in step B2 is preferably the Levenberg-Marquardt method, but other nonlinear least squares methods are also conceived, including but not limited to gradient descent, Gauss-Newton method, etc.

[0270] In one specific embodiment, when using variable half-life fitting, the model function for the field s can be:

[0271]

[0272] In variable half-life fitting, the decay constant is also used as a parameter to be fitted. This is to explain, rather than limit, that small variations in the half-life under actual measurement conditions can be taken into account.

[0273] The parameter vector is:

[0274]

[0275] The constraints are:

[0276]

[0277] In a preferred embodiment, constraints are set to limit the decay constant to the maximum and minimum range of the theoretical value, thereby preventing the half-life from deviating from the physically reasonable range during the fitting process. In one example, the constraint is 95% to 105% of the initial value. This allows for some flexibility to adapt to actual conditions while avoiding non-physical fitting results.

[0278] Taking the nonlinear least squares method as an example, the Levenberg-Marquardt algorithm is used to solve the problem:

[0279]

[0280] This is a nonlinear optimization problem that is solved using an iterative algorithm. The Levenberg-Marquardt algorithm combines the advantages of gradient descent and Gauss-Newton's method, exhibiting fast convergence speed when close to the optimal solution and good stability when far from the optimal solution.

[0281] The Jacobian matrix is:

[0282]

[0283] The Jacobian matrix contains the partial derivatives of the model function with respect to its parameters and is crucial for calculating the search direction and step size in nonlinear optimization algorithms. Matrix element J jm This represents the sensitivity of the model value at the j-th observation point to the m-th parameter.

[0284] For the activity coefficient:

[0285]

[0286] Model function for activity coefficient The partial derivative is the corresponding exponential decay factor, which reflects the linear effect of changes in the activity coefficient on the model output.

[0287] For the decay constant:

[0288]

[0289] The model function with respect to the decay constant λ m The partial derivatives of the function include the time factor Δt, indicating that the influence of the decay constant increases with time, which is consistent with the mathematical properties of exponential functions. The negative sign indicates that the larger the decay constant, the faster the activity decays.

[0290] Here, once the initial activity coefficients and decay coefficients of each nuclide class at the reference time are obtained, the sum of the initial activities of all nuclides can be calculated, which can be used to subsequently determine the decay factor.

[0291] S24324: Using a multi-exponential decay model that includes the determined initial activity coefficients, calculate the sum of the decay activities of all nuclides at any given time for each coincidence event.

[0292] Here, using a multi-exponential decay model with predetermined parameters (e.g., initial activity coefficient or initial activity coefficient and decay coefficient), the total (theoretical) activity at any time t for each field can be obtained:

[0293]

[0294] Thus, we can also obtain the total activity of each coincident event at any time t, that is, the sum of the decay activities of all nuclides.

[0295] Therefore, the decay correction factor can be determined in step S2433 below.

[0296] S2433: Determine the decay correction factor corresponding to each coincidence event in the off-beam interval of each field based on the sum of the initial activities of all nuclides and the sum of the decay activities of all nuclides.

[0297] In this embodiment, the decay correction factor corresponding to each coincidence event in the off-beam interval of each field can be determined based on the ratio of the sum of the initial activities of all nuclides to the sum of the decay activities of all nuclides.

[0298] As an explanation and not a limitation, the decay correction factor is used to compensate for the decrease in count value due to radioactive decay, and ideally (without decay), the activity at all times should be equal to the total activity at the reference time.

[0299] Therefore, in some specific embodiments, the decay correction factor (DCF) is defined as the ratio of the total activity at the reference time to the total activity at any time:

[0300]

[0301] in The total initial activity is the beam off state of the s-th field at the start of the beam off.

[0302] In this embodiment, after applying decay correction, the correction weight for the i-th coincidence event is:

[0303]

[0304] In the embodiments of this application, during image reconstruction, the decay correction factor can be used to perform decay correction on the weights of each coincident event.

[0305] As mentioned earlier, depending on the needs, for the field s, the reference time can be the start time of the beam separation interval of the corresponding field. However, it can also be or include other times, such as the start time of the beam interval of other fields. Especially in some embodiments, for example, when performing cross-field correction, the reference time can be the start time of the beam interval of the field to which the correction is to be performed. For example, assuming that the conformity data of the beam interval of field s-1 is to be corrected to field s, the reference time is the start time of the beam interval of field s. As mentioned below. In this case, the aforementioned DecayCorrection Factor (DCF) can still be determined accordingly, only the reference time has changed.

[0306] S244: Use coincidence event data and corresponding decay correction factors in the off-beam interval of each field to perform image reconstruction and obtain the off-beam reconstructed image corresponding to each field.

[0307] In the embodiments of this application, attenuation correction can be performed on coincidence event data in the off-beam interval of each field at or before image reconstruction based on the attenuation correction factor obtained above.

[0308] In some embodiments, decay correction can be in-field correction.

[0309] Accordingly, in these embodiments, step S244 may include: step C1 (not shown): obtaining the decay correction factor at the start time of the off-beam interval of each field, correcting the activity of each nuclide corresponding to the coincidence event in the off-beam interval of each field using the corresponding decay correction factor, and reconstructing the first off-beam reconstruction image based on the coincidence event of the off-beam interval of each field after correction.

[0310] In these embodiments, for the field s, the decay correction factor determined above can be used. Or, more specifically, using decay-corrected correction weights, the activities of each nuclide corresponding to the i-th coincidence event within the off-beam interval of the field s are corrected to the activities of each nuclide corresponding to the start time of the off-beam interval of the field s, which is the activities of each nuclide corresponding to the reference time (in this embodiment, the reference time is the start time of the off-beam interval of the corresponding field). ).

[0311] Furthermore, the first off-beam reconstructed image is obtained by using the corrected coincidence data for image reconstruction, which is an off-beam reconstructed image corrected for in-field decay. In this embodiment, any suitable reconstruction algorithm can be used for image reconstruction, such as the OSEM TOF algorithm mentioned above, which will not be elaborated here.

[0312] In some embodiments, decay correction may also include inter-field correction. By way of explanation and not limitation, in a multi-field delivery mode, the positron-emitting nuclide activity generated in each field will remain in the next field.

[0313] Accordingly, in these embodiments, step S244 may include:

[0314] Step D1 (not shown): For the first field, obtain the decay correction factor at the start time of the off-beam interval of the first field, use the decay correction factor to correct the activity of each nuclide corresponding to the coincidence event in the off-beam interval of the first field, and reconstruct the second off-beam reconstruction image of the first field based on the corrected coincidence event.

[0315] Step D2 (not shown): For any subsequent field s, obtain the decay correction factor at the start time of the beamout interval of the s-th field, use the decay correction factor to correct the activity of each nuclide corresponding to the coincidence event in the beamout interval of the s-th field, and reconstruct the first intermediate reconstructed image of the s-th field based on the coincidence event in the beamout interval of the s-th field after correction. Use the decay correction factor to correct the activity of each nuclide corresponding to the coincidence event in the beamout interval of the (s-1)-th field, and reconstruct the second intermediate reconstructed image of the (s-1)-th field based on the coincidence event in the beamout interval of the (s-1)-th field. Obtain the second beamout reconstructed image of the s-th field based on the difference between the first intermediate reconstructed image and the second intermediate image.

[0316] In these embodiments, for the first field of fire, the decay correction factor determined above can be used. Or, more specifically, using decay-corrected correction weights, the activities of each nuclide corresponding to the i-th coincidence event within the off-beam interval of the first field are corrected to the activities of each nuclide corresponding to the start time of the off-beam interval of the first field, which is the activity of each nuclide corresponding to the reference time (in this embodiment, the reference time is the start time of the off-beam interval of the corresponding field). Furthermore, the corrected coincidence data is used to perform image reconstruction to obtain a second off-beam reconstructed image. In this embodiment, any suitable reconstruction algorithm can be used for image reconstruction, such as the OSEMTOF algorithm mentioned above, which will not be elaborated here. In a specific instance, the OSEM-ToF method is used to reconstruct the beamoff region of the first field. The event subset is reconstructed to obtain the Beam-off reconstructed image of the first field after decay correction. .

[0317] In these embodiments, for the field s, the decay correction factor determined above can be used. Or, more specifically, using the decay-corrected correction weights, the i-th coincidence event within the off-beam interval of the field s is corrected to the start time of the off-beam interval of that field s, which is the reference time (in this embodiment, the reference time is the start time of the off-beam interval of the corresponding field). ).

[0318] Further, the first intermediate reconstructed image is obtained by using the beamoff coincidence data of the beamoff field s after in-field correction, which is the beamoff reconstructed image after in-field decay correction. In this embodiment, any suitable reconstruction algorithm can be used for image reconstruction, such as the OSEM-ToF algorithm mentioned above, which will not be elaborated here. In a specific instance, the beamoff interval of the s-th beamoff field is analyzed using the OSEM-ToF method. Reconstructing the image by reconstructing the beamoff image under the s-th field after decay correction of the s-th field. .

[0319] Furthermore, for field s, the previous field, i.e., the (s-1)th field, can be corrected to the beam-off start time of the sth field by referring to the above definition. Here, Figure 13 This diagram illustrates inter-field correction according to a specific embodiment of the present application, where the correction of the (s-1)th field is performed to the beam-off start time of the s-th field. At this point, for the beamout data within the (s-1)th field, the reference time is the start time of the beamout interval of the corrected field, that is, the reference time is the start time of the beamout interval of field s. As mentioned above.

[0320] Here, decay correction factors can be used. Or, more specifically, using the decay-corrected correction weights, the i-th coincidence event within the off-beam interval of field s-1 is corrected to the start time of the off-beam interval of field s. Accordingly, field s-1 is corrected to the start time of the off-beam interval of field s. The decay correction factor is:

[0321]

[0322] At this point, the correction weight corresponding to the i-th coincidence event within the beamout interval of field s-1 can be expressed as:

[0323]

[0324] Each matching event has a weight. The weights after decay correction are .

[0325] Furthermore, the second intermediate reconstructed image can be obtained by using the beamoff data of field s-1 after inter-field correction, which is the beamoff reconstructed image after inter-field decay correction. In this embodiment, any suitable reconstruction algorithm can be used for image reconstruction, such as the OSEM-ToF algorithm mentioned above, which will not be elaborated here. In a specific example, the beamoff interval of the (s-1)th field is analyzed using the OSEM-ToF method. The matching event subset is reconstructed to obtain the Beam-off PET image of the (s-1)th field after decay correction to the s-th field. .

[0326] Furthermore, for any subsequent shooting field s, the image can be reconstructed based on the first intermediate image. Second intermediate image The difference yields the second off-beam reconstruction image of the s-th field. .

[0327] In summary, the beam separation formula for the off-beam interval used for inter-beam correction and reconstruction can be expressed as:

[0328] .

[0329] This application's embodiments effectively solve the signal aliasing problem in multi-field delivery modes through innovative field separation technology. On one hand, by identifying the beam-on interval and extracting coincidence events during high-count periods for image reconstruction, interference from low-count periods is avoided, ensuring effective utilization of the peak signal. In a further embodiment, based on the identified coincidence events corresponding to the beam-off interval, a decay correction technique based on multi-exponential fitting is used. This involves establishing a multi-exponential decay model corresponding to multiple nuclide categories, fitting the temporal distribution count data of coincidence events in the beam-off interval, determining the initial activity coefficient of each nuclide category at the reference time, and then calculating a decay correction factor to correct the residual signal. This compensates for the decrease in count value caused by the decay of radionuclides over time, enabling the low-activity signal in the beam-off interval to be effectively used for reconstruction, further improving the image reconstruction effect.

[0330] Furthermore, this embodiment of the application corrects the residual signal of the previous radiation field to the reference time of the current radiation field, and eliminates signal interference between radiation fields through differential processing, thereby achieving precise separation and reconstruction of the independent dose distribution of each radiation field. Thus, through the separated independent radiation field images, the actual range of each radiation field (such as the Bragg peak position) can be accurately verified, providing real-time feedback for adaptive radiotherapy and significantly improving the accuracy of particle therapy range verification and the safety and effectiveness of treatment.

[0331] Furthermore, the technical solution of this application embodiment also has good compatibility, is compatible with existing fully digital PET equipment, can realize innovative field separation function without hardware modification, and is applicable to various particle radiotherapy scenarios such as proton therapy and heavy ion therapy, and has strong universality.

[0332] The steps and sub-steps described in the embodiments of this application can be executed independently or separately, or they can be combined or merged without contradiction. Furthermore, the order of steps in the embodiments of this application is not absolutely limited; the execution order of some steps can be adjusted or they can be executed in parallel according to actual needs, provided that the technical solution is not affected.

[0333] In some embodiments of this application, an electronic device is also provided, which may include a processor and a memory storing a computer program, the processor being configured to perform the methods of any embodiment of this application when running the computer program.

[0334] Figure 14 A schematic diagram of an exemplary electronic device 1800 that can implement the methods of embodiments of this application is shown. In some embodiments, it may include more or fewer electronic devices than shown. In some embodiments, it may be implemented using a single or multiple electronic devices. In some embodiments, it may be implemented using cloud-based or distributed electronic devices.

[0335] like Figure 14 As shown, the electronic device 1800 includes a processor 1801, which can perform various appropriate operations and processes based on programs and / or data stored in read-only memory (ROM) 1802 or programs and / or data loaded from storage portion 1808 into random access memory (RAM) 1803. The processor 1801 can be a single-core or multi-core processor, or may include multiple processors. In some embodiments, the processor 1801 may include a general-purpose main processor (such as a CPU) and one or more special coprocessors, such as a graphics processing unit (GPU), a neural network processor (NPU), a digital signal processor (DSP), or other general-purpose or application-specific integrated circuits. Various programs and data required for the operation of the electronic device 1800 are also stored in RAM 1803. The processor 1801, ROM 1802, and RAM 1803 are interconnected via bus 1804. An input / output (I / O) interface 1805 is also connected to bus 1804.

[0336] The processor and memory described above are used together to execute a program stored in the memory. When the program is executed by a computer, it can implement the steps or functions of the methods described in the above embodiments.

[0337] The following components are connected to I / O interface 1805: an input section 1806 including a keyboard, mouse, etc.; an output section 1807 including a display and speakers, etc.; a storage section 1808 including a hard disk, etc.; and a communication section 1809 including a network interface card such as a LAN card and a modem, etc. The communication section 1809 performs communication processing via a network such as the Internet. Drive 1810 is also connected to I / O interface 1805 as needed. Removable media 1811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 1810 as needed so that computer programs read from them can be installed into storage section 1808 as needed.

[0338] Figure 14 The electronic device shown in this application is merely illustrative; however, the electronic device in embodiments of this application may also include components that are more advanced than those described above. Figure 14 The electronic device shown has more or fewer components or has more or fewer components than the one shown. Figure 14 The embodiments shown have the same, partially the same, or different architectures.

[0339] Although not shown, some embodiments also provide a computer-readable storage medium storing a computer program configured to be run to perform the methods of any of the embodiments of this application. When executed, the computer program is capable of performing the functions corresponding to the various steps of the methods described in the above embodiments. The computer program can also run on electronic devices as described in the embodiments of this application.

[0340] The storage medium in embodiments of this application includes non-volatile and / or volatile articles that can store information by any method or technology. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0341] Those skilled in the art will understand that the embodiments of this specification can be implemented in various forms, such as methods, systems, or computer program products. Therefore, those skilled in the art will realize that the functional modules / units or controllers and related method steps described in the above embodiments can be implemented in software, hardware, or a combination of software and hardware.

[0342] Unless explicitly stated otherwise, the actions or steps of the methods and procedures described in the embodiments of this application do not necessarily have to be performed in a specific order and can still achieve the desired results. In some implementations, multitasking and parallel processing are also possible or may be advantageous.

[0343] This document describes several embodiments, but for the sake of brevity, the descriptions of the embodiments are not exhaustive, and identical or similar features or parts between the embodiments may be omitted. In this document, "one embodiment," "some embodiments," "example," "specific example," or "some examples" refers to at least one embodiment or example applicable to this application, but not all embodiments. The above terms do not necessarily mean referring to the same embodiment or example. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of the different embodiments or examples.

[0344] The exemplary systems and methods of this application have been specifically shown and described with reference to the above embodiments, which are merely examples of the best mode for implementing the systems and methods. Those skilled in the art will understand that various changes can be made to the embodiments of the systems and methods described herein without departing from the spirit and scope of the invention as defined in the appended claims when implementing the systems and / or methods.

Claims

1. A method of PET imaging for multi-field particle therapy, characterized by, The method comprises the following steps: According to the time distribution characteristics and the count value size of the collected single event data, one or more time intervals of each field are identified, the time intervals include a high count value period and a low count value period, and each time interval corresponds to a particle beam delivery state; Coincidence processing is performed on the single event data to generate coincidence event data; Image reconstruction is performed on the coincidence event data in the identified time interval of each field to obtain a corresponding reconstructed image of each field.

2. The PET imaging method of claim 1, wherein, According to the time distribution characteristics of the collected single event data, one or more time intervals of each field are identified, which comprises the following steps: Statistical histograms of single event data of each field are generated by counting the distribution of single event data of each field over time; A set count threshold is determined and intersected with the statistical histogram of single event data of each field; The one or more time intervals of each field are determined according to the intersection.

3. The PET imaging method of claim 2, wherein, The one or more time intervals of each field are determined according to the intersection, which comprises the following steps: A period between an initial intersection and a final intersection is determined as an on-beam interval of each field.

4. The PET imaging method of claim 3, wherein, The one or more time intervals of each field are determined according to the intersection, which comprises the following steps: A period between an odd-numbered intersection and a subsequent adjacent even-numbered intersection is determined as a high count value period of the on-beam interval.

5. The PET imaging method of claim 2, wherein, The one or more time intervals of each field are determined according to the intersection, which comprises the following steps: A period after the final intersection in each field is determined as an off-beam interval of each field.

6. The PET imaging method of claim 1, wherein, The identified time interval includes an on-beam interval; The image reconstruction of the coincidence event data in the identified time interval of each field to obtain a corresponding reconstructed image of each field comprises the following steps: The image reconstruction of the coincidence event data in the on-beam interval of each field to obtain a corresponding on-beam reconstructed image of each field.

7. The PET imaging method of claim 6, wherein, The on-beam interval includes a plurality of high count value periods; The image reconstruction of the coincidence event data in the on-beam interval of each field to obtain a corresponding on-beam reconstructed image of each field comprises the following steps: The coincidence event data in the plurality of high count value periods of the on-beam interval of each field is extracted to obtain a first coincidence event set of each field; The first coincidence event set of each field is used for image reconstruction to obtain the on-beam reconstructed image of each field.

8. The PET imaging method of claim 1, wherein, The identified time interval also includes an off-beam interval; The image reconstruction of the coincidence event data in the identified time interval of each field to obtain a corresponding reconstructed image of each field comprises the following steps: The time distribution-based count data of the coincidence event in the off-beam interval of each field is decay corrected to obtain a decay correction factor; The coincidence event data in the off-beam interval of each field and the corresponding decay correction factor are used for image reconstruction to obtain a corresponding off-beam reconstructed image of each field.

9. The PET imaging method of claim 8, wherein, The decay correction processing of the coincidence event data in the off-beam interval of each field comprises the following steps: According to prior information, a plurality of nuclide categories corresponding to the coincidence event data in the off-beam interval of each field are obtained; A multi-exponential fitting is performed according to the time distribution-based count data of the coincidence event in the off-beam interval of each field and the plurality of nuclide categories to obtain a sum of initial activities of all nuclides and a sum of decay activities of all nuclides. The decay correction factor corresponding to each coincidence event in the off-beam interval of each field is determined according to the sum of initial activities of all the nuclides and the sum of decay activities of all the nuclides.

10. The PET imaging method of claim 9, wherein, The multi-exponential fitting based on the time distribution based count data of the coincidence events in the off-beam interval of each field and the nuclide categories comprises: A multi-exponential decay model corresponding to the plurality of nuclide categories is established, wherein the parameters of the multi-exponential decay model include the initial activity coefficients and decay coefficients of each nuclide category; The multi-exponential fitting based on the time distribution based count data of the coincidence events at least determines the initial activity coefficients of each nuclide category at a reference time based on the multi-exponential decay model; The sum of initial activities of all the nuclides is calculated according to the initial activity coefficients of each nuclide category at the reference time; The sum of decay activities of all the nuclides of each coincidence event at any time is calculated by using the multi-exponential decay model containing the determined initial activity coefficients.

11. The PET imaging method of claim 10, wherein, The decay coefficient of each nuclide category is determined according to the half-life of the nuclide category; The multi-exponential fitting based on the time distribution based count data of the coincidence events at least determines the initial activity coefficients of each nuclide category at a reference time based on the multi-exponential decay model, comprising: The initial activity coefficients of each nuclide category at the reference time are estimated by using the linear least square method to perform multi-exponential fitting with the time distribution based count data of the coincidence events as input to obtain the best square approximation.

12. The PET imaging method of claim 10, wherein, The multi-exponential fitting based on the time distribution based count data of the coincidence events at least determines the initial activity coefficients of each nuclide category at a reference time based on the multi-exponential decay model, comprising: The initial activity coefficients of each nuclide category at the reference time and the decay coefficients of each nuclide category are estimated by using the nonlinear least square method to perform multi-exponential fitting with the time distribution based count data of the coincidence events as input to obtain the best square approximation.

13. The PET imaging method of claim 12, wherein, The initial activity coefficients of each nuclide category at the reference time and the decay coefficients of each nuclide category are estimated by using the nonlinear least square method to perform multi-exponential fitting with the time distribution based count data of the coincidence events as input to obtain the best square approximation, comprising: When performing the multi-exponential fitting, the decay coefficients of each nuclide category are limited within a preset constraint range.

14. The PET imaging method of claim 9, wherein, The coincidence event data in the identified time interval of each field is subjected to image reconstruction to obtain the corresponding reconstructed image of each field, and the method further comprises: Before the decay correction processing, the coincidence event data in the off-beam interval of each field is subjected to random correction and / or background correction.

15. The PET imaging method of claim 14, wherein, The coincidence event data in the off-beam interval of each field is subjected to random correction and / or background correction, comprising: At least the coincidence event data in the off-beam interval of each field is subjected to statistical processing to form a first event count set based on time distribution; At least the coincidence event data in the off-beam interval of each field is subjected to delay processing and statistical processing to form a second event count set based on time distribution; The random correction and / or background correction is performed according to the difference between the first event count set and the second event count set.

16. The PET imaging method of claim 15, wherein, the coincidence event data of each of the beam-on intervals is statistically processed to form a first event count set based on time distribution, including: at least equally interval time binning the coincidence event data of each of the beam-on intervals by a preset time length to obtain a first bin set including a plurality of first bins, wherein each first bin includes coincidence event counts falling within a respective first bin time range; the coincidence event data of each of the beam-on intervals is at least delay processed and statistically processed to form a second event count set based on time distribution, including: at least delay processing the coincidence event data of each of the beam-on intervals to obtain delay coincidence event data; equally interval time binning the delay coincidence event data by the preset time length to obtain a second bin set including a plurality of second bins, wherein each second bin includes delay coincidence event counts falling within a respective second bin time range, and the number of first bins is equal to the number of second bins and the bin time corresponds one by one.

17. The PET imaging method of claim 16, wherein, the random correction based on the difference between the first event count set and the second event count set, including: subtracting the delay coincidence event counts of the corresponding second bins from the coincidence event counts of each first bin in the first bin set to obtain first difference counts of each third bin after random correction.

18. The PET imaging method of claim 16, wherein, the background correction based on the difference between the first event count set and the second event count set, including: subtracting the delay coincidence event counts of the corresponding second bins from the coincidence event counts of each first bin in the first bin set to obtain first difference counts of each third bin; determining the minimum value of the first difference counts of each third bin as the background level; subtracting the background level from the coincidence event counts of each first bin in the first bin set to obtain second difference counts of each fourth bin after background correction.

19. The PET imaging method of claim 17 or 18, wherein, the random correction and / or background correction based on the difference between the first event count set and the second event count set, including: removing bins before the start time of each beam-on interval, and using the remaining bin set for subsequent decay correction processing.

20. The PET imaging method of claim 19, wherein, the random correction and / or background correction based on the difference between the first event count set and the second event count set, further including: converting the start time of each bin in the remaining bin set to a relative start time relative to the start time of each beam-on interval to form a relative time vector; forming an observation value vector from the bin counts of the remaining bin set after random correction and / or background correction, wherein the relative time vector and the observation value vector are used for subsequent multi-exponential fitting.

21. The PET imaging method of any one of claims 8 to 18, wherein, the image reconstruction using the coincidence event data in the beam-on intervals of each field of view and the corresponding decay correction factors to obtain the corresponding beam-on reconstruction images of each field of view, including: correcting the coincidence events in the beam-on intervals of each field of view to the start time of the beam-on intervals of the field of view using the decay correction factors and reconstructing to obtain a first beam-on reconstruction image.

22. The PET imaging method of any one of claims 8 to 18, wherein, The image reconstruction using coincidence event data in the off-beam interval of each field of view and the corresponding decay correction factor obtains an off-beam reconstruction image corresponding to each field of view, and comprises: For the first field of view, the coincidence event in the off-beam interval of the first field of view is corrected using the decay correction factor to correct the start time of the off-beam interval of the first field of view, and a second off-beam reconstruction image of the first field of view is obtained by reconstruction; For any subsequent field of view s, the coincidence event in the off-beam interval of the s-th field of view is corrected using the decay correction factor to correct the start time of the off-beam interval of the s-th field of view, and a first intermediate reconstruction image of the s-th field of view is obtained by reconstruction, the coincidence event in the off-beam interval of the s-1-th field of view is corrected using the decay correction factor to correct the start time of the off-beam interval of the s-th field of view, and a second intermediate reconstruction image of the s-1-th field of view is obtained by reconstruction, and a second off-beam reconstruction image of the s-th field of view is obtained according to the difference between the first intermediate reconstruction image and the second intermediate reconstruction image.

23. The PET imaging method of any one of claims 1 to 18, wherein, Further comprising: Collecting single event data generated by the PET detector during multi-field-of-view particle therapy, the single event data including time, energy and position information of the event.

24. An electronic device, comprising: Comprise: A processor and a memory storing a computer program, the processor being configured to implement the method of any one of claims 1 to 23 when running the computer program.

25. A storage medium, characterized by The storage medium stores a computer program configured to be run to implement the method of any one of claims 1 to 23.

Images