SPECT image continuous acquisition and quantitative reconstruction analysis method

Abstract

The invention provides an SPECT image continuous acquisition and quantitative reconstruction analysis method, and relates to the technical field of nuclear medicine image analysis, and the method comprises the following steps: S1, continuous acquisition: obtaining original data under SPECT continuous acquisition, enabling a probe and a scanning bed to synchronously and continuously move in a scanning process, and collecting the data, the method comprises the steps of S1, data preprocessing: obtaining corrected table mode data used for reconstruction, and generating file data recording SPECT mechanical motion information, and S3, quantitative reconstruction: obtaining a high-resolution quantitative activity distribution image in combination with a table mode iteration reconstruction technology provided by the invention. On one hand, the SPECT data acquisition efficiency is improved, the conventional SPECT acquisition time and dosage are effectively reduced, and the image signal-to-noise ratio is improved; on the other hand, original data information is accurately reserved, high-precision physical modeling and correction are matched, high-precision quantitative reconstruction can be achieved, image blurring caused by continuous rotation discretization is effectively avoided, and advanced clinical requirements including dynamic imaging can be better met.

Description

Technical Field

[0001] The invention relates to the technical field of nuclear medicine image analysis, in particular to a SPECT image continuous acquisition and quantitative reconstruction analysis method. Background Art

[0002] Single Photon Emission Computed Tomography (SPECT) is an important nuclear medicine imaging technology. It reflects physiological metabolic information by tracking the specific distribution of radioactive drugs in the human body. It currently plays an important role in the diagnosis and treatment of tumors and cardiovascular diseases. The most mainstream clinical system is the dual-probe parallel hole SPECT system. The direction information of the incident gamma photons is obtained through the parallel hole collimator. The data at different angles are collected in conjunction with the step rotation of the probe. Combined with the three-dimensional tomographic reconstruction technology, the three-dimensional distribution of radioactive drugs in the body can be visualized. The reconstruction algorithm commonly used in early clinical practice is the filtered back-projection (FBP) algorithm. However, due to the inability to accurately correct the physical effects such as photon attenuation, collimator response, and detector blur during the imaging process, the FBP reconstructed image has low resolution and poor signal-to-noise ratio, and the true activity distribution information of the drug cannot be obtained. The maximum likelihood expectation maximization (MLEM) algorithm based on ordered subset acceleration is commonly used in clinical practice. Combined with accurate physical correction, accurate nuclear medicine quantitative reconstruction can be achieved.

[0003] Traditional parallel hole SPECT adopts a step-by-step acquisition mode of rotation, static acquisition, and then rotation to the next angle. There are two main problems: 1) Low efficiency. The mechanical movement time can reach 25% of the actual acquisition time, and the probe does not collect data during the mechanical movement, which not only reduces the acquisition efficiency, but also affects the quality of the final reconstructed image. 2) Slow speed. Since SPECT needs to collect projection data within a range of at least 180 degrees for effective imaging, the step-by-step acquisition speed of continuous rotation and stopping is slow, and it takes a long time to complete a single tomographic acquisition. It also cannot meet clinical needs with high requirements for time characteristics, including dynamic imaging. In addition to the acquisition mode, the traditional reconstruction algorithm is based on multi-angle projection data for reconstruction and cannot effectively process continuously rotating acquisition data. If it is discretized into projection data at several specific angles, some position information will inevitably be lost, resulting in blurred reconstructed images; and the discretization processing with high position accuracy will lead to excessive projection data, increased calculation amount, and significantly reduced reconstruction speed.

[0004] Table pattern reconstruction algorithms have been widely used in positron emission tomography (PET), but rarely in SPECT. There are two main considerations. First, in terms of technical difficulty, the table pattern algorithm needs to know the relatively accurate position of each (coincidence) event to calculate projection and back projection, and the crosstalk between the detection units of the SPECT detector brings difficulty to table pattern reconstruction; in contrast, PET usually uses discrete crystal readout, which can better avoid the position readout error between each detection unit, so the technical difficulty of PET using table pattern reconstruction is relatively lower. In addition, in terms of clinical manifestations, PET imaging uses coincidence signals between detection units, and the number of coincidence units is usually much larger than the number of acquisition events, so PET table pattern reconstruction can greatly reduce the amount of calculation and save reconstruction time; while the number of acquisition events in SPECT is usually not significantly reduced compared with the number of projection detection units, so from the perspective of calculation, it is not meaningful to use table pattern reconstruction. However, considering the urgent clinical need for SPECT signal-to-noise ratio improvement and dynamic imaging, as well as the importance of continuous acquisition for SPECT signal-to-noise ratio improvement and dynamic imaging, the advantages of table-mode reconstruction in high reconstruction resolution in the above scenarios will be demonstrated. Since PET itself uses a probe-fixed continuous acquisition mode, the significance of table-mode reconstruction for signal-to-noise ratio and dynamic imaging is almost the same as that of traditional reconstruction.

[0005] In summary, the existing SPECT acquisition method has low efficiency and poor signal-to-noise ratio; when used for dynamic imaging, rapid rotation of the probe will cause blurred image reconstruction, and the traditional commonly used SPECT quantitative reconstruction technology cannot solve this problem well. Summary of the invention

[0006] Technical issues solved

[0007] In view of the shortcomings of the prior art, the present invention provides a method for continuous acquisition and quantitative reconstruction and analysis of SPECT images, which solves the problems of low efficiency and poor signal-to-noise ratio of traditional SPECT acquisition methods, as well as poor resolution of traditional reconstruction methods under continuous rotation acquisition conditions.

[0008] Technical Solution

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for continuous acquisition and quantitative reconstruction and analysis of SPECT images, comprising the following steps:

[0010] S1, continuous acquisition: obtain the raw data under SPECT continuous acquisition. During the scanning process, the probe and the scanning bed move synchronously and continuously to collect data;

[0011] S2, data preprocessing: obtaining the corrected table mode data for reconstruction and generating file data recording the SPECT mechanical motion information;

[0012] S3. Quantitative reconstruction: A high-resolution quantitative activity distribution image is obtained by combining the table pattern iterative reconstruction technology proposed in the present invention.

[0013] Preferably, step S1 comprises the following steps:

[0014] S11, initialization settings: the probe is placed in L mode or H mode, and the height and axial position of the scanning bed are adjusted;

[0015] S12, motion acquisition: the probe moves continuously, including rotation, tangential translation and radial movement, and the scanning bed is stationary or moves continuously synchronously;

[0016] S13. Signal acquisition and storage: Synchronously collect photon signals during mechanical movement, record information such as the photon action position and energy, and continuously record the system's mechanical movement information, including the probe rotation angle, tangential position, radial position, and axial position of the scanning bed.

[0017] Preferably, step S2 comprises the following steps:

[0018] S21, photon information correction: including but not limited to uniformity correction, energy correction; distinguishing main energy window and scattering window data according to energy; transforming photon position coordinates to meet the SPECT standardized projection format;

[0019] S22, mechanical motion correction: correcting the mechanical motion position information, including but not limited to probe tilt correction and rotation center correction;

[0020] S23, generating table mode data: for each photon, the above corrected information is recorded to generate table mode data;

[0021] S24, generate orbital parameter file: generate an orbital parameter file including information such as probe angle position, radial position, bed axial position, acquisition time, acquisition frame duration, etc.

[0022] Preferably, step S3 comprises the following steps:

[0023] S31, sensitivity map calculation: read the orbit parameter file and calculate the sensitivity map according to the scanning orbit information;

[0024] S32, subset division: Different from the traditional parallel hole SPECT which divides the projection data into subsets according to the angle, the present invention divides the SPECT table mode data into different subsets in sequence according to the acquisition time sequence for algorithm acceleration;

[0025] S33, Iterative Reconstruction: The present invention proposes a system modeling method based on line tracking for iterative reconstruction, which incorporates collimator response, attenuation effect and detector blur effect into the same modeling framework for synchronous parallel calculation to achieve precise physical correction. Combined with the quantitative conversion factor and acquisition parameters of the system calibration, a high-resolution quantitative reconstructed image in Bq / mL is output.

[0026] Preferably, step S33 includes the following steps:

[0027] S3301, initialize the line-tracing based projector, and calculate the photon position (r x , r y ) Initialize the line tracking projector, the initialization parameters include the number of tracking lines, angle information, endpoint positions and corresponding weights; sample the possible tracking line angles according to the collimator parameters, and perform the sampling in (r x , r y ) neighborhood, the tracking line endpoint position is sampled, and the tracking line weight factor is determined by the collimator response and the detector blur effect;

[0028] S3302, performing forward projection in combination with a line tracing algorithm, wherein the contribution factor of a pixel to the photon position described in S3301 is proportional to the intersection length of the pixel and the tracing line, and synchronously calculating the attenuation effect of objects at different depths based on the line tracing process;

[0029] S3303, generating a projection correction factor according to the real table mode data compared with the previous projection result;

[0030] S3304, performing back projection in combination with a line tracking algorithm, and synchronously calculating the attenuation effect of objects at different depths based on the line tracking process;

[0031] S3305, correcting and updating the current image in combination with the sensitivity map and the back-projection result;

[0032] S3306, quantitative conversion: according to the quantitative calibration factor of the system and in combination with the acquisition parameter settings, the iteratively reconstructed output image is converted into an image in Bq / mL.

[0033] Preferably, the sensitivity map calculation process in S31 is implemented by a back-projection operation in S3304.

[0034] Preferably, the real table mode data in S3303 takes a negative value when the source is a scattering window.

[0035] Beneficial Effects

[0036] The present invention provides a method for continuous acquisition and quantitative reconstruction and analysis of SPECT images.

[0037] Beneficial effects:

[0038] On the one hand, it improves the efficiency of SPECT data acquisition, effectively reduces the conventional SPECT acquisition time and dose, and improves the image signal-to-noise ratio; on the other hand, it accurately retains the original data information, and combined with high-precision physical modeling and correction, it can achieve high-precision quantitative reconstruction, effectively avoiding image blur caused by continuous rotation discretization, and can better meet advanced clinical needs including dynamic imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Flow chart for continuous acquisition and quantitative reconstruction analysis of SPECT images;

[0040] Figure 2 Schematic diagram for modeling the response of the SPECT system;

[0041] Figure 3 Spiral acquisition and step acquisition projections for the Jaszczak model;

[0042] Figure 4 Reconstruct different cross-sectional slices for Jaszczak spiral acquisition + table mode;

[0043] Figure 5 NEMA three-point source images with different acquisition conditions and reconstruction methods.

[0044] in Figure 3 (a) simulates spiral orbit acquisition. Figure 3 (b) is a single bed step-by-step acquisition projection diagram. Figure 5 From left to right, they correspond to traditional step acquisition + projection mode reconstruction, continuous acquisition + table mode reconstruction, and continuous acquisition + projection mode reconstruction. From top to bottom, they correspond to three point sources at different positions. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] Embodiment 1:

[0047] like Figure 1 , Figure 2 As shown, an embodiment of the present invention provides a method for continuous acquisition and quantitative reconstruction and analysis of SPECT images, comprising the following steps:

[0048] S1, continuous acquisition: obtaining raw data under SPECT continuous acquisition. During the scanning process, the probe and the scanning bed move synchronously and continuously and collect data. Different from the conventional parallel hole SPECT scanning method using rotational stepping, data is not collected during the rotation of the probe. Step S1 includes the following steps:

[0049] S11, initialization settings: the probe is placed in L mode or H mode, and the height and axial position of the scanning bed are adjusted;

[0050] S12, motion acquisition: the probe moves continuously, including rotation, tangential translation and radial movement, and the scanning bed is stationary or moves continuously synchronously;

[0051] S13. Signal acquisition and storage: Synchronously collect photon signals during mechanical movement, record information such as the photon action position and energy, and continuously record the system's mechanical movement information, including the probe rotation angle, tangential position, radial position, and axial position of the scanning bed.

[0052] S2, data preprocessing: obtain the corrected table mode data for reconstruction, and generate file data recording the SPECT mechanical motion information. Different from the traditional parallel hole SPECT preprocessing to obtain projections at various angles for subsequent reconstruction, the present invention retains the table mode data format, and compared to the PET table mode data which only needs to record the photon position and energy, the table mode data proposed by the present invention also includes the SPECT mechanical motion information. Step S2 includes the following steps:

[0053] S21, photon information correction: including but not limited to uniformity correction, energy correction; distinguishing main energy window and scattering window data according to energy; transforming photon position coordinates to meet the SPECT standardized projection format;

[0054] S22, mechanical motion correction: correcting the mechanical motion position information, including but not limited to probe tilt correction and rotation center correction;

[0055] S23, generating table mode data: for each photon, the above corrected information is recorded to generate table mode data;

[0056] S24, generate orbital parameter file: generate an orbital parameter file including information such as probe angle position, radial position, bed axial position, acquisition time, acquisition frame duration, etc.

[0057] S3, quantitative reconstruction: Combined with the table mode iterative reconstruction technology proposed in the present invention, a high-resolution quantitative activity distribution image is obtained. Step S3 includes the following steps:

[0058] S31, sensitivity map calculation: read the orbit parameter file, calculate the sensitivity map according to the scanning orbit information, and the sensitivity map calculation process in S31 is realized by the back projection operation in S3304.

[0059] S32. Subset division: Different from the traditional parallel hole SPECT which divides the projection data into subsets according to the angle, the present invention divides the SPECT table mode data into different subsets in sequence according to the acquisition time sequence for algorithm acceleration.

[0060] S33, iterative reconstruction. Different from the traditional method of using the point spread function method to model the collimator response and the line tracking method to calculate the attenuation effect, the present invention adopts an improved line tracking method to simultaneously model the collimator response and the attenuation effect. In addition, the present invention also incorporates the problem that the detector blur is not conducive to the table mode reconstruction calculation into the framework of the line tracking model to solve it. Step S33 includes the following steps:

[0061] S3301, initialize the projector based on line tracking. Different from the traditional method based on the discrete SPECT pixel number or PET crystal number, the present invention is based on the table mode data photon position (r x , r y ) initializes the line tracking projector, and avoids the additional convolution operation performed by the traditional SPECT reconstruction algorithm after calculating the projection to characterize the detector blur effect. The initialization parameters include the number of tracking lines, angle information, endpoint positions and corresponding weights, as shown in the following figure. Figure 2 Taking into account the blurring effect of the detector's inherent resolution, we Figure 2 The figure shows photons incident at different positions, and the final position is recorded as (r x , r y ) and in (r x , r y ) The endpoints of the tracking lines are sampled within a neighborhood. Different weights are then assigned to each tracking line. Different from the PET line tracking algorithm, which mainly considers the tracking line length and crystal attenuation when calculating the tracking line weight, the present invention mainly considers the influence of angle distribution and detector blur effect when calculating the SPECT tracking line weight. Where θ represents the angle between the tracking line and the detector normal, is the position vector of the endpoint of the tracing line relative to the photon action point, Indicates the sampling surface element corresponding to this position.

[0062] S3302, combining line tracking algorithm for front projection, unlike PET where the attenuation effect is the same everywhere on the response line, the SPECT attenuation effect changes with depth; also unlike the traditional SPECT reconstruction where the attenuation effect is only calculated along the detector normal direction, the present invention considers the attenuation effect of various possible photon incident trajectories (including oblique incidence), and initializes L atten =0, Y proj =0; the length of the intersection of the tracking line and pixel j is recorded as l j , the current pixel value is The linear attenuation coefficient is μ j , represents the attenuation effect at the current pixel j, and the geometric response factor p j =W lj , W is the weight factor of the line, calculated as in step 3301), the current projection is updated to The projection calculation here is to accumulate the different tracking lines of the current photon signal.

[0063] S3303: Generate projection correction factors based on real table mode data compared with previous projection results Where v represents the value of the current photon event after preprocessing correction (if it is a scattered photon, the value is negative).

[0064] S3304, combine the line tracking algorithm to perform back projection and initialize the back projection correction factor of pixel j but a j , p j The calculation method is the same as step 3302), where the summation is the accumulation of different photon signals and different tracking lines of each photon signal.

[0065] S3305: Correct and update the current image by combining the sensitivity map and the back-projection result s j is the sensitivity factor at pixel j. If the set iteration termination condition is not met, return to step 3301) and repeat a new round of iteration.

[0066] S3306, quantitative conversion: according to the quantitative calibration factor of the system and in combination with the acquisition parameter settings, the iteratively reconstructed output image is converted into an image in Bq / mL.

[0067] The table mode data of the Jaszczak model was collected based on the continuous rotation of the probe, and the table mode data of the spiral track of the synchronous movement of the bed was simulated and generated on this basis. In order to illustrate the acquisition characteristics, multi-angle projection data of the spiral track was generated at the same time, and compared with the traditional single-bed step-by-step acquisition projection map, such as Figure 3 shown. Figure 4The test results of the continuous acquisition + table mode reconstruction method proposed in the present invention on the Jaszczak model are shown, and it can be seen that the different cross-sectional features of the Jaszczak model are well demonstrated. This embodiment shows that the present invention can well meet the actual imaging requirements.

[0068] Embodiment 2

[0069] To illustrate the improvement of the SPECT imaging performance of the present invention, Figure 5 Another set of embodiments is shown, specifically comparing the resolution of point source imaging with reference to the NEMA standard. The three groups compared in the figure respectively use traditional step-by-step acquisition + traditional projection mode reconstruction, continuous acquisition + table mode reconstruction proposed by the present invention, and continuous acquisition + traditional projection mode reconstruction. Figure 5 It can be seen that for point sources far away from the center of the field of view, the arrow-indicated part of the reconstruction result of continuous acquisition + traditional projection mode is significantly brighter, indicating that its tangential resolution is worse, while the difference between the other two groups is not much. This is because for projection mode reconstruction, continuous acquisition brings additional rotational motion blur, while traditional step acquisition is performed in a stationary state after the probe completes rotation, so there is no problem of rotational motion blur. This embodiment shows that the resolution of continuous acquisition imaging is worse than that of step-by-step acquisition when using traditional projection mode reconstruction, and the table mode reconstruction proposed in the present invention can significantly improve the resolution characteristics of continuous acquisition imaging.

[0070] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for continuous acquisition and quantitative reconstruction and analysis of SPECT images, characterized in that: The following steps are involved: S1, continuous acquisition: obtain the raw data under SPECT continuous acquisition. During the scanning process, the probe and the scanning bed move synchronously and continuously to collect data; S2, data preprocessing: obtaining the corrected table mode data for reconstruction and generating file data recording the SPECT mechanical motion information; S3. Quantitative reconstruction: A high-resolution quantitative activity distribution image is obtained by combining the table pattern iterative reconstruction technology proposed in the present invention.

2. A method for continuous acquisition and quantitative reconstruction and analysis of SPECT images according to claim 1, characterized in that: Step S1 includes the following steps: S11, initialization settings: the probe is placed in L mode or H mode, and the height and axial position of the scanning bed are adjusted; S12, Motion acquisition: The probe moves continuously, including rotation, tangential translation and radial movement, and the scanning bed is stationary or moves continuously synchronously; S13. Signal acquisition and storage: Synchronously collect photon signals during mechanical movement, record information such as the photon action position and energy, and continuously record the system's mechanical movement information, including the probe rotation angle, tangential position, radial position, and axial position of the scanning bed.

3. A method for continuous acquisition and quantitative reconstruction and analysis of SPECT images according to claim 1, characterized in that: Step S2 includes the following steps: S21, photon information correction: including but not limited to uniformity correction, energy correction; distinguishing main energy window and scattering window data according to energy; transforming photon position coordinates to meet the SPECT standardized projection format; S22, mechanical motion correction: correcting the mechanical motion position information, including but not limited to probe tilt correction and rotation center correction; S23, generating table mode data: for each photon, the above corrected information is recorded to generate table mode data; S24, generate orbital parameter file: generate an orbital parameter file including information such as probe angle position, radial position, bed axial position, acquisition time, acquisition frame duration, etc.

4. The method for continuous acquisition and quantitative reconstruction and analysis of SPECT images according to claim 1, characterized in that: Step S3 includes the following steps: S31, sensitivity map calculation: read the orbit parameter file and calculate the sensitivity map according to the scanned orbit information; S32, subset division: Different from the traditional parallel hole SPECT which divides the projection data into subsets according to the angle, the present invention divides the SPECT table mode data into different subsets in sequence according to the acquisition time sequence for algorithm acceleration; S33, Iterative Reconstruction: The present invention proposes a system modeling method based on line tracking for iterative reconstruction, which incorporates collimator response, attenuation effect and detector blur effect into the same modeling framework for synchronous parallel calculation to achieve precise physical correction. Combined with the quantitative conversion factor and acquisition parameters of the system calibration, a high-resolution quantitative reconstructed image in Bq / mL is output.

5. A method for continuous acquisition and quantitative reconstruction and analysis of SPECT images according to claim 4, characterized in that: Step S33 includes the following steps: S3301, initialize a line-tracking projector (projector), initialize the line-tracking projector according to the photon position (rx, ry) of the table mode data, and initialize the parameters including the number of tracking lines, angle information, endpoint positions and corresponding weights; sample possible tracking line angles according to collimator parameters, sample the tracking line endpoint positions in the (rx, ry) neighborhood according to the inherent resolution characteristics of the detector, and the tracking line weight factor is jointly determined by the collimator response and the detector blur effect; S3302, performing forward projection in combination with a line tracing algorithm, wherein the contribution factor of a pixel to the photon position described in S3301 is proportional to the intersection length of the pixel and the tracing line, and synchronously calculating the attenuation effect of objects at different depths based on the line tracing process; S3303, generating a projection correction factor according to the real table mode data compared with the previous projection result; S3304, performing back projection in combination with a line tracking algorithm, and synchronously calculating the attenuation effect of objects at different depths based on the line tracking process; S3305, correcting and updating the current image in combination with the sensitivity map and the back-projection result; S3306, quantitative conversion: according to the quantitative calibration factor of the system and in combination with the acquisition parameter settings, the iteratively reconstructed output image is converted into an image in Bq / mL.

6. A method for continuous acquisition and quantitative reconstruction and analysis of SPECT images according to claim 5, characterized in that: The sensitivity map calculation process in S31 is implemented through the back-projection operation in S3304.

7. A method for continuous acquisition and quantitative reconstruction and analysis of SPECT images according to claim 5, characterized in that: When the real table mode data source in S3303 is a scattering window, a negative value is taken.

Images