Search

EP-4737949-A1 - IMAGE RECONSTRUCTION METHOD AND SYSTEM, DETECTION DEVICE, ELECTRONIC DEVICE AND STORAGE MEDIUM

EP4737949A1EP 4737949 A1EP4737949 A1EP 4737949A1EP-4737949-A1

Abstract

The application discloses an image reconstruction method, a system, a detection device, an electronic device, and a storage medium, relating to the field of data processing. The image reconstruction method comprises: obtaining TOF resolutions of a portion of lines of response according to prior information; obtaining timing resolutions of all scintillation crystals according to the TOF resolutions of the portion of lines of response and a predetermined model; obtaining TOF resolutions of remaining lines of response according to the timing resolutions of all scintillation crystals and the predetermined model; and reconstructing an image according to all TOF resolutions and a first predetermined algorithm. The solution of the application applies differentiated TOF resolutions to different lines of response. In particular, by correlating the TOF resolutions of lines of response with the timing resolutions of corresponding scintillation crystals, the TOF resolutions of lines of response may be accurately estimated, thereby effectively enhancing image reconstruction quality .

Inventors

  • LI, ANG
  • FANG, LEI
  • LI, Bingxuan
  • ZHANG, BO

Assignees

  • RAYSOLUTION Healthcare Co., Ltd

Dates

Publication Date
20260506
Application Date
20241129

Claims (20)

  1. An image reconstruction method, characterized by comprising: obtaining TOF resolutions of a portion of lines of response according to prior information; obtaining timing resolutions of all scintillation crystals according to the TOF resolutions of the portion of lines of response and a predetermined model; obtaining TOF resolutions of remaining lines of response according to the timing resolutions of all scintillation crystals and the predetermined model; and reconstructing an image according to all TOF resolutions and a first predetermined algorithm.
  2. The image reconstruction method according to claim 1, characterized in that the obtaining timing resolutions of all scintillation crystals according to the TOF resolutions of the portion of lines of response and a predetermined model comprises: inputting the TOF resolutions of the portion of lines of response as input values respectively into the predetermined model to obtain the timing resolutions of all scintillation crystals; and the obtaining TOF resolutions of remaining lines of response according to the timing resolutions of all scintillation crystals and the predetermined model comprises: grouping the timing resolutions corresponding to two scintillation crystals located on the same line of response as one group, and inputting all groups of timing resolutions as input values respectively into the predetermined model to obtain the TOF resolutions of the remaining lines of response.
  3. The image reconstruction method according to claim 1, characterized in that the predetermined model is: σ ij 2 = σ i 2 + σ j 2 where σ i represents the timing resolution of scintillation crystal i, σ j represents the timing resolution of scintillation crystal j, and σ ij represents the TOF resolution of the line of response formed by scintillation crystal i and scintillation crystal j.
  4. The image reconstruction method according to claim 1, characterized in that the predetermined model is: σ ij 2 = σ i 2 + σ j 2 + δ r where σ i represents the timing resolution of scintillation crystal i, σ i represents the timing resolution of scintillation crystal j, δ r represents a variable related to photon incident angle, and σ ij represents the TOF resolution of the line of response formed by scintillation crystal i and scintillation crystal j.
  5. The image reconstruction method according to claim 1, characterized in that the prior information is obtained according to the following steps: conducting an image detection test under test conditions to obtain the number of coincidence events on all lines of response and time difference corresponding to two crystals of each coincidence event ; selecting lines of response where the number of coincidence events reaches a first predetermined threshold as target lines of response; and obtaining the TOF resolution of all the target lines of response as the prior information according to a second predetermined algorithm and the time difference of each coincidence event.
  6. The image reconstruction method according to claim 5, characterized in that the second predetermined algorithm is: σ = ∑ i = 1 n Δt i − Δt ¯ 2 n − 1 where n is the total number of coincidence events on a certain target line of response, Δt i is the time difference of the i-th coincidence event, Δt is the mean value of the coincidence time differences on the target line of response, σ is the standard deviation of the time differences on the target line of response, and the TOF resolution of the target line of response is 2.355σ.
  7. The image reconstruction method according to claim 5, characterized in that the first predetermined threshold is greater than or equal to 5.
  8. The image reconstruction method according to claim 5, characterized in that the test conditions comprise: a radioactive source that only emits positrons being selected as a target source; and/or the shape of the target source comprising a shell source, a line source, or a point source.
  9. The image reconstruction method according to claim 8, characterized in that when the target source is the shell source, before selecting lines of response where the number of coincidence events exceeds the first predetermined threshold as target lines of response, the steps for obtaining the prior information further comprise: excluding lines of response at a distance from the central axis of the shell source that reaches a second predetermined threshold.
  10. The image reconstruction method according to claim 9, characterized in that the second predetermined threshold is 80% of the maximum distance corresponding to the field of view of the detector ring.
  11. The image reconstruction method according to claim 8, characterized in that when the target source is the shell source, after selecting lines of response where the number of coincidence events exceeds the first predetermined threshold as target lines of response, the steps for obtaining the prior information further comprise: performing time-of-flight compensation on all coincidence events using a predetermined method.
  12. The image reconstruction method according to claim 11, characterized in that the predetermined method comprises: obtaining two intersection points between the line of response and the shell source, and obtaining the time-of-flight difference of coincidence events generated by annihilation at the two intersection points; plotting a time difference spectrum according to the time differences of the coincidence events to obtain two peaks, where the two peaks respectively correspond to the two intersection points between the line of response and the shell source; classifying the coincidence events into one of the two peaks using a k-means algorithm; and compensating the time-of-flight differences of the coincidence events according to the intersection point to which the respective coincidence events correspond and the calculated time-of-flight difference.
  13. The image reconstruction method according to claim 8, characterized in that the test conditions further comprise: the size of the target source being less than a predetermined requirement.
  14. The image reconstruction method according to claim 13, characterized in that the predetermined requirement comprises: the thickness of the shell source being less than or equal to 1 cm; or the diameter of the line source being less than or equal to 1 cm; or the diameter of the point source being less than or equal to 0.5 cm.
  15. The image reconstruction method according to claim 8, characterized in that the test conditions further comprise: the position of the target source ensuring that the connecting lines between any single detection unit and at least two detection units at different positions all pass through the target source.
  16. The image reconstruction method according to claim 15, characterized in that : the height extent of the shell source is parallel to the axis of the PET; or among the intersection points between the cross-section of the detector ring and the line source, at least three points are not collinear; or at least four of the point sources are not coplanar.
  17. The image reconstruction method according to claim 8, characterized in that the target source comprises Fluorine-18 ( 18 F), Sodium-22 ( 22 Na), or Germanium-68 ( 63 Ge).
  18. The image reconstruction method according to claim 1, characterized in that the first predetermined algorithm is: x j n + 1 = x j n ∑ i H ij N i A i ∑ i k ∈ S H ijk 1 ∑ j H ijk x j n + r ik + s ik N i A i where x j n is the j-th pixel value of the n-th iteration, N i and A i are respectively the normalization factor and attenuation factor of the i-th line of response, r ik and s j . are respectively the number of random and scatter coincidence events on the k-th time bin of the i-th line of response, S is the set of line of response numbers and time bin numbers where all list-mode events are located, H ij is the system matrix for non-TOF reconstruction, and H ijk is the system matrix for TOF reconstruction.
  19. The image reconstruction method according to claim 1, characterized in that the first predetermined algorithm is: x j l + 1 , n = x j l , n ∑ i H ij N i A i ∑ i k ∈ S l H ijk 1 ∑ j H ijk x j l , n + r ik + s ik N i A i where x j l , n is the j-th pixel value of the l-th subset of the n-th iteration, N i and A i are respectively the normalization factor and attenuation factor of the i-th line of response, r ik and s ik are respectively the number of random and scatter coincidence events on the k-th time bin of the i-th line of response, S l is the set of line of response number i and time bin number k where all list-mode events in the l-th subset are located, H ij is the system matrix for non-TOF reconstruction, and H ijk is the system matrix for TOF reconstruction.
  20. The image reconstruction method according to claim 1, characterized in that the first predetermined algorithm is: x j n + 1 = x j n ∑ i H ij N i A i ∑ i ∑ k H ijk Y ik ∑ j H ijk x j n + r ik + s ik N i A i where x j n is the j-th pixel value of the n-th iteration, N i and A i are respectively the normalization factor and attenuation factor of the i-th line of response, r ik and s ik are respectively the number of random and scatter coincidence events on the k-th time bin of the i-th line of response, H ij is the system matrix for non-TOF reconstruction, and H ijk is the system matrix for TOF reconstruction.

Description

The application claims priority to Chinese Patent Application No. 202311859527.5, filed on December 30, 2023, the entire contents of which are incorporated herein by reference. Technical Field The application relates to the field of image reconstruction, and specifically relates to an image reconstruction method, a system, a detection device, an electronic device, and a storage medium. Background Positron Emission Tomography (PET) is one of the most advanced molecular imaging technologies in the world today, with numerous functions such as early cancer detection, therapeutic efficacy evaluation, pharmacokinetics observation and the like. By imaging compounds labeled with radionuclides in living organisms, PET is capable of non-invasively, quantitatively, and dynamically evaluating the metabolic level, biochemical reaction, and functional activity of various functional organs in living organisms. With high sensitivity and accuracy, PET has an irreplaceable position in clinical medicine and drug development. At the beginning of development of PET, image reconstruction mainly relies on information regarding the number of coincidence events on the connecting line between detectors. However, early detectors had insufficient photon time of flight (Time of Flight, TOF) resolution performance, which may hardly bring any gain to image reconstruction. But with the development of PET detector technology in recent years, the TOF resolution of PET detectors has reached 200ps-600ps. With the support of such TOF information, the quality of TOF image reconstruction has been significantly improved compared to non-TOF image reconstruction. TOF resolution has also become a key metric for evaluating PET system performance, and meanwhile TOF image reconstruction has been widely used. In existing methods, the coincidence events detected along a line of response (LOR) are sorted into discrete time bins, with each time bin representing a corresponding spatial segment of the LOR. TOF image reconstruction requires modeling the contribution Hijk of gamma-ray photon emitted from the j-th voxel to the k-th time bin on the i-th LOR, where Hijk = HijKijk, which Kijk is commonly referred to as the TOF kernel. Typically, a Gaussian distribution is used to model the value of the TOF kernel in the k-th time bin, with the full width at half maximum (FWHM) of the Gaussian distribution being the TOF resolution of the PET system. It is generally assumed that all LORs have the same TOF resolution. Under this assumption, the TOF kernel of any LOR is modeled with the same standard deviation of the Gaussian distribution (i.e., the FWHM of the Gaussian distribution divided by 2.355). However, in actual systems, the TOF resolution differs among individual LORs. These differences distort the modeling of TOF information and therefore prevent TOF image reconstruction from achieving optimal performance. The content described in the background is provided solely to facilitate understanding of the related technologies in this field and shall not be construed as an admission of prior art. Summary Therefore, the application intends to provide an image reconstruction method and system, a detection device, an electronic device, and a storage medium, which may solve at least one problem existing in the prior art. In a first aspect, an image reconstruction method is provided, comprising: obtaining TOF resolutions of a portion of lines of response according to prior information; obtaining timing resolutions of all scintillation crystals according to the TOF resolutions of the portion of lines of response and a predetermined model; obtaining TOF resolutions of remaining lines of response according to the timing resolutions of all scintillation crystals and the predetermined model; and reconstructing an image according to all TOF resolutions and a first predetermined algorithm. In an embodiment of the application, the obtaining timing resolutions of all scintillation crystals according to the TOF resolutions of the portion of lines of response and a predetermined model comprises: inputting the TOF resolutions of the portion of lines of response as input values respectively into the predetermined model to obtain the timing resolutions of all scintillation crystals; and the obtaining TOF resolutions of the remaining lines of response according to the timing resolutions of all scintillation crystals and the predetermined model comprises: grouping the timing resolutions corresponding to two scintillation crystals located on the same line of response as one group, and inputting all groups of timing resolutions as input values respectively into the predetermined model to obtain the TOF resolutions of the remaining lines of response. In an embodiment of the application, the predetermined model is: σij2=σi2+σj2 where σi represents the timing resolution of scintillation crystal i, σj represents the timing resolution of scintillation crystal j, and σij represents the TOF reso