Search

CN-119228770-B - CT-guidance-based senile brain PET image space standardization system, method and device

CN119228770BCN 119228770 BCN119228770 BCN 119228770BCN-119228770-B

Abstract

The invention discloses a CT-guided senile brain PET image space standardization system, a method and a device, wherein the system comprises a PET/CT image acquisition and format conversion module, a PET/CT preprocessing module, a PET/CT image coarse standardization module, a PET/CT image fine standardization module and an automatic SUV extraction module, and (3) carrying out the senile brain PET molecular image space standardization by taking the low-dose CT brain structure image as an aid, and automatically extracting the SUV of the region of interest by adopting brain map partition of a standard space. The invention adopts a coarse-to-fine two-step walking strategy, provides an accurate, stable and user-friendly space standardization system for the senile brain PET image through an optimized process, and has important clinical application value and scientific research significance.

Inventors

  • ZHANG HONG
  • Hu Daoyan
  • ZHONG YAN
  • WANG JING
  • JIN CHENTAO

Assignees

  • 浙江大学

Dates

Publication Date
20260508
Application Date
20240929

Claims (9)

  1. 1. A senile brain PET image space standardization system based on CT guidance, characterized in that the system comprises: the PET/CT image acquisition and format conversion module is used for acquiring a PET image and a low-dose CT image paired with an elderly patient and performing format conversion treatment; the PET/CT preprocessing module is used for carrying out rigid transformation preprocessing on the PET image and the CT image after format conversion; the PET/CT image coarse standardization module is used for optimizing a nonlinear deformation field through nonlinear registration, applying the nonlinear deformation field to the rigidly transformed PET image and CT image and generating coarse standardized PET and CT images; The PET/CT image fine standardization module is used for carrying out tissue classification on a coarse registered CT image by adopting a mixed Gaussian model, using one Gaussian distribution on gray matter and white matter, using two Gaussian distributions on cerebrospinal fluid, bones, external tissues and air, aligning the two Gaussian distributions with a tissue probability map of a senile brain standard space, updating voxel classification probability based on space prior information and segmentation probability of the CT image, further generating a fine standardized CT and PET image registered to a senile brain MIITRA standard space by optimizing and regularizing iterative nonlinear deformation fields, and further generating a fine standardized CT and PET image registered to the senile brain MIITRA standard space, wherein the method comprises the following steps of: 1) Tissue classification is performed on the coarsely registered CT images by optimized segmentation parameters: Wherein, the The gaussian distribution is represented by the formula, And Respectively brain tissue class Specifically, the Gaussian parameters are chosen based on the basis that in CT imaging, only one average intensity value for each of gray and white matter, one Gaussian function is used, two Gaussian functions are used for cerebrospinal fluid because cerebrospinal fluid near the bone may have higher intensity due to the overflow effect, two Gaussian functions are used because the bone has a very wide range of values in CT, the external tissue is considered to consist of fat and muscle regions, two Gaussian functions are used, one Gaussian function is used to model most of-1024 HU in air, and the other Gaussian function is used to model pixels with other intensities between-50 and-300 HU; 2) Aligning the roughly registered CT image with a tissue probability map TPM of a senile brain standard space MIITRA, updating voxel classification probability, wherein the process combines space prior information And segmentation probability of CT image ; Wherein, the Is a category of Is a spatial prior probability of (1); 3) Optimizing nonlinear deformation fields by minimizing objective functions : Wherein, the Is a category in the template image Is used for the image of the probability of a (b), Is a regularization term used to constrain the smoothness of the deformation field, Is a regularization parameter; 4) Regularization term The regularization term is used for restraining the smoothness and continuity of the deformation field, and the regularization form is as follows: The regularization term ensures the continuity and smoothness of the deformation field by inhibiting the magnitude of the gradient of the deformation field; 5) Using an optimization algorithm, levenberg-Marquardt iteratively optimizes the deformation field: Wherein, the Is the rate of learning to be performed, Is the gradient of the objective function to the deformation field; 6) Applying the fine normalized nonlinear deformation field to the coarse registered CT and PET images to generate fine normalized CT and PET images: the automatic SUV extraction module is used for extracting SUV values of cortical and subcortical ROIs by adopting a MIITRA-space brain map for the fine standardized PET image for subsequent SUVR calculation and disease diagnosis analysis.
  2. 2. The CT-guided senile brain PET image space normalization system according to claim 1, wherein the PET/CT images are converted into nifi format after acquisition and the normalized uptake value SUV is calculated voxel by voxel for the PET data.
  3. 3. A CT-guided senile brain PET image space normalization system according to claim 1 in which the preprocessing procedure includes PET image resampling to match the voxel resolution of the CT image and rigid transformation of the CT image and the resampled PET image.
  4. 4. The aged brain PET image space standardization system based on CT guidance according to claim 1, wherein the PET/CT image coarse standardization module uses Old Normalize method to standardize the CT image after rigid transformation onto the CT brain template image, further carries out nonlinear registration, and obtains nonlinear deformation field through optimization and regularization.
  5. 5. The CT guidance-based senile brain PET image space normalization system of claim 1, wherein the automated SUV extraction module uses a MIITRA spatial brain atlas to segment the finely normalized PET image into a plurality of brain regions, and calculates an average SUV value for each brain region.
  6. 6. A senile brain PET image space normalization method based on a CT guidance-based senile brain PET image space normalization system according to any one of claims 1 to 5, characterized in that it comprises the following steps: (1) Acquiring PET images and CT images paired with the elderly patient, and performing format conversion treatment; (2) Carrying out rigid transformation pretreatment on the PET image and the CT image after format conversion; (3) Optimizing a nonlinear deformation field through nonlinear registration, and applying the nonlinear deformation field to the rigidly transformed PET image and CT image to generate a coarse standardized PET image and a coarse standardized CT image; (4) Tissue classification is carried out on the CT images subjected to coarse registration, the tissue classification probability is aligned with a tissue probability map of a brain standard space of the elderly, voxel classification probability is updated, and further an iterative nonlinear deformation field is optimized and applied to the CT images and the PET images subjected to coarse registration, so that fine standardized CT images and PET images are generated; (5) For fine standardized PET images, a MIITRA-space brain map is used to extract SUV values of cortical and subcortical ROIs for subsequent SUVR calculation and disease diagnosis analysis.
  7. 7. A CT guidance-based senile brain PET image space standardization apparatus comprising a memory and one or more processors, the memory storing executable code, characterized in that the processor implements a CT guidance-based senile brain PET image space standardization method as claimed in claim 6 when executing the executable code.
  8. 8. A computer readable storage medium having a program stored thereon, wherein the program, when executed by a processor, implements a CT guidance based senile brain PET image space normalization method according to claim 6.
  9. 9. A computer program product comprising a computer program which, when executed by a processor, implements a CT guidance based senile brain PET image space normalization method according to claim 6.

Description

CT-guidance-based senile brain PET image space standardization system, method and device Technical Field The invention relates to the field of PET image registration, in particular to a senile brain PET image space standardization system, method and device based on CT guidance. Background Positron emission computed tomography (Positron Emission Tomography, PET) is a leading edge molecular imaging technique that utilizes a radioisotope labeled tracer to visualize the distribution of specific molecules in vivo, quantitatively reflecting physiological and pathological processes. PET technology plays an important role in the scientific research and clinical diagnosis of brain diseases common to the elderly population, such as alzheimer's disease and parkinson's disease. However, current PET image diagnosis mainly relies on visual assessment by doctors, which is not only subjective but also prone to missed diagnosis and misdiagnosis. To overcome these limitations, automated extraction of standard uptake values (Standardized Uptake Value, SUV) or standard uptake value ratios (Standardized Uptake Value Ratio, SUVR) of brain regions becomes critical. The method is not only helpful for doctors to make objective diagnosis and reduce the dependence on subjective experience of doctors, but also can explore physiological and pathological changes of different brain areas in scientific researches. The key to achieving this goal is spatial normalization of the PET images, i.e., registering the individual images to a standard brain template space. At present, two main methods for spatial standardization of PET images are brain structure image-aided spatial standardization and PET brain template-based spatial standardization. Wherein the brain structure images include magnetic resonance imaging (Magnetic Resonance Imaging, MRI) structure images (e.g., T1-weighted image and T2-weighted image) and computed tomography (Computed Tomography, CT) structure images. The brain structure image aided method is generally divided into three steps, namely, firstly, registering an individual PET image to an individual structure brain image rigidly, secondly, registering the individual structure brain image to a standard brain structure template in a nonlinear manner, and finally, registering the PET image to a standard space by using a nonlinear deformation field generated in the last step. The PET brain template-based method comprises the steps of constructing PET standard brain templates of specific groups, and then non-linearly registering individual PET images to the templates. The precise brain anatomy information provided by high resolution MRI is considered to be the gold standard for PET standardization. Although the PET brain template-based method is computationally faster, it requires the construction of a specific population of PET standard brain templates and may produce large registration errors when processing PET images of great variability. On the other hand, CT brain structure image-aided PET normalization methods often require further optimization methods to approximate the effects of MRI due to the resolution limitations of CT images on brain anatomy. In particular, the method firstly carries out a cleaning procedure, sets all values lower than-300 HU to be-1024 HU so as to avoid the influence algorithm of a low-density structure outside the head, loads the preprocessed CT images into the optimized SPM12 segmentation algorithm, disables bias field correction on optimized parameters in SPM12 segmentation, only uses 1 Gaussian distribution on a Gaussian mixture model of grey matter and white matter, generates a spatially standardized deformation field by tissue segmentation of CT for other 4 tissues (cerebrospinal fluid, bone, external tissue and air) and carries out deformation field on the deformation field and CT, thereby realizing that the deformation field can be directly applied to the PET field of the aged patients in the case of the aging registration, but the problem of the aged can be solved, and the precision of the PET can be obviously improved when the PET field is not in the aged registration, because the PET field is directly used for the aged patients. In view of this, in the PET/CT scanning system, it is particularly urgent to develop an accurate and stable spatial standardization method for the PET image of the aged brain for the aged brain patients having only a low-dose CT structure map. The method utilizes the low-dose CT image, thereby not only avoiding the need of extra MRI scanning, but also improving the accuracy and stability of the standardization process, and having great significance for clinical practice and scientific research. The existing PET standardization techniques have several limitations that are particularly evident in clinical applications. First, standardized methods based on MRI structural brain images require the patient to receive additional high quality MRI scans in a short per