Osteoporosis
Conditions
Brief summary
In this study, we use conditional generation adversarial network to enhance the resolution of MSCT images and obtain micro-CT-like images. Based on this, we measure the bone structure indexes of micro-CT-like images and analyzed the correlation between bone structure and bio-mechanical indexes.
Detailed description
In this study, we expect to collect 10 sets of vertebrae (between T6 and L5) from 10 formalin-fixed human cadavers. The study protocol was reviewed and approved by the local Institutional Review Boards.The collected specimens subject to normalized Micro-CT and MSCT image scanning, image reconstruction using standard algorithms, and bone structure observation using bone algorithms to obtain high-quality, standardized axial images. We will use a conditional generation adversarial network-based image mapping technique to train the mapping model between MSCT images and Micro-CT images, find the correspondence between the image information contained in each of MSCT and Micro-CT, and finally obtain high-resolution micro-CT-like images. After obtaining the Micro-CT-like images of the samples, we will segment and annotate the images (including different structures such as vertebral body, bone cortex) and train the Teacher-Student & U-Net++ deep learning architecture to achieve accurate segmentation of vertebral body regions, and obtain cancellous bone regions of interest. Based on this, we will analyze the bone structure characteristics of each spatial region in the images, including the thickness, spacing, orientation and distribution pattern of bone trabeculae and other indicators. Ultimately, we will cut a standardized cubic sample from the vertebral cancellous bone , obtain the bio-mechanical performance index of this cancellous bone sample by mechanical experiments, and analyze the correlation between bone structure of Micro-CT-like images and bio-mechanical index.
Interventions
DIAGNOSTIC_TESTRadiological scanning
The collected specimens subject to normalized Micro-CT and MSCT image acquisition, image reconstruction using standard algorithms, and bone structure observation using bone algorithms to obtain high-quality, standardized axial images.
OTHERDeep learning
We will use a conditional generation adversarial network-based image mapping technique to train the mapping model between MSCT images and Micro-CT images, find the correspondence between the image information contained in each of MSCT and Micro-CT, and finally obtain high-resolution micro-CT-like images.
PROCEDUREBio-mechanical testing
we will cut a standardized cubic sample from the vertebral cancellous bone , obtain the bio-mechanical performance index of this cancellous bone sample by mechanical experiments, and analyze the correlation between bone structure of Micro-CT-like images and bio-mechanical index.
Sponsors
Peking University Third Hospital
Study design
Observational model
CASE_CROSSOVER
Time perspective
CROSS_SECTIONAL
Eligibility
Sex/Gender
ALL
Healthy volunteers
No
Inclusion criteria
* We will collect human formalin-fixed vertebral specimens provided by the Department of Anatomy, Peking University School of Medicine. The expected number of enrolled specimens is approximately 100 vertebrae (between T6 and L5) .
Exclusion criteria
* The vertebrae with significant compression fractures, bone neoplasms, or other causes of significant bone destruction.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Consistency or correlation | From the beginning of the experiment to 8 months | Analyze the consistency or correlation between micro-CT and micro-CT-like bone structure indexes; analyze the correlation between micro-CT-like bone structure indexes and bio-mechanical indexes |
Countries
China
Contacts
Primary ContactDan Jin
Outcome results
None listed