Automatic Vertebral Rotation Angle Measurement of 3D Vertebrae Based on an Improved Transformer Network
The measurement of vertebral rotation angles serves as a crucial parameter in spinal assessments, particularly in understanding conditions such as idiopathic scoliosis. Historically, these angles were calculated from 2D CT images. However, such 2D techniques fail to comprehensively capture the intri...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-01-01
|
| Series: | Entropy |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1099-4300/26/2/97 |
| _version_ | 1827343667596099584 |
|---|---|
| author | Xing Huo Hao Li Kun Shao |
| author_facet | Xing Huo Hao Li Kun Shao |
| author_sort | Xing Huo |
| collection | DOAJ |
| description | The measurement of vertebral rotation angles serves as a crucial parameter in spinal assessments, particularly in understanding conditions such as idiopathic scoliosis. Historically, these angles were calculated from 2D CT images. However, such 2D techniques fail to comprehensively capture the intricate three-dimensional deformities inherent in spinal curvatures. To overcome the limitations of manual measurements and 2D imaging, we introduce an entirely automated approach for quantifying vertebral rotation angles using a three-dimensional vertebral model. Our method involves refining a point cloud segmentation network based on a transformer architecture. This enhanced network segments the three-dimensional vertebral point cloud, allowing for accurate measurement of vertebral rotation angles. In contrast to conventional network methodologies, our approach exhibits notable improvements in segmenting vertebral datasets. To validate our approach, we compare our automated measurements with angles derived from prevalent manual labeling techniques. The analysis, conducted through Bland–Altman plots and the corresponding intraclass correlation coefficient results, indicates significant agreement between our automated measurement method and manual measurements. The observed high intraclass correlation coefficients (ranging from 0.980 to 0.993) further underscore the reliability of our automated measurement process. Consequently, our proposed method demonstrates substantial potential for clinical applications, showcasing its capacity to provide accurate and efficient vertebral rotation angle measurements. |
| first_indexed | 2024-03-07T22:33:38Z |
| format | Article |
| id | doaj.art-74c9f1d1ca52493e99616e28e013ae25 |
| institution | Directory Open Access Journal |
| issn | 1099-4300 |
| language | English |
| last_indexed | 2024-03-07T22:33:38Z |
| publishDate | 2024-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Entropy |
| spelling | doaj.art-74c9f1d1ca52493e99616e28e013ae252024-02-23T15:15:35ZengMDPI AGEntropy1099-43002024-01-012629710.3390/e26020097Automatic Vertebral Rotation Angle Measurement of 3D Vertebrae Based on an Improved Transformer NetworkXing Huo0Hao Li1Kun Shao2School of Mathematics, Hefei University of Technology, Hefei 230601, ChinaSchool of Mathematics, Hefei University of Technology, Hefei 230601, ChinaSchool of Software, Hefei University of Technology, Hefei 230601, ChinaThe measurement of vertebral rotation angles serves as a crucial parameter in spinal assessments, particularly in understanding conditions such as idiopathic scoliosis. Historically, these angles were calculated from 2D CT images. However, such 2D techniques fail to comprehensively capture the intricate three-dimensional deformities inherent in spinal curvatures. To overcome the limitations of manual measurements and 2D imaging, we introduce an entirely automated approach for quantifying vertebral rotation angles using a three-dimensional vertebral model. Our method involves refining a point cloud segmentation network based on a transformer architecture. This enhanced network segments the three-dimensional vertebral point cloud, allowing for accurate measurement of vertebral rotation angles. In contrast to conventional network methodologies, our approach exhibits notable improvements in segmenting vertebral datasets. To validate our approach, we compare our automated measurements with angles derived from prevalent manual labeling techniques. The analysis, conducted through Bland–Altman plots and the corresponding intraclass correlation coefficient results, indicates significant agreement between our automated measurement method and manual measurements. The observed high intraclass correlation coefficients (ranging from 0.980 to 0.993) further underscore the reliability of our automated measurement process. Consequently, our proposed method demonstrates substantial potential for clinical applications, showcasing its capacity to provide accurate and efficient vertebral rotation angle measurements.https://www.mdpi.com/1099-4300/26/2/97predictive modelsdeep learningattention workspoint cloudautomatic measurement |
| spellingShingle | Xing Huo Hao Li Kun Shao Automatic Vertebral Rotation Angle Measurement of 3D Vertebrae Based on an Improved Transformer Network Entropy predictive models deep learning attention works point cloud automatic measurement |
| title | Automatic Vertebral Rotation Angle Measurement of 3D Vertebrae Based on an Improved Transformer Network |
| title_full | Automatic Vertebral Rotation Angle Measurement of 3D Vertebrae Based on an Improved Transformer Network |
| title_fullStr | Automatic Vertebral Rotation Angle Measurement of 3D Vertebrae Based on an Improved Transformer Network |
| title_full_unstemmed | Automatic Vertebral Rotation Angle Measurement of 3D Vertebrae Based on an Improved Transformer Network |
| title_short | Automatic Vertebral Rotation Angle Measurement of 3D Vertebrae Based on an Improved Transformer Network |
| title_sort | automatic vertebral rotation angle measurement of 3d vertebrae based on an improved transformer network |
| topic | predictive models deep learning attention works point cloud automatic measurement |
| url | https://www.mdpi.com/1099-4300/26/2/97 |
| work_keys_str_mv | AT xinghuo automaticvertebralrotationanglemeasurementof3dvertebraebasedonanimprovedtransformernetwork AT haoli automaticvertebralrotationanglemeasurementof3dvertebraebasedonanimprovedtransformernetwork AT kunshao automaticvertebralrotationanglemeasurementof3dvertebraebasedonanimprovedtransformernetwork |