Compression Fracture
Conditions
Brief summary
This project aims to develop and validate an automatic detection and classification system for vertebral compression fractures on computer tomography (CT) images using an artificial intelligence (AI) system (named Smart Bone) by Quanta.
Detailed description
A computer search of CT scans (2010.01.01-2018.09.30) was performed in Wan Fang Hospital. Those CT images that were retrospectively reviewed by experienced radiologists. The CT scans of 1000-1500 subjects aged 50 and above with and without thoracic or lumbar compression fractures were included in this project for machine learning and deep learning. The control group included those without compression fractures while the patient group were those with compression fractures. Subjects that did not meet the inclusion criteria were excluded. The cortical layer of the T12-L5 spine images were manually labelled with the labeling software by the the technologists and confirmed the correctness of the image by an experienced radiologist. All the de-linked and completed images were provided to Quanta Computer Inc. for subsequent classification and analysis of AI machines for deep learning to facilitate the development of a system for automatic detection of pressure fractures by CT. This newly developed automatic system will be of valuable clinical impact in assisting radiologists to detect and classify vertebral compression fractures precisely and accurately.
Interventions
OTHERComputer Assisted Detection Software For Vertebral Fractures
A device named Smart Bone that is using CT image retrospectively acquired to entails a second review of CT images with Vertebral Compression Fractures through an interactive AI program developed by Quanta Computer Inc. was applied to the CT images. The device is a computer-aided detection (CADe) software application and is designed to assist radiologists to analyze Spine CT images. The device uses deep learning methods to perform vertebrae detection and classification of images.
Sponsors
Quanta Computer Inc.
Taipei Medical University WanFang Hospital
Study design
Observational model
COHORT
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
50 Years to 90 Years
Healthy volunteers
Yes
Inclusion criteria
* Cases with CT examinations acquired between 2010.01-2018.09 * Cases with CT examinations performed with one of the following protocol: whole body, abdomen, and spine * Cases must be \>/= 50 years of age * Cases with reports from CT examinations ditched as positive or negative compression fractures within a search range from T12 to L5 vertebrae. * CT images with raw data that are allowed to be reconstructed in axial view with a slice thickness of 1.3 mm * CT images with raw data that are allowed to be reconstructed in sagittal view with a slice thickness of 2.5 mm
Exclusion criteria
* CT images with imaging artifacts, foreign bodies, or implants * Cases with comorbid conditions, such as infection, cancer metastasis, chronic osteomyelitis, or other nonosteoporotic compression fracture
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Concordance rate | 2019.06 to 2020.03 | CT Imaging Reporting and Data System descriptors suggested by Smart Bone are in good agreement with those selected by experts. In other words, the CT Imaging Reporting and Data System generated by Smart Bone are not statistically different from the consensus of experts. CT Imaging Reporting and Data System Assessment Category Score: The user makes the final decision on the Assessment Category Score. Using this Score, Smart Bone displays the assessment description. Grade 0: Normal Vertebrae Grade 1: Mild Fracture, 20-25% Grade 2: Moderate Fracture, 26-40% Grade 3: Severe Fracture, \>40% |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Accuracy | 2019.06 to 2020.03 | Comparing to the accuracy of CT imaging results by radiologists with and without CADe will be evaluated. |
Countries
Taiwan
Outcome results
None listed