[Artificial intelligence for diagnosis of vertebral compression fractures using a morphometric analysis model, based
A V Petraikin1, Zh E Belaya2, A N Kiseleva3
1Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies.
Problemy Endokrinologii
|December 28, 2020
Summary
An AI model, Comprise-G, effectively detects vertebral compression fractures (VFs) on chest CT scans, improving osteoporosis complication diagnosis. This AI assistant shows high accuracy, aiding radiologists in preventing future fractures.
Area of Science:
- Radiology
- Artificial Intelligence
- Medical Imaging
Background:
- Pathological low-energy vertebral compression fractures (VFs) are common osteoporosis complications.
- VFs are often unreported on chest CT (CCT), necessitating advanced diagnostic tools.
- An AI assistant can improve the detection of osteoporosis complications and prevent new fractures.
Purpose of the Study:
- To develop an AI model for automated diagnosis of thoracic spine compression fractures.
- The model aims to assist radiologists in identifying VFs on CCT images.
Main Methods:
- A retrospective study used 160 anonymized CCT scans.
- Morphometric analysis and semiquantitative assessment were performed.
- A CNN-based AI model (Comprise-G) was developed to measure vertebral bodies and calculate compression degree.
Main Results:
- The Comprise-G model achieved high sensitivity and specificity in detecting VFs.
- On test data, the model showed 83.2% sensitivity and 90.0% specificity for VFs in patients.
- For individual vertebrae, the model demonstrated 79.3% sensitivity and 98.7% specificity on test data.
Conclusions:
- The Comprise-G AI model exhibits strong diagnostic performance for VFs on CCT.
- The model shows potential for clinical application and warrants further validation.
