Predicting Secondary Vertebral Compression Fracture After Vertebral Augmentation via CT-Based Machine Learning Radiomics-Clinical Model

Xiaokun Wang ¹, Wu Ye ¹, Yao Gu ¹

⁰Department of Orthopedics, First Affiliated Hospital of Nanjing Medical University, Nanjing 210029, China (X.W., W.Y., Y.G., Y.G., H.W., Y.Z., D.P., X.G., W.C.).

Academic Radiology +

|

July 11, 2024

View abstract on PubMed

Summary

A new radiomics-clinical model can predict secondary vertebral compression fractures (SVCF) after vertebral augmentation (VA). This tool helps personalize treatment strategies for patients undergoing VA surgery.

Area Of Science

Radiology
Medical Imaging
Machine Learning in Medicine

Background

Secondary vertebral compression fractures (SVCF) are a common complication following vertebral augmentation (VA).
Predicting SVCF is crucial for optimizing patient outcomes and treatment planning.
Current methods for SVCF prediction require enhancement for improved accuracy.

Purpose Of The Study

To develop a radiomics-based predictive model for SVCF after VA.
To integrate radiomics features with clinical data for a comprehensive predictive tool.
To establish a model that can guide precise treatment strategies for patients at risk of SVCF.

Main Methods

Retrospective analysis of 470 patients undergoing VA for osteoporotic vertebral compression fractures (OVCF).
Extraction of radiologic features from CT images (T6-L5 vertebrae) using radiomics.
Construction of a predictive model using machine learning algorithms (XGBoost) and clinical variables.

Main Results

A logistic nomogram incorporating radiomics signature, bone cement volume, and L1-L4 T-scores demonstrated high predictive capability.
The model achieved a prediction capability of 0.986 in the training set and 0.884 in the verification set.
Eight radiomics features were identified as significant predictors of SVCF.

Conclusions

The developed radiomics-clinical model shows significant potential for prospectively predicting SVCF post-VA.
This model can aid in tailoring treatment strategies for individual patients.
Machine learning-driven radiomics analysis offers a promising approach for managing complications after vertebral augmentation.

Keywords:

Machine learning Predictive model Radiomics Secondary vertebral compression fracture Vertebral augmentation