Related Experiment Video
Updated: Jun 26, 2026
07:44
Evaluation of Patients' Posture and Gait Profile After Lumbar Fusion Surgery by Video Rasterstereography and Treadmill Gait Analysis
Published on: March 23, 2019
Deep learning for predicting lumbar segmental instability using neutral lateral lumbar radiographs: a retrospective
Jiajun Song1, Jiawei Du1, Shengwei Liu1
1154 Anshan Road, Heping District, Tianjin, China, 300052, Tianjin Medical University General Hospital, Tianjin, China.
Summary
A deep learning model accurately predicts lumbar segmental instability (LSI) from X-rays, identifying key features like facet joints and disc degeneration. This tool aids in precise clinical screening and personalized treatment decisions.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Medicine
- Spinal Biomechanics
Background:
- Lumbar segmental instability (LSI) poses diagnostic challenges.
- Accurate prediction of LSI is crucial for effective patient management.
- Current diagnostic methods may lack precision and efficiency.
Purpose of the Study:
- To develop a deep learning model for LSI prediction using lateral lumbar radiographs.
- To identify key radiographic features contributing to LSI.
- To validate the model's performance and clinical utility.
Main Methods:
- A DenseNet121-based stacking ensemble model integrated with SVM, Random Forest, and Softmax classifiers was utilized.
- 10-fold cross-validation assessed model performance (AUC, accuracy, sensitivity, specificity, F1-score).
- Gradient-weighted Class Activation Mapping (Grad-CAM) identified critical anatomical regions, validated through machine learning frameworks.
Main Results:
- The stacking ensemble model achieved an AUC of 0.838, accuracy of 77.5%, sensitivity of 75.9%, and specificity of 78.4%.
- Grad-CAM highlighted facet joints (34.93%), intervertebral discs (25.99%), and osteophytes (22.66%) as primary contributors.
- Consistent performance was observed across age and gender subgroups, with decision curve analysis confirming clinical utility.
Conclusions:
- The developed stacking ensemble model effectively predicts LSI from neutral lateral lumbar radiographs.
- Key imaging biomarkers include facet joint hypertrophy, disc degeneration, and osteophyte formation.
- The model's generalizability and stable performance offer a reliable tool for clinical screening and decision-making.
