Authors (including presenting author) :
WL Yeung
Affiliation :
Innovation and technology laboratory
Introduction :
Osteoporosis, characterized by low bone mass and increased fracture risk, often presents with vertebral collapse leading to significant morbidity. Early detection is crucial for timely intervention. At Princess Margaret Hospital, we evaluated an AI-driven screening system on lumbar spine (LS) radiographs to detect vertebral fractures and predict osteoporosis (T-score < –1.0).
Objectives :
1. Assess the AI program’s ability to identify vertebral fractures (osteoporotic collapse). 2. Determine the program’s performance in predicting osteoporosis (T-score < –2.0).
Methodology :
• Study Design & Population: Retrospective review of 3,000 LS spine radiographs from adults (≥ 50 years). • Data Review: Radiographs were independently assessed by an orthopaedic and traumatology team for vertebral fractures and signs of osteoporosis. • AI Model Development: A labeled dataset (normal vertebrae, osteoporotic changes, vertebral fractures) was used to train the model. • Performance Metrics: • Area Under the Curve (AUC) for detecting T-score < –1.0. • Accuracy in identifying vertebral collapse.
Result & Outcome :
Results • Osteoporosis Detection: The AI achieved an AUC of 0.91 for predicting T-score < –1.0. • Vertebral Collapse Detection: The model attained 85% accuracy for detecting osteoporotic vertebral fractures. • Clinical Correlation: AI findings were consistent with specialist readings, reducing oversight and improving efficiency in clinic. Outcome and Conclusion An AI-based approach demonstrated high accuracy in detecting osteoporotic vertebral fractures and predicting low T-scores. This promising tool can streamline diagnosis, minimize human error, and improve patient care through earlier intervention and treatment. Future work will focus on refining the model’s performance and validating these results in larger data set.
