Artificial intelligence is showing promise in identifying vertebral fractures on CT scans, achieving accuracy comparable to seasoned medical specialists. This development could significantly aid in the diagnosis of osteoporosis, a condition weakening bones and increasing fracture risk.
Vertebral fractures, often missed on scans, are critical indicators of underlying osteoporosis. A recent study compared eight human raters with four deep learning models and one commercial AI tool. Using a large dataset of thoracic and lumbar vertebrae, the AI models demonstrated diagnostic accuracy for moderate to severe fractures at 0.988, statistically on par with radiology residents at 0.991.
These findings suggest AI-assisted CT analysis could streamline the detection and grading of vertebral fractures. While further clinical validation is necessary, these deep learning tools are moving closer to integration into routine imaging workflows, offering a valuable support system for clinicians to identify fractures that might otherwise go unnoticed.