Opportunistic detection and risk assessment of vertebral fractures in high-risk populations using machine learning

Project Description

Training a machine learning model that can identify vertebral fractures in clinical CT scans…

Training a machine learning model that can identify vertebral fractures in clinical CT scans. This model could either identify the presence of fractures OR identify the fractures and give the fracture a grade based on Genant classification. Once the ML model is developed, it could be used on RETRO2 to create associations with fracture risk and individuals with various spinal conditions (Schmorl’s Nodes, Ankylosing Spondylitis, Diffuse Idiopathic Skeletal Hyperostosis, etc.). Or could be integrated into the entire opportunistic CT pipeline, since clinicians would be interested in finding ways to identify fractures opportunistically.

Scope

• MSc: 2-year master project
• Summer student: extracted sub-projects
• Includes grading, model development, identifying spine diseases in scans, running inference on dataset, statistics

Data Source

• RETRO
• RETRO2
• Any other opportunistic CT clinical cohort

Resources

• ML computer (GROOT!)