Artificial intelligence data-driven 3D model for AIS.

Area of Science:

• Biomedical Engineering
• Computational Biology
• Orthopedics

Background:

• Scoliosis is a complex 3D spinal deformity, with Adolescent Idiopathic Scoliosis (AIS) being the most prevalent form in children.
• Current AIS treatment selection relies heavily on surgeon experience, highlighting a need for objective, predictive tools.
• Patient-specific models can enhance understanding and treatment of AIS during critical growth periods.

Purpose of the Study:

• To develop a clinically validated, patient-specific Reduced Order Finite Element Model (ROFEM) for spine analysis.
• To predict the progression of Adolescent Idiopathic Scoliosis (AIS) using data mining techniques.
• To propose an improved, data-driven method for AIS treatment selection.

Main Methods:

• Integrating Finite Element (FE) analysis with biomechanical data, image processing, and data science.
• Generating patient-specific spine geometry from X-ray images for FE modeling.
• Developing a RO model from detailed FE models and employing neural networks for curvature prediction.

Main Results:

• A patient-specific ROFEM of the spine was developed, integrating biomechanical and imaging data.
• A neural network approach was implemented to predict spinal curvature progression in AIS.
• The study established a foundation for predicting AIS severity and guiding treatment decisions.

Conclusions:

• Clinically validated patient-specific spine models can significantly improve the understanding and management of AIS.
• Predictive modeling empowers physicians to make more informed treatment choices and educate families effectively.
• This approach promises to enhance surgical intervention assessment and personalized care for AIS patients.