Deep learning-based multimodal image analysis predicts bone cement leakage during percutaneous kyphoplasty: protocol...

Area of Science:

• Orthopedics
• Radiology
• Artificial Intelligence

Background:

• Bone cement leakage (BCL) is a common complication of percutaneous kyphoplasty (PKP) for osteoporotic vertebral compression fractures (OVCF).
• BCL can lead to severe secondary complications and negatively impact patient outcomes.
• Existing machine learning models have limitations in clinical application for predicting BCL.

Purpose of the Study:

• To develop and validate a deep learning (DL) model for predicting BCL occurrence and classification during PKP.
• To analyze preoperative CT and MRI scans for BCL prediction.
• To evaluate the model's performance against clinician predictions and reference standards.

Main Methods:

• Retrospective and prospective internal datasets for training and validation.
• Cross-center external dataset for model testing.
• Deep learning analysis of preoperative CT and MRI data.

Main Results:

• (Results will be detailed upon study completion)

Conclusions:

• The DL model is expected to significantly improve preoperative risk assessment for BCL.
• This tool can enable personalized treatment strategies, reducing BCL incidence and enhancing clinical outcomes.
• The model has potential for broader application, especially in resource-limited settings, and can facilitate better clinician-patient communication.