A Fully Automated Pipeline for Vertebral Structural Assessment From Medical Images. Application Under Metastatic...

Area of Science:

• Biomedical Engineering
• Computational Mechanics
• Medical Imaging Analysis

Background:

• Spinal bone metastases frequently cause vertebral fractures, significantly impacting patient quality of life.
• Current methods for predicting skeletal events have limited accuracy, necessitating advanced computational approaches.
• Accurate prediction of vertebral structural failure is crucial for effective clinical management and complication prevention.

Purpose of the Study:

• To develop a fully automated, patient-specific computational framework for predicting vertebral structural behavior using CT data.
• To integrate deep learning, geometric normalization, and finite element analysis for enhanced fracture risk assessment.
• To support clinical decision-making and treatment planning for patients with metastatic spinal disease.

Main Methods:

• A deep neural network was employed for semantic segmentation of vertebrae and metastatic regions from CT scans.
• The Coherent Point Drift (CPD) algorithm was utilized for automated alignment and boundary condition definition.
• Cartesian Grid Finite Element Method (cgFEM) simulations were performed to assess vertebral mechanical response under metastatic conditions.

Main Results:

• The developed workflow achieved full automation from CT imaging to fracture risk estimation with high segmentation accuracy.
• cgFEM simulations provided clinically relevant metrics, including safety factors and stability variations, linked to tumor characteristics.
• The analysis successfully identified scenarios associated with an increased risk of vertebral fracture.

Conclusions:

• This study presents an end-to-end, patient-specific framework for automated fracture risk evaluation in metastatic vertebrae.
• The combination of deep learning and computational mechanics offers clinically relevant outputs to guide therapeutic strategies.
• Future work will focus on integrating patient-specific loading data for enhanced predictive modeling and clinical decision support.