An articulated statistical shape model for accurate hip joint segmentation.

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Accurate segmentation of pelvic and proximal femur anatomy in CT data is crucial for various clinical applications.
• Existing statistical shape models often lack flexibility in representing complex anatomical joints like the hip.

Purpose of the Study:

• To develop a fully automatic, robust, and accurate framework for segmenting the human pelvis and proximal femur in CT images.
• To introduce a composite statistical shape model with an explicitly modeled flexible hip joint.
• To evaluate a graph optimization-based multi-object segmentation method for improved accuracy.

Main Methods:

• Development of a composite statistical shape model for the femur and pelvis, incorporating explicit rotational parameters for hip joint flexibility.
• Image-driven adaptation of the composite shape model.
• Evaluation of a fine-grain multi-object segmentation method using graph optimization, leveraging initializations from the shape model.

Main Results:

• The proposed composite shape model adaptation demonstrated robust and accurate segmentation of both the proximal femur and pelvis in a leave-one-out evaluation on 50 CT volumes.
• Simultaneous optimization of femur and pelvis using graph-based methods yielded more accurate results compared to separate optimizations.
• The framework integrates shape model adaptation and graph optimization for fully automatic segmentation.

Conclusions:

• The developed framework provides a robust and accurate solution for automatic segmentation of the pelvis and proximal femur in CT data.
• Explicitly modeling hip joint flexibility within a composite statistical shape model enhances segmentation performance.
• The combined approach of shape model adaptation and graph optimization offers a significant advancement in medical image segmentation.