Automatic segmentation of the thumb trapeziometacarpal joint using parametric statistical shape modelling and random...

Area of Science:

• Medical imaging
• Biomedical engineering
• Computational anatomy

Background:

• The trapeziometacarpal (TMC) joint is crucial for hand function, and its accurate modeling is essential for clinical analysis.
• Current methods for creating 3D models of the TMC joint from CT images can be time-consuming and labor-intensive.
• Parametric meshes are valuable for statistical shape modeling and quantitative analysis of joint morphology.

Purpose of the Study:

• To develop and validate an automated pipeline for generating parametric meshes of the TMC joint bones from clinical CT images.
• To enable efficient batch processing and analysis of TMC joint geometry.
• To assess the accuracy and efficiency of the proposed automated segmentation method.

Main Methods:

• The pipeline utilizes 3D random forest regression voting (RFRV) combined with statistical shape model (SSM) segmentation.
• The method was trained and validated on a dataset of 65 clinical CT images.
• A subset of images was held out from the training set to rigorously test the segmentation performance.

Main Results:

• The automated pipeline achieved mean root mean squared (RMS) errors of 1.066 mm for the first metacarpal and 0.632 mm for the trapezial bone.
• The segmentation process required approximately 2 minutes per CT image, demonstrating high efficiency.
• Preliminary results indicate promising accuracy for generating 3D meshes of TMC joint bones.

Conclusions:

• The proposed automated pipeline provides an accurate and efficient method for creating parametric meshes of the TMC joint from CT images.
• This approach facilitates large-scale batch processing and analysis of TMC joint morphology.
• The developed technique holds potential for advancing research and clinical applications related to TMC joint pathologies.