Modelling individual geometric variation based on dominant eigenmodes of organ deformation: implementation and...

Area of Science:

• Medical Imaging
• Computational Anatomy
• Radiotherapy Physics

Background:

• Inter-fractional organ motion and deformation significantly impact radiotherapy accuracy.
• Accurate modeling of organ variability is crucial for adaptive radiotherapy planning.
• High-dimensional geometric data from CT scans presents a challenge for motion modeling.

Purpose of the Study:

• To develop and validate a statistical method for modeling inter-fractional organ deformation and correlated motion.
• To reduce the dimensionality of organ geometry information using principal component analysis (PCA).
• To characterize patient-specific geometric variability using eigenmodes for improved radiotherapy planning.

Main Methods:

• Application of principal component analysis (PCA) to organ shapes from multiple CT studies.
• Development of a few-parametric statistical model based on eigenmodes representing correlated displacements.
• Quantification of geometric variability using eigenvalues and weighted sums of eigenmodes.
• Validation on prostate, rectum, and bladder datasets from four patients (15-18 CTs each).

Main Results:

• The eigenmode decomposition effectively reduced the dimensionality of geometric data.
• Geometric variability of prostate, bladder, and rectum was found to be governed by a few dominant patient-specific eigenmodes.
• Residual errors converged rapidly with an increasing number of eigenmodes.
• Using four dominant modes, residual errors were within 1.3-2.0 mm for prostate, 1.4-1.9 mm for rectum, and 1.5-1.9 mm for bladder.

Conclusions:

• A few dominant eigenmodes accurately describe individual geometric variations in organs from multiple imaging data.
• This eigenmode-based approach effectively characterizes deformable organ motion.
• The method provides key factors to assist in adaptive radiotherapy planning.