Uncertainty-aware asynchronous scattered motion interpolation using Gaussian process regression.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Anatomy

Background:

• Soft tissue motion estimation is crucial for image-guided interventions and surgical navigation.
• Accurate registration of pre-operative data onto deformable organs requires precise motion tracking.
• Existing methods struggle with randomly scattered, non-uniform, and uncertain motion measurements.

Purpose of the Study:

• To develop a novel method for interpolating uncertain, spatiotemporally scattered motion measurements.
• To enhance the accuracy and reliability of soft tissue motion estimation.
• To enable uncertainty quantification in surgical navigation information.

Main Methods:

• Proposed a novel motion field representation for scattered motion signal samples.
• Devised a new Gaussian process (GP) regression model with a non-constant-mean prior and anisotropic covariance function.
• Performed extensive evaluations comparing the proposed GP model against state-of-the-art methods.

Main Results:

• The novel GP regression model significantly outperforms existing GP models for soft tissue motion interpolation.
• The proposed method effectively handles randomly non-uniformly scattered and uncertain motion data.
• The model provides uncertainty quantification for the interpolated motion field.

Conclusions:

• The developed Gaussian process regression model offers a superior approach for soft tissue motion estimation.
• This advancement improves the reliability of navigation information in image-guided surgery.
• Uncertainty quantification aids surgeons in making informed decisions during interventions.