X-Metric: An N-Dimensional Information-Theoretic Framework for Groupwise Registration and Deep Combined Computing.

Area of Science:

• Medical Imaging Analysis
• Computational Anatomy
• Machine Learning for Healthcare

Background:

• Accurate statistical dependency estimation and anatomical correspondence finding are crucial for analyzing multiple medical images.
• Existing methods often struggle with computational complexity and scalability for large datasets.

Purpose of the Study:

• To present a generic probabilistic framework for groupwise registration and anatomical correspondence finding in N medical images.
• To develop a computationally efficient and versatile method applicable to various medical imaging scenarios.

Main Methods:

• A novel N-dimensional joint intensity distribution formulation using latent variables and nonparametric estimators.
• Induction of an information-theoretic X-metric and X-CoReg algorithm via maximum likelihood and expectation-maximization.
• Extension to a weakly-supervised scenario with a deep learning-based combined-computing framework for simultaneous registration and segmentation.

Main Results:

• The proposed X-CoReg algorithm achieves groupwise registration with O(N) computational complexity.
• Demonstrated versatility in multimodal groupwise registration, motion correction for dynamic contrast enhanced MRI, and deep combined computing.
• Achieved superior accuracy and efficiency compared to existing methods in diverse applications.

Conclusions:

• The developed probabilistic framework offers a powerful and efficient solution for groupwise medical image analysis.
• The combined-computing deep learning framework enables simultaneous, collaborative registration and segmentation.
• The method's adaptability and performance highlight the advantages of the proposed imaging process representation.