A PDE-regularized smoothing method for space-time data over manifolds with application to medical data.

Area of Science:

• Statistics
• Numerical Analysis
• Differential Geometry

Background:

• Modeling spatio-temporal data on manifolds presents significant challenges.
• Existing methods may struggle with computational efficiency and large datasets.

Purpose of the Study:

• To introduce an innovative statistical-numerical method for spatio-temporal data analysis on generic 2D Riemannian manifolds.
• To develop an efficient and scalable approach for handling massive datasets in complex domains.

Main Methods:

• A regression model incorporating a heat equation-based regularizing term.
• Discretization using a finite element scheme on the manifold.
• Solution via a fixed-point iterative algorithm for enhanced efficiency.

Main Results:

• The proposed method demonstrates high efficiency compared to monolithic approaches.
• The technique effectively handles massive spatio-temporal datasets.
• Successful application shown in simulation studies, neuroimaging, and hemodynamic data analysis.

Conclusions:

• The novel statistical-numerical method provides an efficient and scalable solution for modeling spatio-temporal data on manifolds.
• The approach is well-suited for complex, large-scale applications in fields like neuroimaging and hemodynamics.