Universality in the merging dynamics of parametric active contours: a study in MRI-based lung segmentation

TL;DR

This paper reveals that the dynamics of parametric active contours in MRI lung segmentation exhibit universal scaling behavior akin to classical nucleation, providing a fundamental understanding of the method's underlying physics.

Contribution

It demonstrates the universal scaling behavior in active contour dynamics, linking it to classical nucleation theory, which is a novel insight in medical image segmentation.

Findings

Active contour dynamics follow a nucleation-like universal scaling law.

The numerical results align with theoretical nucleation model exponents.

This understanding enhances the fundamental comprehension of segmentation processes.

Abstract

Measurement of lung ventilation is one of the most reliable techniques of diagnosing pulmonary diseases. The time consuming and bias prone traditional methods using hyperpolarized H${}^{3}$He and ${}^{1}$H magnetic resonance imageries have recently been improved by an automated technique based on multiple active contour evolution. Mapping results from an equivalent thermodynamic model, here we analyse the fundamental dynamics orchestrating the active contour (AC) method. We show that the numerical method is inherently connected to the universal scaling behavior of a classical nucleation-like dynamics. The favorable comparison of the exponent values with the theoretical model render further credentials to our claim.