Nonrigid registration of 2-D and 3-D dynamic cell nuclei images for improved classification of subcellular particle motion

Abstract

The observed motion of subcellular particles in fluorescence microscopy image sequences of live cells is generally a superposition of the motion and deformation of the cell and the motion of the particles. Decoupling the two types of movements to enable accurate classification of the particle motion requires the application of registration algorithms. We have developed an intensity-based approach for nonrigid registration of multichannel microscopy image sequences of cell nuclei. First, based on 3-D synthetic images we demonstrate that cell nucleus deformations change the observed motion types of particles and that our approach allows to recover the original motion. Second, we have successfully applied our approach to register 2-D and 3-D real microscopy image sequences. A quantitative experimental comparison with previous approaches for nonrigid registration of cell microscopy has also been performed.

Fig. 1

Example images from a multichannel microscopy image sequence: (a) cell nucleus at time point 0, (b) cell nucleus at time point 80, (c) subcellular particles at time point 0, (d) subcellular particles at time point 80.

Fig. 2

Averaged mean squared displacement (MSD) of synthetic particle motion: (a) original data, (b) deforming cell (unregistered), (c) after registration using our nonrigid registration approach.

Fig. 3

(a) Histogram of relative shape anisotropy for directed, diffusive, and obstructed particle motion data, (b) histogram and cumulative histogram of the maximum distance of the particle position to the position at the first time point for 1000 particles with diffusive motion.

Fig. 4

(a) Original synthetic image at time point 0, (b) original image at time point 22, (c) registered image at time point 22 using our nonrigid registration approach.

Fig. 5

(a) Original image at time point 0, (b) original image at time point 70, (c) difference of the original images, (d) difference of the registered images, (e) registration transformation visualized as a vector field.

Fig. 6

(a) Root mean squared (RMS) intensity error and (b) correlation coefficient (CC) averaged over a real image sequence.

Fig. 7

Error of the position of a nucleus structure to the ground truth position without and with registration of a 2-D image sequence using the different nonrigid registration approaches: (a) without temporal regularization, (b) with temporal regularization using σ_T = 2.0.

Fig. 8

Error of the position of a nucleus structure to the ground truth position without and with registration of a 3-D image sequence using the different nonrigid registration approaches: (a) without temporal regularization, (b) with temporal regularization using σ_T = 2.0.