Case study: cellar scaffold extraction using crest point for volume rendering

Abstract

Extraction of scaffolds, such as the meiotic spindles, a 3D tubular framework consisting of the microtubules, from conforcal laser scanning microscopy (CLSM) data of a cell is a challenge in biological image processing. It is of major importance in the research of microtubule anchor proteins, and molecular motor mechanics. However, the scaffold is hidden within CLSM data due to the nature of light excitation, and is difficult to visualize using traditional opacity and color transfer functions that depend only on local intensity. In this paper, we treat 3D CLSM data as a hyper-surface in R4, and show that the crest points of the hyper-surface correspond to the centerline of the hidden scaffold. We propose an automatic approach to extract the hidden scaffold from CLSM data. First, the spindle from the large data set is segmented using Weibull E-SD fields. We, next, apply the Savitzky-Golay (S-G) filter and Gaussian convolution to reduce the noise in the data and calculate the first and second derivatives. Lastly, direct volume rendering using ray casting is applied to visualize the volume data. We combine the local intensity and maximum curvature information to decide the opacity transfer function. Promising results are shown on simulated data sets as well as real CLSM data of mouse egg.