A Gaussian Mixture Model for Automated Vesicle Fusion Detection and Classification

Author(s): Haohan Li, Zhaozheng Yin, and Yingke Xu
Date of Publication: October 2015
Abstract:

Accurately detecting and classifying vesicle-plasma membrane fusion events in fluorescence microscopy, is of primary interest for studying biological activities in a close proximity to the plasma membrane. In this paper, we present a novel Gaussian mixture model for automated identification of glucose transporter 4 (GLUT4) vesicle and plasma membrane fusions and classification of full fusion and partial fusion events in Total Internal Reflection Fluorescence microscopy (TIRFM) image sequences. Image patches of fusion event candidates are detected in individual images and linked over consecutive frames. A Gaussian mixture model is fit on each image patch of the patch sequence with outliers rejected for robust Gaussian fitting. The estimated parameters of Gaussian functions over time are catenated into feature vectors for classifier training. Applied on three challenging datasets, our method achieved competitive results on detecting and classifying fusion events compared with two other state-of-the-art methods.

Citation: Haohan Li, Zhaozheng Yin and Yingke Xu, “A Gaussian Mixture Model for Automated Vesicle Fusion Detection and Classification,” the 18th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) workshop on Computational Methods for Molecular Imaging (CMMI), 2015.
Team(s): Plant Team