Contour Analysis: Cell Membrane Edge Detection
Dr. Young-Zoon Yoon, Aidan Brown, and I developed an algorithm that efficiently detects the contour of cell membranes from phase contrast microscopy images using correlation and quadratic fitting under Dr. Pietro Cicuta at the Cavendish Lab of the University of Cambridge from July to August 2008. The use of correlation was first suggested by Aidan Brown, and I further elaborated its efficiency by introducing the normalization and shift factors for input profiles. Moreover, I also introduced the quadratic fitting in the final stage in order to obtain sub-pixel level results and other devices (e.g., sub-pixel resampling, etc.) for the minimization of unwanted aliasing effect. The algorithm sweeps every frame of the captured movie and automatically delimits the contour of the membrane for each frame, thereby providing the complete positional information of the cell membranes as time evolves. I implemented the algorithm as a multi-threaded C++ module for MATLAB (supported platform: BSD-UNIX, Mac OS X, Linux, Windows; Fig. 1). Currently, researchers at the Cavendish Lab utilize the algorithm to assess the thermal fluctuations of human red blood cells.
Figure 1: the implemented algorithm on MATLAB.
Figure 2: the detected cell membrane is depicted as a red contour. The algorithm produces sub-pixel level results.
