"
Team:ETHZ Basel/InformationProcessing/CellDetection
From 2010.igem.org
Cell Detection and Tracking
Background
One important input of the controller is the current location and the swimming direction of the E. lemming that we decide to control. However, due to experimental limitations, also other E. coli cells will by chance be on the microscope image. We therefore decided to implement a cell detection algorithm which detects all cells in the current frame and combine it with a cell tracking algorithm that aligns the cells in the adjacent frames. Thus it is possible to keep track of the E. lemming without confuse it with other cells. Since the controller has to react in real-time on changes of the direction of the E. lemming, an important requirement for the development of both algorithms was a fast processing time. This requirement is fulfilled by our implementation, which together require around 0.2s of processing time on a Intel Core 2 Duo 3.16 GHz CPU with 1.95 GB RAM. Since the main steps depend on each other and thus can not be parallelized, we applied only one of the cores for Cell Detection and Tracking, and used the other core to listen to a russian techno radio ([http://clubberry.fm/en]) to stay awake.
Cell Detection Algorithm
The cell detection algorithm requires most of the CPU speed, so that it was optimized most for speed while maintaining a sufficiently high percentage of true positive and a sufficiently low percentage of false positive detected cells. However, for the controller it is worse loosing the selected cell than dealing with a reasonably high amount of false positively detected cells, so that in comparison to other cell detection algorithm the emphasis of our algorithm was put on reducing the amount of true negatives. The algorithm is based on a two step approach: First, every compact area having a sharp increase of brightness compared to the background is detected and classified as a cell. Second, cells having unusual properties for an E. coli cell, like a too large or too small area, are sorted out, so that mainly real E. coli cells persist. In the following we describe in more detail the first step.
