Full success for FRIAS and BIOSS teams at the Third Cell Tracking Challenge 2015 in Brooklyn
The Cell Tracking Challenge is a programming competition, requiring participants to track moving cells in time-lapse video sequences. It was organised in the framework of the ISBI 2015 from April 16 – 19, 2015 in Brooklyn, USA. The conference consists of a range of lectures, tutorials, workshops and several challenges in different fields of bioimaging.
Tiago Esteves, Maja Temerinac-Ott and Pedro Quelhas have been working on the project since October 2014, when the challenge outline was handed out to the participants. All participants then had to send their test data back before the start of the symposium in April 2015. After receiving the award, Tiago Esteves presented the results of the team at the symposium in Brooklyn.
Another team from Freiburg, consisting of Olaf Ronneberger, Robert Bensch, Philipp Fischer and Thomas Brox from the Computer Science Department and BIOSS Centre for Biological Signaling Studies at the University of Freiburg, was also among the winners.
Maja Temerinac-Ott is a postdoc in computational biology with FRIAS External Senior Fellow Prof. Robert Murphy, focusing on “active learning-driven perturbagen analysis”. Robert Murphy’s laboratory combines research in cell with computational biology, specifically by linking fluorescence-based cell measurement methods with quantitative and computational methods. Tiago Esteves has been working together with Robert Murphy and Maja Temerinac-Ott as a guest researcher at FRIAS from October 2014 until January 2015.
The Third Cell Tracking Challenge 2015 built on the experiences, methods and results of the previous two rounds in 2013 and 2014 by using the existing dataset repository. These datasets included sequences of 2D and 3D fluorescently counterstained nuclei or whole cells, moving on top or immersed in a substrate, along with 2D Phase Contrast and Differential Interference Contrast (DIC) microscopy videos and realistic simulations of moving nuclei. The videos cover a wide range of cell types and quality. What made the third round especially demanding was the inclusion of new light-sheet microscopy 3D embryonic developmental data, which is probably the most challenging cell-tracking problem existing today.
Just like in a game of blind man’s bluff, the participants had to find ways to track the movement of cells with different properties. The team of Tiago Esteves and Maja Temerinac-Ott received the Third Cell Tracking Challenge Bitplane attendance awards for their cell-tracking method, which is based on the Laplacian of Gaussians (LoG) local detection and detection-association tracking. This method allows for the detection of regions in a digital image with different characteristics, such as brightness or color. To locate the cells, they used LoG filtering to enhance the image’s blob like structure, which in turn corresponds to cell locations. In order to analyse the mobility of the cells, they applied a detection-association tracking approach, where cells are first detected in each frame and cell-to-cell associations are performed for consecutive frames, based on a Euclidian distance nearest neighbour search. To estimate the cell’s shape they resorted to automatic image threshold or the Sliding Band Filter (SBF), which is a convergence filter and thus able to detect the convex shape of cells.
It is important to understand the movement of cells as they interact with their surrounding environment, in order to understand the mechanobiology of cell migration and its multiple implications in both tissue development and many diseases. Thus, tracking moving cells in time-laps video sequences is required for various applications both for scientific and industrial purposes.