2020/03/17 by Jean Ollion, Ollion, Jean, Charles Ollion +1
Biochemistry, Genetics and Molecular Biology · Engineering · #Cell Image Analysis Techniques #Image Processing Techniques and Applications #Advanced Fluorescence Microscopy Techniques
paper · pdf · doi:10.48550/arxiv.2003.07790
The mother machine is a popular microfluidic device that allows long-term\ntime-lapse imaging of thousands of cells in parallel by microscopy. It has\nbecome a valuable tool for single-cell level quantitative analysis and\ncharacterization of many cellular processes such as gene expression and\nregulation, mutagenesis or response to antibiotics. The automated and\nquantitative analysis of the massive amount of data generated by such\nexperiments is now the limiting step. In particular the segmentation and\ntracking of bacteria cells imaged in phase-contrast microscopy---with error\nrates compatible with high-throughput data---is a challenging problem.\n In this work, we describe a novel formulation of the multi-object tracking\nproblem, in which tracking is performed by a regression of the bacteria's\ndisplacement, allowing simultaneous tracking of multiple bacteria, despite\ntheir growth and division over time. Our method performs jointly segmentation\nand tracking, leveraging sequential information to increase segmentation\naccuracy.\n We introduce a Deep Neural Network architecture taking advantage of a\nself-attention mechanism which yields extremely low tracking error rate and\nsegmentation error rate. We demonstrate superior performance and speed compared\nto state-of-the-art methods. Our method is named DiSTNet which stands for\nDISTance+DISplacement Segmentation and Tracking Network.\n While this method is particularly well suited for mother machine microscopy\ndata, its general joint tracking and segmentation formulation could be applied\nto many other problems with different geometries.\n