Model-Based Generation of Synthetic 3D Time-Lapse Sequences of Multiple Mutually Interacting Motile Cells with Filopodia

• Published in: Simulation and Synthesis in Medical Imaging Vol. 11037; pp. 71 - 79
• Main Authors: Peterlík, Igor, Svoboda, David, Ulman, Vladimír, Sorokin, Dmitry V., Maška, Martin
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2018; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: 3D time-lapse sequence; Cell deformation; Cell interaction; Filopodia; Simulation
• ISBN: 3030005356; 9783030005351
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-00536-8_8

Complementing collections of 3D time-lapse image data with comprehensive manual annotations is an extremely laborious and often impracticable task, which hinders objective benchmarking of bioimage analysis workflows as well as training of widespread deep-learning-based approaches. In this paper, we present a novel simulation system capable of generating synthetic 3D time-lapse sequences of multiple mutually interacting cells with filopodial protrusions, accompanied by inherently generated reference annotations, in order to stimulate the development of fully 3D bioimage analysis workflows for filopodium segmentation and tracking in complex scenarios with multiple mutually interacting cells. The system integrates its predecessor, which was designed for single-cell, collision-unaware scenarios only, with proactive, mechanics-based handling of collisions between multiple filopodia, multiple cell bodies, or their combinations. We demonstrate its potential on two generated 3D time-lapse sequences of multiple lung cancer cells with curvilinear filopodia, which visually resemble confocal fluorescence microscopy image data.