Quantitative comparison of tracking performance using TrackMate-Helper.

<p>TrackMate now offers multiple detections and linking algorithms. Each of these algorithms needs to be configured with a parameter set that can take a wide range of values. While adequate values for these parameters can often be estimated intuitively, a systematic approach is often desirable. Indeed optimizing tracking parameters through visual assessment can be difficult when following many objects.</p> <p>To this end, we developed a new tool called TrackMate-Helper, which performs automatic parameter sweeps over any combination of detector and tracker available in TrackMate and measures the tracking results accuracy using the Cell-Tracking-Challenge (CTC) metrics. TrackMate-Helper requires an input image and the corresponding tracking ground-truth. Importantly, TrackMate-Helper is built as an end-user tool with a user-friendly interface that conveniently allows configuring parameter sweeps over many combinations of tracking parameters. We envision that this systematic approach will benefit medium and high-throughput automatic tracking studies. By optimizing the tracking parameters on one movie, users will be able to find optimal tracking parameters for the rest of their dataset.&nbsp;</p> <p>Here, we used TrackMate-Helper to assess the performance of TrackMate on four datasets that cover a wide range of biological and imaging situations:</p> <ol> <li> <p>Migrating cancer cells imaged using fluorescence microscopy to visualize their nuclei. In this dataset, the cells are densely packed and divided during the experiment .&nbsp;</p> </li> <li> <p>Migrating T-cells imaged using phase-contrast microscopy).&nbsp;</p> </li> <li> <p>Neisseria meningitidis bacterial growth on agar pads. The bacteria are fluorescently labeled to visualize their membrane. Starting from a few single bacteria, the cells quickly divide several times and swarm the field of view.</p> </li> <li> <p>Glioblastoma-astrocytoma cells imaged using bright-field microscopy at high resolution.</p> </li> </ol> <p>The table included reports the best results for each detector and tracker combination. We tested from 3000 to 20000 different parameter settings for each combination.</p>