Increases in the data rates of detectors for electron microscopy (EM) have outpaced increases in network, mass storage and memory bandwidth by two orders of magnitude between 2009(Weber, 2018. The LiberTEM open source platform (Clausen et al., 2020) is designed to match the growing performance requirements of EM data processing (Weber, Clausen, & Dunin-Borkowski, 2020). MotivationThe data rate of the fastest detectors for electron microscopy that are available in 2019 exceeds 50 GB/s, which is faster than the memory bandwidth of typical personal computers (PCs) at this time. Applications from ten years before that ran smoothly on a typical PC have evolved into numerical analysis of complex multidimensional datasets (Ophus, 2019) that require distributed processing on high-performance systems. Furthermore, electron microscopy is interactive and visual, and experiments performed inside electron microscopes (so-called in situ experiments) often rely on fast on-line data processing as the experimental parameters need to be adjusted based on the observation results. As a consequence, modern data processing systems for electron microscopy should be designed for very high throughput in combination with short response times for interactive GUI use and closed-loop feedback. That requires fundamental changes in the architecture and programming model, and consequently in the implementation of algorithms and user interfaces for electron microscopy applications.

No abstract

In this paper, we propose a hybrid approach for multi-object microbial cell segmentation. The approach combines an MLbased detection with a geometry-aware variational-based segmentation using B-splines that are parametrized based on a geometric model of the cell shape. The detection is done first using YOLOv5. In a second step, each detected cell is segmented individually. Thus, the segmentation only needs to be done on a per-cell basis, which makes it amenable to a variational approach that incorporates prior knowledge on the geometry. Here, the contour of the segmentation is modelled as closed uniform cubic B-spline, whose control points are parametrized using the known cell geometry. Compared to purely ML-based segmentation approaches, which need accurate segmentation maps as training data that are very laborious to produce, our method just needs bounding boxes as training data. Still, the proposed method performs on par with ML-based segmentation approaches usually used in this context. We study the performance of the proposed method on time-lapse microscopy data of Corynebacterium glutamicum.

No abstract

To address the need for ground truth data for deep learning based segmentation algorithms in microfluidic live-cell imaging, we present CellSium, a cell simulator primarily aimed at synthesizing realistic image sequences of bacterial microcolonies suitable for training neural networks. Availability and Implementation: CellSium is free and open source software under the BSD license, implemented in Python, available at https://github.com/modsim/cellsium (DOI: 10.5281/zenodo.6193033), along with documentation, usage examples and Docker images.