This years invited speakers are (in alphabetical order):
- Auguste Genovesio, Institut de Biologie, ENS Paris
- Large Scale Image Analysis for Computational Biology
During the last 15 years automated image analysis has invaded research in cell biology in such a way that it now has become fully part of it. In this talk, we will visit a few example of image analysis approaches we designed to address computational biology challenges. We will describe a multi target tracking algorithm we developed to study the HIV virus dynamic during infection. We will also show how the miniaturization of samples and the full automation of their analysis led to process half million cell based conditions in parallel to discover therapeutic targets or decipher the mechanism of action of a drug. We will also show that both approaches can be combined to operate tracking on a large number of miniaturized experiments. Then we will describe technics we are developing to make possible the analysis of irregular tissue samples.
- Large Scale Image Analysis for Computational Biology
- Anna Kreshuk, HCI Heidelberg
- Image Analysis for Neural Circuit Reconstruction – and everything else!
Reconstruction of the directed graph of neural connections requires analysis of huge 3D stacks of electron microscopy images. The main components are segmentation of neurons (the nodes of the graph), detection of synapses (the edges of the graph) and estimation of synaptic partners (the directions of the graph edges). This talk is going to cover our efforts to solve these three problems using CNNs for pixelwise prediction and graphical models for higher level prior information. Besides, I will talk about ilastik – the software we write to make these and other learning-based methods accessible to life scientists, things we learned while making it and our future plans for it.
- Image Analysis for Neural Circuit Reconstruction – and everything else!
- Michael Liebling, IDIAP
- Leveraging motion redundancy to achieve fast volumetric imaging in cardiac development microscopy
Imaging the developing heart in vertebrate models such as the zebrafish and characterize its shape and motion patterns, in three-dimensions, could provide key insights to better understand normal and diseased cardiac development. Imaging the heart at early stages involves capturing cellular motion at slow temporal scales while the beating heart undergoes rapid contractions. These rapid motions make the imaging particularly challenging, as they can lead to motion artefacts—such as blur—and poor image quality due to low photon count. We have devised methods to improve specific imaging limitations (e.g. frame rate, limited dimensionality, low signal-to-noise) by leveraging the repeating nature of the cardiac heartbeat. In this presentation, I will highlight how we relaxed requirements on the redundancy assumption to allow us to use only quasi periodic signals (as opposed to strictly periodic signals) via an elastic registration procedure and how we derived efficient procedures to limit the contribution of outliers in reconstruction procedures by considering L1-norm data fidelity terms. Taken together, these techniques have allowed us to assemble composite multi-dimensional (volumetric) image sequences of the beating heart at resolutions exceeding those possible by direct imaging devices alone. Our results show that spatial and temporal redundancy in the sample motion can be exploited to computationlly push the limits of optical instruments. I will conclude by mentioning remaining limitations, inherent to our approach, and opportunities for further developments.
- Leveraging motion redundancy to achieve fast volumetric imaging in cardiac development microscopy
- Stephan Preibisch, MDC Berlin
- Image Reconstruction and Application to Whole-Organism Transcription Imaging
Modern developmental biology increasingly relies on imaging of large samples with high spatial and temporal resolution. Scientists can acquire high-resolution images of large living and fixed samples by extending the field-of-view of high-resolution optical microscopes using multi-tile or multi-view acquisition modes. We present software to interactively view and reconstruct such terabyte-sized acquisitions of developing organisms and entire cleared samples such as mouse brains or mouse embryos.
We will illustrate how we apply these powerful imaging tools to study transcriptional regulation underlying C. elegans development. We will present approaches that combine single-molecule fluorescent in-situ hybridization to detect individual mRNA molecules with lightsheet and electron microscopy acquisitions of entire C. elegans specimen. These powerful multi-modal approaches will enable us to decipher molecular mechanisms underlying important developmental events such as dosage compensation or choice of alternative developmental stages, and thereby ensure proper development of organisms.
- Image Reconstruction and Application to Whole-Organism Transcription Imaging
- Yannick Schwab, EMBL
- Capturing rare cells in heterogeneous samples with Correlative Light and Electron Microscopy (CLEM).
CLEM uniquely combines fluorescence imaging of living cells or tissues with electron microscopy to link functional assays on biological systems with high resolution ultrastructural information. My team is actively working on developing new methods that enhance the throughput for collecting structure-function data from both adherent cultured cells or from multicellular specimens. In this talk, I will present our latest progresses and applications for screening subcellular phenotypes on cultured cells challenged with genes knockdowns and for tracking single tumor cells in the mouse brain. Besides CLEM, this talk will also introduce volume electron microscopy, a set of techniques the impact of which is significantly increasing in the Cell Biology community.
- Capturing rare cells in heterogeneous samples with Correlative Light and Electron Microscopy (CLEM).
- Pavel Tomancak, MPI-CBG
- Developmental Lineage Tracing in Big Image Data
New microscopy paradigms, such as light sheet microscopy, enable capturing of complete developmental lineages of living organisms. That means that for a given biological system, be it an embryo or an organ, we can potentially follow the patterns of cell behaviour that build up its final shape and form. In combination with genetic labelling techniques, we can connect the cellular level description with the molecular one and assess its impact on the macroscopic tissue level. The computer vision challenges are to accurately segment and track cells in big 4D image data, to map static gene expression patterns to the dynamic lineages, to extract quantitative information about cell behaviour from the recordings and finally, to build multiscale models of tissue morphogenesis. I will present computational tools that were developed to meet these challenges and explain their impact on our understanding of enigmatic morphogenetic processes such as appendage formation and gastrulation during early embryonic development.
- Developmental Lineage Tracing in Big Image Data
- Natasa Sladoje, Uppsala University
- Tackling a major bottleneck in the BioImage Computing community – Network of European BioImage Analysts (NEUBIAS)
A pilot survey conducted in 2015, including 1841 researchers from the European bioimaging community, showed that more than 70% of all high-impact bioscience publications rely on advanced microscopy techniques, and that 95% of those surveyed considered image analysis important-to-essential in their research. Image analysis provides the key to boosting innovative microscopy technology even further. However, the huge investments made by research institutes in advanced microscopes (roughly 1-2 billion Euros in Europe), are not well balanced by the capacity to analyse the data produced. Image analysis is the final step in imaging-based research workflows and appears to be the narrowest bottleneck; only ca. 42% of captured images are quantitatively analysed. The survey also revealed that 68% of respondents considered support and training in image analysis lacking.
This situation with no doubts calls for engaged development of advanced and well suited methods to handle the sophisticated, challenging, and valuable bioimage data. However, reliable and fast automated data analysis requires not only advanced computational methods, but also high level of understanding of the details of sample preparation, as well as particular imaging processes. Bio-image computing grows as an interdisciplinary field which imposes demands on involved researchers (biologists, computer scientists, engineers) to collaborate, communicate, share and protect data, develop and benchmark algorithms and tools, and spread knowledge through the whole community of developers as well as users, via training courses, conferences, and publications.
This talk will present the activities of NEUBIAS – Network of European BioImage Analysts, initiated with an aim to address these challenges in BioImage computing.
- Tackling a major bottleneck in the BioImage Computing community – Network of European BioImage Analysts (NEUBIAS)
Previous editions...