Abstractocyte: A Visual Tool for Exploring Nanoscale Astroglial Cells

Abstract

This thesis presents the design and implementation of Abstractocyte, a system for the visual analysis of astrocytes, and their relation to neurons, in nanoscale volumes of brain tissue. Astrocytes are glial cells, i.e., non-neuronal cells that support neurons and the nervous system. Even though glial cells make up around 50 percent of all cells in the mammalian brain, so far they have been far less studied than neurons. Nevertheless, the study of astrocytes has immense potential for understanding brain function. However, the complex and widely-branching structure of astrocytes requires high-resolution electron microscopy imaging and makes visualization and analysis challenging. Using Abstractocyte, biologists can explore the morphology of astrocytes at various visual abstraction levels, while simultaneously analyzing neighboring neurons and their connectivity. We define a novel, conceptual 2D abstraction space for jointly visualizing astrocytes and neurons. Neuroscientists can choose a joint visualization as a specific point in that 2D abstraction space. Dragging this point allows them to smoothly transition between different abstraction levels in an intuitive manner. We describe the design of Abstractocyte, and present three case studies in which neuroscientists have successfully used our system to assess astrocytic coverage of synapses, glycogen distribution in relation to synapses, and astrocyticmitochondria coverage.

Interactive Volumetric Visual Analysis of Glycogen‐derived Energy Absorption in Nanometric Brain Structures

“…Neuroscientists normally use standard surface visualization frameworks, like Blender, for performing visual inspection of neural structures decorated with glygogen distribution, in a way to speculate about proximity patterns and formulate hypothesis, even with the help of VR setups [CBB *16]. Recently, these surface rendering frameworks were enriched by highlighting absorption on surface reconstructions ( GLAM , proposed by Agus et al [ABG*18]), or by mapping glycogen content in abstract neurite representations ( Abstractocyte , proposed by Mohammed et al [MAAB*18]). After analysis sessions, neuroscientists concluded that all the frameworks can be considered complementary tools, since they provide different view of the same data.…”

“…After analysis sessions, neuroscientists concluded that all the frameworks can be considered complementary tools, since they provide different view of the same data. The main difference of our framework is that it works directly on EM volume data, that we use to compute the absorption information more accurately than the models proposed by Agus et al [ABG*18] and Mohammed et al [MAAB*18]. Figure left shows similar structures represented with Abstractocyte [MAAB*18](left) and the proposed framework (right): in this case a Multi Synaptic Bouton from mouse 3 connected to a dendrite.…”

“…The framework was moreover extended by providing visual tools for representing connectivity information by using subway lines metaphors [AABS*14], and visual tools in forms of blocks for tracking the segmentation and proof‐reading for large projects [AABH*16]. Similarly, techniques for exploring 2D abstract representations of neural structures have been recently introduced to help neuroscientists in correlation analysis of morphology [MAAB*18], while methods for analyzing absorption patterns have been proposed to enrich dense and sparse surface reconstructions of rodent brain structures [ABG*18]. Our solution extends a volume visualization framework [HBJP12] with a method for highlighting local absorption patterns and tools for performing connection based and absorption based queries, able to provide neuroscientists visual hints for comparative and correlative analysis of morphology structures and glycogen distributions.…”

“…On the basis of the requirement analysis, contrary to other systems [ABG*18, MAAB*18], which are designed around surface representations coming out of careful segmentation processes, we decided to employ and enrich volumetric representations. This is because, in general, domain scientists feel more comfortable with raw EM data, in which they can exploit their expertise for recognizing small‐scale specific features that cannot be easily highlighted in surface reconstructions (for example bouton vesicles and mitochondria).…”

“…To this end, some frameworks have been very recently proposed to support neuroscientists. They are targeted to provide specific abstract representations built around glial cellular structures [MAAB*18], or to improve visual perception of the influence of energy sources on cellular surface reconstructions through immersive environments [ABG*18]. Current approaches, however, do not target general analysis tasks in the neuroenergetics domain.…”