Schweihoff, Jens Florian: Ai-based, behavior dependent approaches for connectomic reconstruction of neuronal circuits. - Bonn, 2022. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5-65918
@phdthesis{handle:20.500.11811/9702,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5-65918,
author = {{Jens Florian Schweihoff}},
title = {Ai-based, behavior dependent approaches for connectomic reconstruction of neuronal circuits},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2022,
month = mar,

note = {Characterizing the functional architecture of neuronal circuits that underly complex behavior requires identifying active neuronal ensembles during behavioral expressions of interest. The recent development of light-induced, activity-dependent labeling enables to capture active neuronal ensembles dependent on ongoing behavior, effectively allowing the behavior-dependent, causal identification of relevant structures for subsequent investigation.
However, the behavior-dependent labeling of active neuronal ensembles was limited so far by a lack of dynamic closed-loop feedback systems that reliably detect unconstrained behavioral expressions. To solve this, I developed DeepLabStream (DLStream). DLStream is a versatile closed-loop toolkit providing real-time pose estimation of animals and conducting behavior-dependent experiments. DLStream has a temporal resolution in the millisecond range, is published open-source, and integrates other open-source projects such as deep learning-based pose estimation networks (DLC, SLEAP, DeepPoseKit), GPIO control (Arduino, Raspberry Pi), and machine learning-based behavior classification (B-SoiD, SimBA). To demonstrate DLStream's capabilities, I used the toolkit to label neuronal ensembles active during specific head directions utilizing Cal-Light, a light-induced, activity-dependent biomolecular labeling system. Behavior-dependent light stimulation resulted in labeling of neuronal ensembles active during specific episodes of head direction. Importantly, this experimental strategy has the potential to untangle previously unknown causal relationships. This can be achieved by combining connectomic analysis of the captured ensembles and consecutive manipulation of their neuronal activity.
Additionally, I established the Tetbow system, a virus-mediated, multicolor labeling system that can eventually be combined with behavior-dependent labeling to allow the anatomic analysis of large-scale tissue samples with behavior-dependent, uniquely labeled neuronal ensembles. Here, the focus lay in the effective use of Tetbow labeled samples in a collaborative attempt to develop an automatic segmentation tool to segment uniquely colored neurons in large tissue samples. Notably, some of the results of this thesis were published, including additional experiments using DLStream.},

url = {https://hdl.handle.net/20.500.11811/9702}
}

The following license files are associated with this item:

InCopyright