Kilias, A. | bioRxiv Science

bioRxiv2021

Incorporating structural knowledge into unsupervised deep learning for two-photon imaging data

Live imaging techniques, such as two-photon imaging, promise novel insights into cellular activity patterns at a high spatio-temporal resolution. While current deep learning approaches typically focus on specific supervised tasks in the analysis of such data, we investigate how structural knowledge can be incorporated into an unsupervised generative deep learning model directly at the level of the video frames. We exemplify the proposed approach with two-photon imaging data from hippocampal CA1 neurons in mice, where we account for spatial structure with convolutional neural network components, disentangle the neural activity of interest from the neuropil background signal with separate foreground and background encoders and model gradual temporal changes by imposing smoothness constraints. Taken together, our results illustrate how such architecture choices facilitate a modeling approach that combines the flexibility of deep learning with the benefits of domain knowledge, providing an interpretable, purely image-based model of activity signals from live imaging data. Teaser sentenceUsing a neural network architecture that reflects domain knowledge provides an interpretable model of live cell imaging data.

bioinformatics↗

bioRxiv2021

Intracortical probe arrays with silicon backbone and microelectrodes on thin polyimide wings enable long-term stable recordings in vivo

ObjectiveRecording and stimulating neuronal activity across different brain regions requires interfacing at multiple sites using dedicated tools while tissue reactions at the recording sites often prevent their successful long-term application. This implies the technological challenge of developing complex probe geometries while keeping the overall footprint minimal, and of selecting materials compatible with neural tissue. While the potential of soft materials in reducing tissue response is uncontested, the implantation of these materials is often limited to reliably target neuronal structures across large brain volumes. ApproachWe report on the development of a new multi-electrode array exploiting the advantages of soft and stiff materials by combining 7-m-thin polyimide wings carrying platinum electrodes with a silicon backbone enabling a safe probe implantation. The probe fabrication applies microsystems technologies in combination with a temporal wafer fixation method for rear side processing, i.e., grinding and deep reactive ion etching, of slender probe shanks and electrode wings. The wing-type neural probes are chronically implanted into the entorhinal-hippocampal formation in the mouse for in vivo recordings of freely behaving animals. Main resultsProbes comprising the novel wing-type electrodes have been realized and characterized in view of their electrical performance and insertion capability. Chronic electrophysiological in vivo recordings of the entorhinal-hippocampal network in the mouse of up to 104 days demonstrated a stable yield of channels containing identifiable multi-unit and single-unit activity outperforming probes with electrodes residing on a Si backbone. SignificanceThe innovative fabrication process using a process compatible, temporary wafer bonding allowed to realize new Michigan style probe arrays. The wing-type probe design enables a precise probe insertion into brain tissue and long-term stable recordings of unit activity due to the application of a stable backbone and 7-m-thin probe wings provoking locally a minimal tissue response and protruding from the glial scare of the backbone.

neuroscience↗