DeepD3, an open framework for automated quantification of dendritic spines.

Dendritic spines are the seat of most excitatory synapses in the brain, and a cellular structure considered central to learning, memory, and activity-dependent plasticity. The quantification of dendritic spines from light microscopy data is usually performed by humans in a painstaking and error-prone process. We found that human-to-human variability is substantial (inter-rater reliability 82.2±6.4%), raising concerns about the reproducibility of experiments and the validity of using human-annotated 'ground truth' as an evaluation method for computational approaches of spine identification. To address this, we present DeepD3, an open deep learning-based framework to robustly quantify dendritic spines in microscopy data in a fully automated fashion. DeepD3's neural networks have been trained on data from different sources and experimental conditions, annotated and segmented by multiple experts and they offer precise quantification of dendrites and dendritic spines. Importantly, these networks were validated in a number of datasets on varying acquisition modalities, species, anatomical locations and fluorescent indicators. The entire DeepD3 open framework, including the fully segmented training data, a benchmark that multiple experts have annotated, and the DeepD3 model zoo is fully available, addressing the lack of openly available datasets of dendritic spines while offering a ready-to-use, flexible, transparent, and reproducible spine quantification method.

Limited functional convergence of eye-specific inputs in the retinogeniculate pathway of the mouse.

Segregation of retinal ganglion cell (RGC) axons by type and eye of origin is considered a hallmark of dorsal lateral geniculate nucleus (dLGN) structure. However, recent anatomical studies have shown that neurons in mouse dLGN receive input from multiple RGC types of both retinae. Whether convergent input leads to relevant functional interactions is unclear. We studied functional eye-specific retinogeniculate convergence using dual-color optogenetics in vitro. dLGN neurons were strongly dominated by input from one eye. Most neurons received detectable input from the non-dominant eye, but this input was weak, with a prominently reduced AMPAR:NMDAR ratio. Consistent with this, only a small fraction of thalamocortical neurons was binocular in vivo across visual stimuli and cortical projection layers. Anatomical overlap between RGC axons and dLGN neuron dendrites alone did not explain the strong bias toward monocularity. We conclude that functional eye-specific input selection and refinement limit convergent interactions in dLGN, favoring monocularity.

Spatially specific vs. unspecific disruption of visual orientation perception using chronometric pre-stimulus TMS.

Transcranial magnetic stimulation (TMS) over occipital cortex can impair visual processing. Such "TMS masking" has repeatedly been shown at several stimulus onset asynchronies (SOAs), with TMS pulses generally applied after the onset of a visual stimulus. Following increased interest in the neuronal state-dependency of visual processing, we recently explored the efficacy of TMS at "negative SOAs", when no visual processing can yet occur. We could reveal pre-stimulus TMS disruption, with results moreover hinting at two separate mechanisms in occipital cortex biasing subsequent orientation perception. Here we extended this work, including a chronometric design to map the temporal dynamics of spatially specific and unspecific mechanisms of state-dependent visual processing, while moreover controlling for TMS-induced pupil covering. TMS pulses applied 60-40 ms prior to a visual stimulus decreased orientation processing independent of stimulus location, while a local suppressive effect was found for TMS applied 30-10 ms pre-stimulus. These results contribute to our understanding of spatiotemporal mechanisms in occipital cortex underlying the state-dependency of visual processing, providing a basis for future work to link pre-stimulus TMS suppression effects to other known visual biasing mechanisms.