DeepEthogram, a machine learning pipeline for supervised behavior classification from raw pixels.

Videos of animal behavior are used to quantify researcher-defined behaviors of interest to study neural function, gene mutations, and pharmacological therapies. Behaviors of interest are often scored manually, which is time-consuming, limited to few behaviors, and variable across researchers. We created DeepEthogram: software that uses supervised machine learning to convert raw video pixels into an ethogram, the behaviors of interest present in each video frame. DeepEthogram is designed to be general-purpose and applicable across species, behaviors, and video-recording hardware. It uses convolutional neural networks to compute motion, extract features from motion and images, and classify features into behaviors. Behaviors are classified with above 90% accuracy on single frames in videos of mice and flies, matching expert-level human performance. DeepEthogram accurately predicts rare behaviors, requires little training data, and generalizes across subjects. A graphical interface allows beginning-to-end analysis without end-user programming. DeepEthogram's rapid, automatic, and reproducible labeling of researcher-defined behaviors of interest may accelerate and enhance supervised behavior analysis. Code is available at: https://github.com/jbohnslav/deepethogram.

Disruption of dorsal hippocampal - prefrontal interactions using chemogenetic inactivation impairs spatial learning.

The hippocampus (HPC) and prefrontal cortex (PFC) are both necessary for learning and memory-guided behavior. Multiple direct and indirect anatomical projections connect the two regions, and HPC - PFC functional interactions are mediated by diverse physiological network patterns, thought to sub serve various memory processes. Disconnection experiments using contralateral inactivation approaches have established the role of direct, ipsilateral projections from ventral and intermediate HPC (vHPC and iHPC) to PFC in spatial memory. However, numerous studies have also prominently implicated physiological interactions between dorsal HPC (dHPC) and PFC regions in spatial memory tasks, and recent reports have identified direct dHPC - PFC connections. Whether dHPC - PFC interactions are necessary for spatial learning and memory has yet to be tested. Here, we used a chemogenetic inactivation approach using virally-expressed DREADDs (designer receptors exclusively activated by designer drugs) in rats to investigate the role of dHPC - PFC interactions in learning a hippocampal - dependent spatial alternation task. We implemented a rapid learning paradigm for a continuous W-track spatial alternation task comprising two components: an outbound, working memory component, and an inbound, spatial reference memory component. We investigated the effect of contralateral inactivation of dHPC and PFC on learning this task as compared with naïve and vehicle injection controls, as well as ipsilateral inactivation of the same regions. Contralateral dHPC - PFC inactivation selectively led to a significant impairment in learning the spatial working memory task compared to control groups, but did not impair learning of the spatial reference memory task. Ipsilateral inactivation animals showed similar learning rates as animals in the control groups. In a separate experiment, we confirmed that bilateral inactivation of PFC also leads to an impairment in learning the spatial working memory task. Our results thus demonstrate that dHPC - PFC interactions are necessary for spatial alternation learning in novel tasks. In addition, they provide crucial evidence to support the view that physiological interactions between dHPC and PFC play a key role in spatial learning and memory.