Part Three: Behavioral Neuroscience

Behavioral neuroscience studies the biological and neural mechanisms that affect thought, behavior, and emotions. By combining psychology and biology, it explores how our nervous system shapes actions. Below are several core areas of study:

Behavioral neuroscience has a vast range of applications. Using neuroimaging via fMRI and EEG, neural markers are identified for disorders such as ADHD and autism. These markers assist with the development of therapies (for instance, Cognitive Behavioral Therapy or CBT) by providing a framework, helping reshape maladaptive neural patterns. Besides this, physical and cognitive therapy strategies can be developed to help patients regain motor and cognitive function after traumatic brain injuries and strokes, and brain. Finally, lesion studies, which examine the impact of brain damage or removal of specific brain areas on behavior, allow researchers to determine specific functions of brain regions [1].

AI can take the vast applications of behavioral neuroscience a step further. Below are several impacts it’s made so far!

With AI, it’s possible to track behavior and motion across any animal species. Quantifying behavior is crucial for many applications in neuroscience. However, extracting certain parts of a behavior for more analysis can be very time consuming. Often, subjects of study are marked with reflective markers to assist with computer-based tracking. However, markers can be painful for the animal and can alter behavior. Additionally, researchers had to predict beforehand what body parts they wanted to track (so, this limited discovery greatly, since new or unexpected behavioral patterns would be completely missed).

DeepLabCut, a deep-learning based software, allows for easy video-based motion tracking across any animal species (for instance, fruit flies, mice, and humans). It tracks important body points, such as the paws, the head, and the tail, with accuracy matching scoring of human experts. Even when only a small amount of frames are labels, tracking performance was comparable [2].

Additionally, several deep learning algorithms have been developed to estimate animal posture during movement, greatly simplifying and speeding up the analysis of several behaviors []. It is challenging to accurately track animal movement under complex conditions. The YOLO (You Only Look Once) algorithm, a deep learning algorithm, was developed to overcome these challenges. The YOLO algorithm was combined with a background subtracting algorithm, labelled DeepBhvTracking.

With DeepBhvTracking, the movement of animals can be tracked accurately in these complex environments. Additionally, it can be used in different behavior frameworks, allowing for broad use in studies of neuroscience, ML (machine learning), and medicine.

This allows behavioral neuroscientists to study anxiety, grooming, locomotion, and social interaction quantitatively and objectively at a scale impossible with manual observation [4]. Through this, common evolutionary neurobiological mechanisms and phenotypes can be identified.

High-throughput screening is when hundreds of thousands of drugs compounds are tested at the same time to see how they affect animal behavior. With AI, this can be possible at very large scales. Zebrafish larvae are put on multi-well plates. Then, cameras record their behavior continuously (swimming patterns, speed, anxiety responses), and AI analyzes these videos. Following this, AI generates a behavioral barcode for each drug (how the compound changed animal behavior), and compared barcode across hundreds of compounds to find similarities, group drugs by mechanism, and find possible therapeutic candidates.

With these zebrafish, rapid phenotypic profiling is possible for thousands of compounds in vivo. These zebrafish have high levels of shared genetics and central nervous system anatomy with humans, and scale to high-throughput testing of complex behavioral readouts. So, researchers can scan entire libraries of drugs for new uses.

A neural network model, Z-LaP Tracker, is capable of quantifying behavior in zebrafish larvae relevant to cognitive function. With this model, a high-throughput screening of FDA-approved drugs was done to identify compounds that affect zebrafish behavior in a manner similar to that induced by calcineurin inhibitors (which may play a role in preventing Alzheimer’s). Behavioral profiles were generated for cluster analysis, then 64 candidate therapeutics were identified for neurodegenerative disorders [5]. With AI, the drug-discovery pipeline can be extremely accelerated for neurological and psychiatric conditions

Despite these advances in research, many gaps in knowledge remain. Currently, behavioral data is very subjective and inconsistent. Often, human experts differ in data interpretation, limiting the generalizability and accessibility of models. Thus, models trained on one lab’s data may not work as well for another lab’s. Additionally, we have an incomplete understanding of the brain itself. This can limit the advances AI makes, since AI can only build off of what we know. AI also cannot scale to study larger animal populations, because most AI models were designed to track one animal at a time. This limits statistical power, since it requires larger sample sizes. Without multiplexing, researchers have to run experiments sequentially (i.e. cage-by-cage), reducing reliability due to fluctuating conditions. This is also expensive and time-consuming. Finally, the lack of regulatory standards for AI behavioral analytic tools makes experiment reproducibility and clinical reliability less concrete [6].

In the future, it’s necessary to work on multimodal integration (combining video, EEG, genomic, and metabolomic data) to reveal the connections between molecules and behavior. Additionally, the development of fully autonomous vivariums (that is, AI-controlled environments that continuously monitor and respond to animal behavior without human intervention) has great promise [5].

Play this custom Connections game to check your understanding of behavioral neuroscience!