AI-Driven Markerless NHP 3D Behavioral Analysis System | Prisys Biotech

Non-Human Primates (NHPs) are indispensable models for studying higher brain functions and neurological disorders due to their neuroanatomical and cognitive similarities to humans. However, capturing and quantifying their natural behavior non-invasively remains a significant challenge.

The next-generation NHP behavioral analysis system by Prisys Biotech integrates computer vision, deep learning, and 3D reconstruction technologies. This convergence allows for the automated, granular, and multi-dimensional analysis of animal behavior, marking a transition from qualitative description to precise quantitative analytics in preclinical research.

Our system is built upon three technological pillars designed to maximize data integrity and animal welfare.

1. Markerless Design: Prioritizing Natural Behavior and Welfare

• Technological Principle: The system eliminates the need for reflective markers or wearable sensors. utilizing high-resolution video streams, advanced deep learning algorithms detect and track anatomical key points directly from the raw footage in real-time.
• Scientific Value: By removing physical markers, we eliminate potential discomfort, behavioral inhibition, or the risk of marker detachment. This ensures animals remain in a natural state, significantly enhancing the ecological validity of the data and adhering to the highest standards of animal ethics and welfare.

2. Multi-View Synchronized 3D Tracking: Beyond the 2D Plane

• Technological Principle: We deploy multiple synchronized, high-frame-rate cameras to capture behavior from distinct angles. Multi-view geometry and reconstruction algorithms fuse these 2D inputs to generate real-time 3D skeletal structures and spatial trajectories.
• Scientific Value: Researchers gain access to a complete spatial reconstruction of behavior. Beyond simple 2D heatmaps, the system quantifies complex kinematic parameters such as joint angles (e.g., knee flexion), limb trajectories, and postural dynamics. This depth of data is critical for analyzing gait abnormalities, motor coordination, and fine motor skills with a precision unattainable by traditional methods.

3. Deep Learning-Based Recognition: From Kinematics to Semantics

• Technological Principle: Trained on extensive datasets of annotated behavioral video, our deep neural networks go beyond skeletal tracking to automatically classify discrete behavioral events (e.g., walking, climbing, grasping, grooming). The system records the temporal sequence, frequency, and duration of these events.
• Scientific Value: This transforms raw kinematic data into biologically meaningful behavioral insights. It allows for the automatic generation of "ethograms" and the analysis of temporal behavioral patterns. The system's evolving AI algorithms ensure continuous improvement in recognition accuracy, enabling the detection of subtle phenotypic changes resulting from disease progression or pharmacological intervention.

This system is currently deployed to assess efficacy in various neuroscientific and disease model contexts:

• Movement Disorders (Parkinson's Disease, SCI): Precise quantification of gait parameters (stride length, cadence, symmetry), tremor amplitude, bradykinesia, and postural stability provides objective biological endpoints for evaluating neuroprotective agents or cell therapies.
• Pain Research & Analgesia: Objective identification of spontaneous pain-related behaviors-such as curling, guarding, weight-bearing avoidance, and site-specific scratching-combined with 3D motion analysis to assess the impact of pain on general activity and mobility.

Neuropsychiatric Models:

• Anxiety/Fear: Analysis of exploration in open-field tests, freezing duration, and risk-assessment behaviors.
• Motivation/Anhedonia: Quantification of approach behaviors toward rewards (food, social interaction), effort expenditure, and reaction velocity.
• Autoimmune Diseases (e.g., Rheumatoid Arthritis): Beyond joint inflammation, the system establishes a "Disease Behavior Ethogram," analyzing compensatory motor functions and behavioral shifts caused by pain, offering a new dimension for evaluating quality-of-life improvements.

The platform's flexible architecture supports applications that extend well beyond current standard models:

• Social Neuroscience: Tracking inter-subject distance, orientation, and interaction patterns in multi-animal setups to study the neural mechanisms of social hierarchy, aggression, and pro-social behavior.
• Sleep & Circadian Rhythms: the fine-grained analysis of sleep architecture and wakefulness, critical for studying circadian disruptions and sleep issues in neurodegenerative diseases.
• Fine Motor Control: High-precision tracking of finger movements and object manipulation offers insights for cortical motor control studies, Brain-Computer Interface (BCI) research, and rehabilitation strategies.
• Multi-Modal Data Fusion: The system's high-precision timestamps allow for millisecond-level synchronization with EEG, EMG, neuroimaging, and in vivo neural recordings. This capability facilitates closed-loop analysis, directly linking behavioral outputs to underlying neural circuit activity.

The AI-based markerless 3D behavioral analysis system by Prisys Biotech represents a methodological advancement in neuroscience. By liberating researchers from labor-intensive and subjective scoring, it delivers objective, high-dimensional datasets. Through the seamless integration of markerless tracking and deep learning, this system bridges the gap between complex NHP behavior and neural mechanisms, accelerating the translation of basic scientific discoveries into effective clinical therapies.