Evaluation of visual performance for RCS rats: A tailor-made OMR setup via DeepLabCut and Psychopy with a customized algorithm

Jing Wang, Thomas Rüland, Sandra Krause, Michael Schöneck, N. Jon Shah, Frank Müller, Björn M. Kampa, Karl‐Josef Langen, Antje WilluweitbioRxiv 2025 · DOI

TL;DR

Researchers have developed a novel, customizable optomotor response (OMR) system for rats using DeepLabCut and PsychoPy. This system accurately tracks visual performance decline in a rat model of retinal degeneration (RCS rats), showing significant vision loss starting at 3 months of age.

What Problem Does This Paper Address?

Evaluating visual function in animal models is crucial for understanding eye diseases and testing new therapies. Traditional methods for assessing visual performance in rodents, like the optomotor response (OMR), often rely on commercial systems with fixed parameters. These systems can be expensive and lack the flexibility to adapt to the specific needs of different rat strains or research questions. For instance, the optimal visual stimuli (like spatial frequency and grating velocity) can vary significantly between rat lines. This study addresses the need for a more adaptable and cost-effective OMR system that can precisely track visual deficits in disease models, specifically the Royal College of Surgeons (RCS) rat, a well-established model for inherited retinal degeneration.

Methods

The core of this work is the development of a tailor-made OMR setup. This setup integrates three key components:

  • DeepLabCut: An open-source tool for markerless pose estimation, used here to track the real-time position of the rats' ears. This allows for precise measurement of head movements in response to visual stimuli.
  • PsychoPy: A popular open-source software for designing psychological experiments, used to generate the visual stimuli (moving gratings) presented to the rats.
  • Customized Algorithms: Developed in-house to automate the analysis of the OMR data collected by DeepLabCut and PsychoPy. This includes calculating the percentage of correct responses (PoC) based on head movements aligning with the visual stimulus.

The OMR box itself was adapted from a mouse OMR setup and scaled up for rats, featuring four 24" monitors within a custom-built PVC enclosure. This design aimed to provide ample space, prevent unintended screen contact, and facilitate video capture. The system also incorporated reflective mirrors to create an illusion of depth, potentially reducing animal stress and platform jumping.

Datasets

The study utilized pigmented Royal College of Surgeons (RCS) rats and wild-type (WT) control rats. Animals were tested at various ages: postnatal days 30 (P30) and 50 (P50), and at 3, 6, and 12 months of age. No specific external datasets were used for training; the DeepLabCut network was trained on videos generated within this study's experimental setup.

Key Insights and Ideas

  • Flexibility is Key: The development of a customizable OMR system is a significant advantage over rigid commercial setups, allowing researchers to fine-tune parameters like spatial frequency and contrast for specific rat strains and research questions.
  • Deep Learning for Behavior: Leveraging DeepLabCut for markerless pose estimation provides a robust and objective way to quantify animal behavior, reducing experimenter bias and enabling more detailed analysis of head movements.
  • RCS Rat Model Behavior: The study provides a detailed timeline of visual decline in RCS rats, showing normal vision in young animals but a sharp drop-off starting around 3 months, leading to near-blindness by 6-12 months.
  • Correlation of Structure and Function: The research links the observed functional decline in OMR performance with histological evidence of photoreceptor degeneration in the RCS rat retina, strengthening the validity of both the OMR model and the histological findings.

Key Results

  • The developed OMR setup demonstrated high efficiency and reproducibility.
  • Optimal spatial frequency (SF) and grating velocity were determined for RCS and WT rats.
  • RCS rats showed normal OMR performance at P30 and P50.
  • A significant decline in OMR performance (Percentage of Correct responses, PoC) in RCS rats was observed starting at 3 months of age.
  • By 6 and 12 months, most RCS rats exhibited PoC values around 50%, indicating blindness (chance level performance).
  • Histological analysis confirmed progressive photoreceptor degeneration in RCS rats, correlating with the observed functional deficits.

Limitations and Future Work

  • Cone Function in Later Stages: While the study identified surviving cone-like cells in older RCS rats, their precise role in the residual vision remains uncertain. The authors suggest that the role of cones in later stages of retinal degeneration requires further investigation.
  • Naturalistic Stimuli: The paper notes that while drifting gratings are standard, more naturalistic visual stimuli (e.g., broadband orientations) might be more efficient in activating visual neurons. The custom-built OMR chamber could facilitate research into these more complex stimuli.
  • Discrepancies in Literature: The authors acknowledge that discrepancies exist between histological and functional read-outs in the literature for retinal degeneration models. While their data suggest surviving cones might account for preserved OMR in 3-month-old RCS rats, further work is needed to fully reconcile these findings.

Summary updated

Abstract

Abstract In this study, we developed an optimized optomotor response (OMR) setup for rats based on existing commercial systems and introduced a novel method that integrates DeepLabCut, PsychoPy, and custom algorithms to automate OMR performance evaluation in rats. Our primary findings revealed the optimal spatial frequency (SF) and grating velocity for pigmented Royal College of Surgeons (RCS) and wild-type (WT) rats. The RCS rats, a model for retinal degeneration, displayed normal OMR performance at postnatal days 30 (P30) and 50 (P50), but a significant decline was observed at 3 months. In most rats aged 6 and 12 months, the percentage of correct responses (PoC) was 50%, indicating blindness. The high efficiency and reproducibility of our non-invasive OMR test, along with its suitability for post-surgical and retinal degeneration models, underscores its potential for broad applications in visual performance evaluation.
BibTeX
@misc{wang2025evaluation,
  author = {Jing Wang and Thomas Rüland and Sandra Krause and Michael Schöneck and N. Jon Shah and Frank Müller and Björn M. Kampa and Karl‐Josef Langen and Antje Willuweit},
  title = {{Evaluation of visual performance for RCS rats: A tailor-made OMR setup via DeepLabCut and Psychopy with a customized algorithm}},
  year = {2025},
  doi = {10.1101/2025.05.24.655932},
  howpublished = {bioRxiv},
}
Figure