Selfee, self-supervised features extraction of animal behaviors

  1. Yinjun Jia  Is a corresponding author
  2. Shuaishuai Li
  3. Xuan Guo
  4. Bo Lei
  5. Junqiang Hu
  6. Xiao-Hong Xu
  7. Wei Zhang  Is a corresponding author
  1. Tsinghua University, China
  2. Chinese Academy of Sciences, China

Abstract

Fast and accurately characterizing animal behaviors is crucial for neuroscience research. Deep learning models are efficiently used in laboratories for behavior analysis. However, it has not been achieved to use an end-to-end unsupervised neural network to extract comprehensive and discriminative features directly from social behavior video frames for annotation and analysis purposes. Here, we report a self-supervised feature extraction (Selfee) convolutional neural network with multiple downstream applications to process video frames of animal behavior in an end-to-end way. Visualization and classification of the extracted features (Meta-representations) validate that Selfee processes animal behaviors in a way similar to human perception. We demonstrate that Meta-representations can be efficiently used to detect anomalous behaviors that are indiscernible to human observation and hint in-depth analysis. Furthermore, time-series analyses of Meta-representations reveal the temporal dynamics of animal behaviors. In conclusion, we present a self-supervised learning approach to extract comprehensive and discriminative features directly from raw video recordings of animal behaviors and demonstrate its potential usage for various downstream applications.

Data availability

Major data used in this study were uploaded to Dryad, including pretrained weights. Data could be accessed via:https://doi.org/10.5061/dryad.brv15dvb8.With the uploaded dataset and pretrained weights, our experiments could be replicated. However, due to its huge size and the limited internet service resources, we are currently not able to share our full training dataset. The full dataset is as large as 400GB, which is hard to upload to a public server and will be difficult for others users to download.For training dataset, it would be available from the corresponding author upon reasonable request (wei_zhang@mail.tsinghua.edu.cn), and then we can discuss how to transfer the dataset. No project proposal is needed as long as the dataset is not used for any commercial purpose.Our Python scripts could be accessed on GitHub: https://github.com/EBGU/SelfeeOther software used in our project include ImageJ(https://imagej.net/software/fiji/) and GraphPad Prism(https://www.graphpad.com/).All data used to plot graphs and charts in the manuscript can be fully accessed on Dryad (DOI 10.5061/dryad.brv15dvb8).

The following data sets were generated
The following previously published data sets were used

Article and author information

Author details

  1. Yinjun Jia

    IDG/McGovern Institute for Brain Research, Tsinghua University, Beijing, China
    For correspondence
    jyj20@mails.tsinghua.edu.cn
    Competing interests
    The authors declare that no competing interests exist.
  2. Shuaishuai Li

    Institute of Neuroscience, Chinese Academy of Sciences, Beijing, China
    Competing interests
    The authors declare that no competing interests exist.
  3. Xuan Guo

    IDG/McGovern Institute for Brain Research, Tsinghua University, Beijing, China
    Competing interests
    The authors declare that no competing interests exist.
  4. Bo Lei

    IDG/McGovern Institute for Brain Research, Tsinghua University, Beijing, China
    Competing interests
    The authors declare that no competing interests exist.
  5. Junqiang Hu

    IDG/McGovern Institute for Brain Research, Tsinghua University, Beijing, China
    Competing interests
    The authors declare that no competing interests exist.
  6. Xiao-Hong Xu

    Institute of Neuroscience, Chinese Academy of Sciences, Beijing, China
    Competing interests
    The authors declare that no competing interests exist.
  7. Wei Zhang

    IDG/McGovern Institute for Brain Research, Tsinghua University, Beijing, China
    For correspondence
    wei_zhang@mail.tsinghua.edu.cn
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0512-3096

Funding

National Natural Science Foundation of China (32022029)

  • Wei Zhang

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Ethics

Animal experimentation: All mating experiments were approved by the Animal Care and Use Committee of the Institute of Neuroscience, CAS Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences, Shanghai, China (IACUC No. NA-016-2016)All studies and experimental protocols of CIS and OFT were approved by Institutional Animal Care and Use Committee (IACUC) at Tsinghua University (No. 19-ZY1). Experiments were performed using the principles outlined in the Guide for the Care and Use of Laboratory Animals of Tsinghua University.

Copyright

© 2022, Jia et al.

This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 4,044
    views
  • 756
    downloads
  • 15
    citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.

Download links

Share this article

https://doi.org/10.7554/eLife.76218

Further reading

    1. Neuroscience
    Yisi Liu, Pu Wang ... Hongwei Zhou
    Short Report

    The increasing use of tissue clearing techniques underscores the urgent need for cost-effective and simplified deep imaging methods. While traditional inverted confocal microscopes excel in high-resolution imaging of tissue sections and cultured cells, they face limitations in deep imaging of cleared tissues due to refractive index mismatches between the immersion media of objectives and sample container. To overcome these challenges, the RIM-Deep was developed to significantly improve deep imaging capabilities without compromising the normal function of the confocal microscope. This system facilitates deep immunofluorescence imaging of the prefrontal cortex in cleared macaque tissue, extending imaging depth from 2 mm to 5 mm. Applied to an intact and cleared Thy1-EGFP mouse brain, the system allowed for clear axonal visualization at high imaging depth. Moreover, this advancement enables large-scale, deep 3D imaging of intact tissues. In principle, this concept can be extended to any imaging modality, including existing inverted wide-field, confocal, and two-photon microscopy. This would significantly upgrade traditional laboratory configurations and facilitate the study of connectomes in the brain and other tissues.

    1. Medicine
    2. Neuroscience
    Ayni Sharif, Matthew S Jeffers ... Manoj M Lalu
    Research Article

    C-C chemokine receptor type 5 (CCR5) antagonists may improve both acute stroke outcome and long-term recovery. Despite their evaluation in ongoing clinical trials, gaps remain in the evidence supporting their use. With a panel of patients with lived experiences of stroke, we performed a systematic review of animal models of stroke that administered a CCR5 antagonist and assessed infarct size or behavioural outcomes. MEDLINE, Web of Science, and Embase were searched. Article screening and data extraction were completed in duplicate. We pooled outcomes using random effects meta-analyses. We assessed risk of bias using the Systematic Review Centre for Laboratory Animal Experimentation (SYRCLE) tool and alignment with the Stroke Treatment Academic Industry Roundtable (STAIR) and Stroke Recovery and Rehabilitation Roundtable (SRRR) recommendations. Five studies representing 10 experiments were included. CCR5 antagonists reduced infarct volume (standard mean difference −1.02; 95% confidence interval −1.58 to −0.46) when compared to stroke-only controls. Varied timing of CCR5 administration (pre- or post-stroke induction) produced similar benefit. CCR5 antagonists significantly improved 11 of 16 behavioural outcomes reported. High risk of bias was present in all studies and critical knowledge gaps in the preclinical evidence were identified using STAIR/SRRR. CCR5 antagonists demonstrate promise; however, rigorously designed preclinical studies that better align with STAIR/SRRR recommendations and downstream clinical trials are warranted. Prospective Register of Systematic Reviews (PROSPERO CRD42023393438).