Our goal is to use deep learning networks to understand which neurons in the brain encode fine motor movements in mice. We collected large datasets entailing calcium imaging data of active neurons and high-resolution videos when mice perform motor tasks. We want to use recent advances in deep learning to (1) estimate the poses of mouse body parts at a high spatiotemporal resolution (2) extract behaviorally-relevant information and (3) align them with neural activity data. Behavioral video analysis is made possible by transfer learning, the ability to take a network that was trained on a task with a large supervised dataset and utilize it on a small supervised dataset. This has been used e.g. in a human pose–estimation algorithm called DeeperCut. Recently, such algorithms were tailored for use in the laboratory in a Python-based toolbox known as DeepLabCut, providing a tool for high-throughput behavioral video analysis.

The project has three parts: (1) Train neural networks through frame extraction and posture annotation using DeepLabCut, (2) Apply network to large datasets using DeepLabCut and extract behaviorally-relevant information using Python/Matlab custom-written scripts, (3) Align behavioral data with neural activity data and perform cross-correlations using Python or Matlab custom-written scripts. Ultimately this project will allow to understand the relationship between motor function and neural activity in specific cell-types in the brain. The data science tools acquired by the student have multiple applications, either in circuit neuroscience (e.g. animal behavior, deep brain imaging) or in computer vision (eg animal or human pose/posture estimation).

One selected candidate will receive a stipend via the DSI Scholars program. Amount is subject to available funding.

Faculty Advisor

  • Professor: Christoph Kellendonk
  • Department/School: Department of Psychiatry
  • Location: CUMC, New York State Psychiatric Institute, 40 Haven Avenue, Kolb Annex 3rd Floor
  • Our goal is to characterize novel brain circuits and in vivo brain signatures of behavior in mice with a focus on motor function, motivation or cognition. To this end, we combine mouse genetic tools, in vivo cell-specific brain imaging, behavioral tests and large dataset analyses.

Project Timeline

  • Earliest starting date: 3/1/2020
  • End date: 8/31/2020
  • Number of hours per week of research expected during Spring 2020: ~8
  • Number of hours per week of research expected during Summer 2020: ~Full-time

Candidate requirements

  • Skill sets:

    • Knowledge in github and Python (in particular the anaconda environment) and interest in getting more proficient is highly preferred.
    • Ability to write (or learn how to write) new Python or Matlab scripts for data analyses based on existing scripts is highly preferred.
    • Knowledge in pose estimation / usage of existing deep neural networks, in particular DeepLabCut or equivalent could be helpful.

  • Student eligibility: freshman, sophomore, junior, senior, master’s

  • International students on F1 or J1 visa: eligible

  • Additional comments:

    • High level of independence (the primary mentor is an expert in biological data collection in mouse models, but not an expert in data analysis. There are several people with data experience available in the lab to ask questions). It is expected that the student will be the main driver for the development of the data processing and analysis (with support/discussion with us).
    • Highly self-motivated, passionate, driven, ambitious.
    • Desire to bring projects to completion.
    • Good organizational and communication skills (to present work to scientists)
    • Interest in neuroscience, in particular calcium imaging of individual cell populations or motor function.
    • Pros: possibility to contribute to a larger research project expected to be published in a peer-reviewed publication in 2020-2021, student will be integrated into a neuroscience research team and can learn about the experimental side of neuroscience. Mentoring will include weekly/biweekly discussions with mentor (experimentalist) and a collaborator with previous experience with DeepLabCut. Possibility to present data in the form of posters/talks. Will be given full responsibility and credit for contribution, that can be acknowledged as authorship in publication, letter of recommendations, etc. depending on the level of achievement.