Deep-learning two-photon fiberscopy for video-rate brain imaging in freely-behaving mice

Authors

Honghua Guan, Department of Electrical and Computer Engineering, Johns Hopkins University, Baltimore, MD, 21218, USA.
Dawei Li, Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, 21205, USA.
Hyeon-Cheol Park, Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, 21205, USA.
Ang Li, Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, 21205, USA.
Yuanlei Yue, Department of Pharmacology and Physiology, School of Medicine and Health Sciences, George Washington University, Washington, DC, 20052, USA.
Yungtian A. Gau, Solomon H. Snyder Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, MD, 21205, USA.
Ming-Jun Li, Science and Technology Division, Corning Incorporated, Corning, NY, 14831, USA.
Dwight E. Bergles, Solomon H. Snyder Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, MD, 21205, USA.
Hui Lu, Department of Pharmacology and Physiology, School of Medicine and Health Sciences, George Washington University, Washington, DC, 20052, USA.
Xingde Li, Department of Electrical and Computer Engineering, Johns Hopkins University, Baltimore, MD, 21218, USA. xingde@jhu.edu.

Document Type

Journal Article

Publication Date

3-22-2022

Journal

Nature communications

Volume

13

Issue

1

DOI

10.1038/s41467-022-29236-1

Abstract

Scanning two-photon (2P) fiberscopes (also termed endomicroscopes) have the potential to transform our understanding of how discrete neural activity patterns result in distinct behaviors, as they are capable of high resolution, sub cellular imaging yet small and light enough to allow free movement of mice. However, their acquisition speed is currently suboptimal, due to opto-mechanical size and weight constraints. Here we demonstrate significant advances in 2P fiberscopy that allow high resolution imaging at high speeds (26 fps) in freely-behaving mice. A high-speed scanner and a down-sampling scheme are developed to boost imaging speed, and a deep learning (DL) algorithm is introduced to recover image quality. For the DL algorithm, a two-stage learning transfer strategy is established to generate proper training datasets for enhancing the quality of in vivo images. Implementation enables video-rate imaging at ~26 fps, representing 10-fold improvement in imaging speed over the previous 2P fiberscopy technology while maintaining a high signal-to-noise ratio and imaging resolution. This DL-assisted 2P fiberscope is capable of imaging the arousal-induced activity changes in populations of layer2/3 pyramidal neurons in the primary motor cortex of freely-behaving mice, providing opportunities to define the neural basis of behavior.

APA Citation

Guan, Honghua; Li, Dawei; Park, Hyeon-Cheol; Li, Ang; Yue, Yuanlei; Gau, Yungtian A.; Li, Ming-Jun; Bergles, Dwight E.; Lu, Hui; and Li, Xingde, "Deep-learning two-photon fiberscopy for video-rate brain imaging in freely-behaving mice" (2022). GW Authored Works. Paper 557.
https://hsrc.himmelfarb.gwu.edu/gwhpubs/557

Department

Pharmacology and Physiology