Skip navigation

DSpace JSPUI

DSpace preserves and enables easy and open access to all types of digital content including text, images, moving images, mpegs and data sets

Learn More
DSpace logo
English
中文
  • Browse
    • Communities
      & Collections
    • Publication Year
    • Author
    • Title
    • Subject
  • Search TDR
  • Rights Q&A
    • My Page
    • Receive email
      updates
    • Edit Profile
  1. NTU Theses and Dissertations Repository
  2. 電機資訊學院
  3. 電子工程學研究所
Please use this identifier to cite or link to this item: http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/93895
Title: TAG-SPARK: 深度學習提升高速雙光子體積成像系統應用於活體老鼠小腦運動神經之研究
TAG-SPARK: Empowering High-Speed Volumetric Imaging with Deep-Learning and Spatial Redundancy for the Cerebellar Neuronal Activities
Authors: 謝尹慈
Yin-Tzu Hsieh
Advisor: 林致廷
Chih-Ting Lin
Co-Advisor: 朱士維
Shi-Wei Chu
Keyword: 多光子顯微鏡,可調式聲波梯度折射率透鏡,高速體積成像,深度學習降噪演算法,空間冗餘資訊,神經網絡,Purkinje cell,
Multi-photon microscopy,Tunable acoustic gradient index lens,High-speed volumetric imaging,Deep-learning noise reduction,Spatial redundancy,Neural networks,Purkinje cells,
Publication Year : 2024
Degree: 碩士
Abstract: 揭開人體中複雜的神經網絡是理解大腦功能的重要一步。在神經活動的全面記錄中,同時取得大範圍神經群體,並擁有高時空解析度的體積成像技術至關重要,有助於理解神經訊號的傳遞與細胞之間的溝通。在神經科學領域中,雙光子高速鈣離子成像技術,以其亞微米空間解析度在捕捉神經活動中扮演重要角色。然而,由於取像速度與圖像質量之間的平衡,受限於有限的光子訊號流量,總導致低信噪比和對比度差的問題。
為了提升高速體積影像的品質,我們開發了一套深度學習降噪算法(TAG-lens-based SPAtial redundancy-driven noise Reduction Kernel, TAG-SPARK),並將其與具有視頻速率的雙光子體積成像系統整合。利用可調式聲波梯度折射率透鏡(TAG lens)實現高速連續焦點變化,取得垂直(axial)空間冗餘資訊,進行自監督模型訓練的開發。結合此降噪深度學習演算法的影像處理方法,該技術實現了超過300%的峰值信噪比(PSNR)和超過700%的信噪比 (SNR)增強,同時保持神經活動的訊號變化特徵。
為了展示此技術的能力,我們以觀察活體小鼠的Purkinje cells 個體與群體反應為例。Purkinje cell在小鼠運動時的樹突神經訊號表現被視為其關鍵特徵。樹突細胞的鈣離子變化與complex spike之間存在顯著相關性。儘管相關的研究已經證實神經訊號由樹突到細胞本體的傳遞概念具有基礎性意義,但僅限於二維的平面影像研究,仍需要高速體積成像進行完整的驗證。
因此,我們的研究旨在探索神經訊號如何在獨立神經細胞中從樹突到細胞本體的傳遞,以及群體細胞之間的相互溝通。技術不僅以高品質影像捕捉神經活動,還有助於我們更深入地理解三維神經結構中神經訊號的傳遞途徑。
Comprehensive recording of neural activities across large neuronal populations with high spatiotemporal resolution and volumetric imaging is crucial for understanding neural circuit signal transduction, information processing, and behavior generation. Two-photon high-speed fluorescence calcium imaging stands as a leading technique in neuroscience for capturing neural activities with sub-micrometer spatial resolution. However, challenges arise from the inherent trade-off between acquisition speed and image quality, resulting in a low signal-to-noise ratio (SNR) due to limited signal photon flux.
To enhance the quality of high-speed volumetric imaging, we developed a deep-learning denoising algorithm, TAG-SPARK (TAG-lens-based SPAtial redundancy-driven noise Reduction Kernel). We propose a home-built two-photon volumetric imaging system with video-rate capabilities, integrating image processing methods using a noise reduction deep learning algorithm. Leveraging the high-speed dense z-sampling at sub-micrometer-scale intervals of a TAG lens, we developed a self-supervised denoising algorithm that exploits the spatial redundancy of z-slices. This approach achieves >300% peak signal-to-noise ratio (PSNR) and >700% SNR enhancement while preserving the fast-spiking functional profiles of neuronal activities.
We demonstrated this technology's application by observing individual neurons and populations of Purkinje cells (PCs) in awake mice. PC activities, which induce dendritic calcium spikes, have been identified as a key feature of motor initiation. The correlation between dendritic and somatic complex spikes is significant. While the foundational importance of the concept of neuronal signal transmission from dendrites to soma is acknowledged, validating this requires high-speed volumetric calcium imaging.
Therefore, our study aims to explore how neural signals propagate from dendrites to cell bodies within individual neurons, as well as communication among neuronal populations. This tailored technique enables capturing neuronal activities with high SNR, advancing our understanding of neuronal signal transduction pathways within 3D neuronal architecture.
URI: http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/93895
DOI: 10.6342/NTU202402518
Fulltext Rights: 同意授權(全球公開)
Appears in Collections:電子工程學研究所

Files in This Item:
File SizeFormat 
ntu-112-2.pdf6.91 MBAdobe PDFView/Open
Show full item record


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

社群連結
聯絡資訊
10617臺北市大安區羅斯福路四段1號
No.1 Sec.4, Roosevelt Rd., Taipei, Taiwan, R.O.C. 106
Tel: (02)33662353
Email: ntuetds@ntu.edu.tw
意見箱
相關連結
館藏目錄
國內圖書館整合查詢 MetaCat
臺大學術典藏 NTU Scholars
臺大圖書館數位典藏館
本站聲明
© NTU Library All Rights Reserved