高通量計算顯微影像系統之研製於生物醫學成像與分析;Research and Development of High-throughput Computational Micro-imaging System for Biomedical Imaging and Analysis

NCU Institutional Repository > 生醫理工學院 > 生物醫學工程研究所 > 博碩士論文 > Item 987654321/89276

請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/89276

題名:	高通量計算顯微影像系統之研製於生物醫學成像與分析;Research and Development of High-throughput Computational Micro-imaging System for Biomedical Imaging and Analysis
作者:	楊奕傑;Yang, Yi-Jie
貢獻者:	生物醫學工程研究所
關鍵詞:	顯微影像;計算成像系統;傅立葉疊層成像;深度學習;Micro-imaging;Computational Imaging System;Fourier Ptychography;Deep Learning
日期:	2022-08-03
上傳時間:	2022-10-04 11:08:36 (UTC+8)
出版者:	國立中央大學
摘要:	現行光學顯微影像系統在成像上由於受限於空間帶寬乘積(Space-Bandwidth Product) 效應，因此影像解析度與成像視野範圍間相互受到限制，在影像取得上可獲得具有精細但視野區域有限或是具有廣區域視野但較差解析度之影像，無法同時取得具高解析度與廣視野範圍影像，為此通常需要裝載高倍率顯微鏡與機械掃描裝置組合構建，不僅增加系統的複雜性和成本，同時限制圖像採集速度，大幅度的限制場域之使用。本研究建立出計算顯微成像系統，成功降低了光學成像中空間帶寬乘積之限制，以傅立葉疊層圖像(Fourier Ptychography, FP)為基礎，藉由編成激發光陣列以此提供多重角度之光源，並經由影像感測裝置擷取多重空間角頻率之光場訊息，透過疊代傅立葉頻域之演算過程經收斂後重建出同時具高解析度與廣視場之高通量顯微影像。並且更進一步的，使用深度學習之技術，基於編碼與解碼(Encoder-Decoder)框架下結合平行式模型架構，將原本傅立葉疊層圖像技術對於重建影像之大量低解析度測量影像之需求，簡化為採集單張圖片即可完成，解決了其所造成大量時間花費於收集重建之所需影像上之冗長影像擷取與重建過程之時間與過程。本研究論文成功建立可同時獲得高解析度與廣視野範圍影像之計算成像系統，在生醫相關顯微成像之應用上，有利取得於多種生理條件與環境下之巨量的細胞結構以及其行為動力學的時間與空間資訊，達到無偏見及可系統化的生物過程並進行更進一步之分析與評估。 ;Due to the space-bandwidth product effect in the imaging of optical microscope systems, the image resolution and the field of view (FOV) restrict each other. Hence, it is impossible to obtain images with high resolution and wide FOV simultaneously. For this reason, a high magnification objective lens and a mechanical scanning device are usually needed. This increases the complexity and cost of the system, slows down the image acquisition speed, and greatly limits the actual space available for the object. In this study, we established a computational microscopy imaging system. Based on Fourier Ptychography (FP) technology, the excitation light array was arranged to provide light sources with multiple angles. The image sensor captures the light information of multiple angular spatial frequencies. Through converging the information with an iterative computational process, images with high resolution and wide FOV are reconstructed. Furthermore, deep learning based on an encoder-decoder framework combined with a parallel model architecture can reconstruct using a single image. Leaving out the original redundant stack of low-resolution images reduces the time spent on image acquisition and reconstruction. This research has successfully established a computational imaging system that can automatically obtain high-resolution and wide-field images. When applied to microscopic biomedical imaging, this technique is beneficial for visualizing subtle cell structures and their dynamic behavior under various physiological conditions and environments. In addition, achieving unbiased and systematized biological processes for further analysis and evaluation.
顯示於類別:	[生物醫學工程研究所 ] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	62	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....