利用全域式光學同調斷層掃描術與自編程式進行影像判讀輔助及樣本動態分析

Abstract

本研究利用全域式光學同調斷層掃描術(Full-field optical coherence tomography; FF-OCT)，配合Ce3+:YAG晶體光纖之自發輻射作為系統光源，對巨噬細胞、脂肪間質幹細胞生成之神經球細胞以及誘導型多功能幹細胞生成之類胚胎體等細胞樣本進行掃描和影像建構。透過此系統，我們能夠獲得橫向解析度為0.8 µm，縱向解析度為0.97 µm的橫平面、縱平面以及三維立體影像。 以Ce3+:YAG FF-OCT為基礎，本研究進一步開發動態光學同調斷層掃描術(Dynamic full-field optical coherence tomography; D-FF-OCT)。在D-FF-OCT的掃描過程中，我們將影像擷取平面停留固定於樣本任意深度，並記錄該固定深度下背向散射光訊號隨時間之變化。藉由D-FF-OCT，我們能夠獲得以長度-長度-時間所組成的三維影像，其空間解析度與FF-OCT相同，而時間解析度為7.7 ms。 此外，我們自行撰寫了分析程式，將D-FF-OCT所採集之時域資訊以像素為單位，轉換為頻譜圖，並進行特徵頻率的分析。為了降低影像中的雜訊干擾以提升整體訊噪比，我們在程式中採用縱向一維高斯及橫向二維高斯之卷積平均法對訊號進行預處理。在橫向進行5*5個像素的高斯卷積平均後，系統之橫向解析度變為1.82 µm，而在縱向進行11點高斯卷積平均後，時間解析度則為9.35 ms。 在巨噬細胞的實驗中，透過上色動態影像可有效增強細胞核的對比度，未來可應用於影像判讀的輔助工具。而在神經球細胞和類胚胎體的實驗中，利用上色動態影像，我們能夠針對細胞的動態特性進行分析，並從頻譜圖中區分環境雜訊與動態訊號之差異。 在上色動態影像中，我們設定了低頻閾值與中頻閾值，將所有頻譜圖分為低頻區段、中頻區段及高頻區段。透過閾值的設定，確保三區段之曲線下總面積相等。其中，巨噬細胞的低頻閾值為4.94 ± 0.3 Hz，中頻閾值為12.86 ± 0.53 Hz；神經球細胞的低頻閾值為6.67 ± 0.43 Hz，中頻閾值為14.34 ± 0.24 Hz；類胚胎體的低頻閾值為5.94 ± 0.46 Hz，中頻閾值為14.39 ± 0.37 Hz。透過三種樣本的閾值比較，我們可以推論出相較於巨噬細胞，神經球細胞與類胚胎體具有更多高頻峰值訊號。這些由D-FF-OCT所測得之具空間解析度的動態資訊，未來有機會與細胞生理指標進行比對，以更進一步地揭開細胞的生理運作機制。

In this study, we utilized full-field optical coherence tomography (FF-OCT) combined with spontaneous emission from Ce3+:YAG crystal fiber as the light source to perform scanning and image reconstruction of macrophages, ADMSC-derived neurospheres, and iPSC-derived embryoid bodies. The system achieved high-resolution en face, cross-sectional, and three-dimensional volumetric images with lateral resolution of 0.8 μm and axial resolution of 0.97 μm. Based on Ce3+:YAG FF-OCT, we developed dynamic full-field optical coherence tomography (D-FF-OCT). D-FF-OCT involved fixing the image acquisition plane at a specific depth within the sample and capturing the temporal variation of backscattered light signals. With D-FF-OCT, we obtained three-dimensional images composed of length-length-time dimensions, maintaining the spatial resolution of FF-OCT and achieving a temporal resolution of 7.7 ms. To analyze the D-FF-OCT data, we developed a analysis program that transformed the acquired temporal information into spectrograms on a per-pixel basis. The program included signal pre-processing techniques such as axial one-dimensional Gaussian convolution average and lateral two-dimensional Gaussian convolution average to reduce noise and enhance the signal-to-noise ratio. The temporal resolution resulted in 9.35 ms after applying an 11-point Gaussian convolution average, while the lateral resolution became 1.82 μm through a 5x5-pixel Gaussian convolution average. In the experiments with macrophages, color-coded D-FF-OCT imaging effectively enhanced the contrast of the cell nucleus, providing a potential auxiliary tool for image interpretation. In the experiments with neurospheres and embryoid bodies, color-coded D-FF-OCT imaging allows us to analyze the dynamic characteristics of the cells and differentiate between environmental noise and dynamic signals in the spectrogram. In the color-coded dynamic imaging, we set low-frequency and mid-frequency thresholds to divide all spectrograms into low-frequency, mid-frequency, and high-frequency segments. These thresholds were adjusted to ensure equal total areas under the curve for each segment. Specifically, for macrophages, the low-frequency threshold is 4.94 ± 0.3 Hz, and the mid-frequency threshold is 12.86 ± 0.53 Hz. For neurospheres, the low-frequency threshold is 6.67 ± 0.43 Hz, and the mid-frequency threshold is 14.34 ± 0.24 Hz. For embryoid bodies, the low-frequency threshold is 5.94 ± 0.46 Hz, and the mid-frequency threshold is 14.39 ± 0.37 Hz. By comparing the thresholds among the three samples, we can infer that neurospheres and embryoid bodies exhibit more high-frequency peak signals compared to macrophages.The spatially resolved dynamic information obtained by D-FF-OCT has the potential to be compared with cellular physiological markers in the future, allowing for a deeper understanding of the cellular physiological mechanisms.