Mirau 全域式光學同調斷層掃描術應用於體外黑色素細胞及黑色素癌細胞之分析與判別

Abstract

Mirau全域式光學同調斷層掃描儀(FF-OCT)為具有非侵入式、樣本無須標定及三維空間解析度佳等特性，是現今生醫領域中重要的三維成像技術之一，可應用於體外單細胞之研究或活體疾病診斷之研究。本研究透過此技術，取得黑色素細胞及黑色素癌細胞之三維影像資訊，這些資訊代表著生物體的散射特性，經計算萃取出影像特徵參數，並透過線性判別分析，進行黑色素正常細胞與黑色素癌細胞之判別研究。 實驗規劃分成兩批次進行，第一批實驗設計上，希望判別癌細胞珠(A2058)與含有黑色素的初代細胞；第二批實驗細胞皆為缺乏黑色素的細胞，依照癌細胞之惡性度高低又細分成兩種樣本，A2058及A375，與缺乏黑色素的正常細胞，共有三種樣本。由實驗結果可知，截面訊號相關之特徵參數與生物體是否含有黑色素這類強吸收物質有關聯，第一批細胞間存在著黑色素含量上的差異，透過截面訊號平均及截面訊號標準差進行判別分析，正確分群率為 ，第二批細胞間無黑色素含量上的差異，則是透過訊號整體標準差與訊號整體偏度進行正常細胞與癌細胞之判別分析，正確分群率為 ，推測造成細胞間的差異是來自於癌細胞中粒線體的變異性。 除上述分析之外，將第一批實驗的細胞三維影像資訊進行影像分割處理，獲得三維狀態下的形態特徵，細胞核質體積比，黑色素細胞的平均細胞核質比約為0.15，黑色素癌細胞之平均細胞核質比約為0.22，經分析後發現，於此參數下黑色素正常細胞及其癌細胞具有顯著性差異(P < 0.05)。 本研究使用非侵入式的Mirau全域式光學同調斷層掃描儀，擷取無須標定的體外單細胞三維影像資訊，透過影像特徵參數的萃取及分析，獲得具有判別黑色素正常細胞及黑色素癌細胞的特徵參數資訊，正確分群率達九成以上，此一結果顯示本研究之光學系統及分析方式，極具潛力進行黑色素癌症之活體皮膚組織的早期診斷研究。

Mirau-based full-field optical coherent tomography (Mirau FF-OCT) is a three-dimensional (3D) optical imaging system that has non-invasive, label-free and high spatial resolution characteristics. It is an important 3D imaging technique in biomedical realm. It can be applied to study single cell in vitro and disease diagnosis in vivo. In this study, we have obtained the melanocyte and the melanoma 3D OCT images by using this technique. The images show the scattering property of organisms. To quantitatively the images, features were extracted from the images and linear discriminant analysis (LDA) was adopted to discriminate the melanocyte and the melanoma. Two batches of experiments were conducted. In the first experiment, the single cells we used were the melanoma cell line, A2058, and melanocyte primary cells with melanin. In the second experiment, we used two kinds of melanoma cell line, A2058 and A375, and Caucasian melanocyte primary cells. The experimental results show the correlation between en face features and absorption ingredients, such as melanin. Among the cells of the first batch, there is difference in the quantity of melanin. We employ the en face features, average of en face and standard deviation of en face, to do LDA and the result is 96.9 %. For the second batch of experiments, the cells had rare melanin. We employ the discriminating features, signal standard deviation and signal skewness, to do LDA to discriminate melanocyte and melanoma and the result is 90.7 %. By inference, the difference between the melanocyte and the melanoma might due to the variation of mitochondria. Furthermore, we utilized the image segmentation process on the first batch of data. The 3D cell morphology and N:C ratio (N/C) were acquired. The average N/C of melanocyte is 0.15 and of the melanoma is 0.22. After ANOVA analysis, the N/C of melanoma is significantly higher than melanocyte (P<0.05). In this study, by utilizing the non-invasive Mirau FF-OCT, we obtain the label-free 3D single cell images in vitro. After extracting the features of image and analysis by LDA, we acquire the discriminating features with bioinformation to discriminate the melanocyte and melanoma. The discriminant result is higher than 90 %. These demonstrate our optical system and analytic method have great potential to study the early diagnosis of melanoma in vivo.