偵測三維共軛焦顯微影像內小鼠胚胎幹細胞中心的深度學習方法

Deep learning method for detecting centers of mouse embryonic stem cells in 3D confocal microscopy images

馬國禎(GUO-ZHEN MA)

指導教授 ： 朱守禮 ； 蔡明達

中原大學/電機資訊學院/資訊工程學系/碩士(2022年)

https://doi.org/10.6840/cycu202201701

摘要

小鼠胚胎幹細胞(mouse embryonic stem cells, mESc)在培養過程中會多次分裂並互相擠壓位移，導致在判斷細胞數量及位置上會有困難，為了增進後續分析mESc之準確率，一個能將接觸中細胞辨識並分割的方法會有相當大的助益。

本研究利用深度學習方法，對共軛焦顯微鏡影像，使用兩種不同結構的U-Net進行分析。分別由二維影像取得細胞邊界，接著由三維影像取得細胞中心。最後將基於這兩個模型所提供的細胞中心與細胞區域的結果，分割接觸細胞。

應用於細胞數較多的影像時，本研究所提出的方法能準確地分割細胞，有助於在分析胚胎細胞時，量化其細胞資訊以及細胞分化之狀態。

關鍵字

U-Net ； 深度學習 ； 三維分割 ； 細胞中心偵測 ； 小鼠胚胎幹細胞

並列摘要

Mouse embryonic stem cells (mESc) would divide multiple times and squeeze each other during the culture process, making it difficult to determine the number and location of cells. In order to improve the accuracy of the analysis of mESc, a method that identifies and segments the contacted cells would be of great benefit.

In this study, two different structures of U-Net are adopted to analyze the confocal microscope images by using deep learning methods. The cell boundaries are first obtained from two-dimensional images, then the cell centers are obtained from three-dimensional images. Finally, the contacted cells are separated, based on the cell centers and the cell regions that are provided by the two models.

When applied to images with a large number of cells, the proposed method can accurately segment the cells and help to quantify the cellular information and cell differentiation for the analysis of embryonic cells.

並列關鍵字

U-Net ； Deep Learning ； 3D Segmentation ； Cell Center Detection ； Mouse Embryonic Stem Cells

參考文獻

[1] G. Pan and J. A. Thomson, Nanog and transcriptional networks in embryonic stem cell pluripotency, vol. 17, Nature Publishing Group, 2007, pp. 42-49.

[2] O. Ronneberger, P. Fischer and T. Brox, U-net: Convolutional networks for biomedical image segmentation, 2015, pp. 234-241.

[3] Ö. Çiçek, A. Abdulkadir, S. S. Lienkamp, T. Brox and O. Ronneberger, 3D U-Net: learning dense volumetric segmentation from sparse annotation, 2016, pp. 424-432.

[4] Y. Tokuoka, T. G. Yamada, D. Mashiko, Z. Ikeda, N. F. Hiroi, T. J. Kobayashi, K. Yamagata and A. Funahashi, "3D convolutional neural networks-based segmentation to acquire quantitative criteria of the nucleus during mouse embryogenesis.," npj Syst Biol Appl 6, 32, 2020.

[5] A. Sachinidis, B. K. Fleischmann, E. Kolossov, M. Wartenberg, H. Sauer and J. Hescheler, Cardiac specific differentiation of mouse embryonic stem cells, vol. 58, Elsevier Science, 2003, pp. 278-291.