深度学习驱动的液滴微流控单细胞分选系统

Deep Learning-Actived Sorting System for Droplet Microfluidic Single Cell

单细胞操作和分析对于研究许多基本生物学过程和揭示细胞异质性至关重要,并且在生物医学领域具有巨大的应用潜力。液滴微流控技术在单细胞分析方面具有显著优势。研制了一种深度学习驱动的液滴微流控单细胞分选系统,主要以液滴内所包含的生物样本种类以及数量作为标准分选目标液滴。根据实验需求制作好相应生物样本的数据集,在服务器上训练好对应的网络模型,并将该网络模型转移到NVIDIA Jetson TX2开发板上,利用该网络模型对实验过程中拍摄到的液滴图像进行实时检测判断,最后根据算法对包含特定物质的液滴进行分选,从而得到目标液滴。此方法能够有效地判断并分选出液滴内图像特征有差异的不同生物样本,可以实现对包含单个及2个细胞液滴的分选。该研究为液滴微流控单细胞分选技术在生物学和医学等领域的广泛应用提供了支撑。

Manipulation and analysis of the single cell are crucial for the research of many basic biological processes and revelation of cell heterogeneity,and have great application potential in the field of biomedicine.The droplet microfluidic technology has significant advantages in single cell analysis.A sorting system for droplet microfluidic single cell actived by deep learning was developed,which mainly sorted target droplets according to the type and quantity of biological samples involved in the droplets.The corresponding network model was trained by creating a dataset of corresponding biological samples according to experimental requirements.And the network model was transferred to the NVIDIA Jetson TX2 development board.By using the network model,real-time detection and judgement of the captured droplet images during the experimental process was performed.Finally,the droplets containing specific substances were sorted according to the algorithm,therefor the target droplets were obtained.The method can effectively judge and sort different biological samples with different image features in the droplets,and the sorting of droplets containing single cell and two cells can be realized.The study provides support for the wide application of droplet microfluidic single cell sorting technology in biology and medicine fields.