基于高速视觉的高通量实时细胞筛选系统的设计与实现

In his study, we plan to realize a high-throughput cell sorting system based on high-speed vision and real-time image processing technology. This research has very important significance for the study of ultra-fast cell sorting, especially for promoting the research of high speed visual feedback control..Our target is to improve the speed of traditional sorting system (≤ 0.01 m/s) of a hundred times. This research will have a great pushing effect on rapid diagnosis of disease and cell printing. To achieve ultra-fast cell sorting system for fast moving cell (≥ 1 m/s), we plan to realize sub-millisecond level visual feedback control and millisecond level sorting actuator. In this study, morphological features of fast moving cells in a microchannel are extracted all by hardware circuits. Adaptive frame-straddling imaging technology is employed for accurate speed measurement of high-speed cell flow. Based on extracted morphological features and measured speed, cell tracking, recognition, and classification are realized all by hardware circuits. Finally, ultra-fast cell sorting is realized by changing the flow direction in microchannel by AC Servo, which is controlled by hardware generated sorting signal from image processing circuits. In this study, we also plan to test our sorting system by using living cells in different size, including fertilized sea urchin eggs, red blood cells (RBCs), and breast cancer cells (MCF-7). The research will greatly accelerate the speed of cellular automation.

本项目依托于高速视觉实时图像处理技术，对高速运动的细胞进行快速特征提取、精准测速和位置预估，实现全硬件高通量实时快速细胞筛选系统。本项目对于推动高速视觉反馈在细胞筛选领域的应用与探索，具有十分重要的意义。.本项目旨在实现超高速细胞流（≥ 1 m/s）的快速识别和筛选，将传统细胞筛选系统的筛选速度（≤ 0.01 m/s）提高百倍，使快速疾病诊断和细胞打印成为可能。本研究完全使用硬件化的算法实时提取超高速微流路中细胞的形态学特征，同时使用自适应帧跨越成像技术对高速细胞流进行实时精确测速，基于细胞形态学特征和细胞速度使用硬件电路对细胞进行追踪识别分类，最后产生硬件级别的筛选控制信号控制AC伺服进行毫秒级别的微流路流向控制，从而实现超高速高通量实时细胞筛选。本研究计划采用不同大小的活体细胞进行筛选测试，包括海胆受精卵、血红细胞、乳癌细胞。本研究将极大的加快细胞自动化的速度。

细胞筛选技术是未来医疗，科研领域的关键技术。未来在细胞打印，快速疾病诊断等领域，细胞筛选将起到至关重要的作用。相比于其他筛选技术，基于视觉的筛选技术可以提供清晰的细胞图片，使复杂的模式识别算法可以用于细胞识别。本项目的研究内容包括高速多物体识别技术，亚毫秒级别视觉反馈技术，高速多目标跟踪识别技术。通过本项目支持，研发团队实现了基于图像的超高速细胞识别技术（12,000 fps），将视觉细胞筛选系统的识别速度提升一个数量级，并大幅提升了识别算法的鲁棒性，使快速疾病诊断和细胞打印成为可能。本研究完全使用硬件化的算法实时提取超高速微流路中细胞的形态学特征，同时使用自适应帧跨越成像技术对高速细胞流进行实时精确测速，基于细胞形态学特征和细胞速度使用硬件电路对细胞进行追踪识别分类，最后产生硬件级别的筛选控制信号控制AC伺服进行毫秒级别的微流路流向控制，从而实现超高速高通量实时细胞筛选。本研究将极大的加快细胞自动化的速度，不仅仅是细胞自动化领域的关键技术，同样也是军事领域和工业领域亟待解决的关键技术。