Deep learning has achieved spectacular performance in image and speech recognition and synthesis. It outperforms other machine learning algorithms in problems where large amounts of data are available. In the area of measurement technology, instruments based on the photonic time stretch have established record real-time measurement throughput in spectroscopy, optical coherence tomography, and imaging flow cytometry. These extreme-throughput instruments generate approximately 1 Tbit/s of continuous measurement data and have led to the discovery of rare phenomena in nonlinear and complex systems as well as new types of biomedical instruments. Owing to the abundance of data they generate, time-stretch instruments are a natural fit to deep learning classification. Previously we had shown that high-throughput label-free cell classification with high accuracy can be achieved through a combination of time-stretch microscopy, image processing and feature extraction, followed by deep learning for finding cancer cells in the blood. Such a technology holds promise for early detection of primary cancer or metastasis. Here we describe a new deep learning pipeline, which entirely avoids the slow and computationally costly signal processing and feature extraction steps by a convolutional neural network that directly operates on the measured signals. The improvement in computational efficiency enables low-latency inference and makes this pipeline suitable for cell sorting via deep learning. Our neural network takes less than a few milliseconds to classify the cells, fast enough to provide a decision to a cell sorter for real-time separation of individual target cells. We demonstrate the applicability of our new method in the classification of OT-II white blood cells and SW-480 epithelial cancer cells with more than 95% accuracy in a label-free fashion.
Cameraless High-throughput 3D Imaging Flow Cytometry
对理解细胞表型的巨大异质性的日益增长的需求推动了成像流式细胞术（IFC）的发展，其将流式细胞术的特征与荧光和明场显微镜相结合。IFC将流式细胞术的通量和统计优势与基于真实或计算图像以及常规流式细胞术指标离散测量事件的能力相结合。现有IFC系统的局限性在于，不管检测方法如何，仅获得二维（2D）细胞图像。在没有断层扫描三维（3D）分辨率的情况下，投影问题仍然存在：将3D信息折叠到2D图像上，限制了对细胞表型至关重要的点计数或共定位的可靠性。在这里，我们提出了一个解决方案的投影问题：三维成像流式细胞术（3D-IFC），一个高通量的三维细胞成像仪的基础上光学切片显微镜。我们结合正交光片扫描照明与我们以前的时空变换检测产生3D细胞图像重建从一个无相机的单像素光电探测器读出。我们进一步证明了这种能力，通过共同捕获的3D荧光和无标记的侧散射图像的单细胞在流中的速度为0.2 m s-1，对应于约500个细胞每秒的吞吐量与60,000体素（随后调整大小为106体素）为每个细胞图像在小于1微米的分辨率在X，Y和Z尺寸。需要改进的高通量成像工具来在免疫学、肿瘤学、细胞和基因治疗以及药物发现领域中识别表型-基因型异质性。
Automated Identification of Cell Populations in Flow Cytometry Data with Transformers
TU Wien（维也纳工业大学）；St Anna Children’s Cancer Research Institute（圣安娜儿童癌症研究所）；Hospital de Pediatria “Dr. Juan P. Garrahan”（胡安·P·加拉汉博士儿科医院）；Charité Universitätsmedizin Berlin（柏林夏里特医学院）；HELIOS Klinikum Berlin-Buch（柏林布赫赫利俄斯医院）
Connecting the dots across time: Reconstruction of single cell signaling trajectories using time-stamped data
Battelle Center for Mathematical Medicine, Research Institute at the Nationwide Children's Hospital（全国儿童医院研究所巴特尔数学医学中心）；The Ohio State University（俄亥俄州立大学）；University of Iowa（爱荷华大学）；University of California, San Francisco（加利福尼亚大学旧金山分校）；Institute of Bioinformatics and Applied Biotechnology（生物信息学与应用生物技术研究所）
Pose-Free 3D Quantitative Phase Imaging of Flowing Cellular Populations
1 School of Instrumentation Science \(\&\) Optoelectronics Engineering, Beihang University, Beijing, 2 College of Future Technology, Peking University, Beijing, China 3 National Biomedical Imaging Center, Peking University, Beijing, China 4 Department of Obstetrics；Gynecology, Peking University People’s Hospital, Beijing, China 5 Peking University Third Hospital, Department of Radiation Oncology, Beijing, China. 6 Peking University Third Hospital, Cancer Center, Beijing, China The authors contributed equally to this work. The corresponding authors ( , ).
Deep learning guided image-based droplet sorting for on-demand selection and analysis of single cells and 3D cell cultures
揭示细胞群体的异质性是从抗菌素耐药性到癌症研究等领域的长期目标。液滴微流体等新兴技术平台有望以超高通量破译细胞异质性。然而，缺乏能够快速鉴定和分离单细胞或3D细胞培养物的方法。在这里，我们证明了深度神经网络可以根据包括单个哺乳动物细胞和多细胞球体在内的微物体的存在和数量，实时准确地对单个液滴图像进行分类。这种方法还能够在不同类型和大小的物体混合物中识别特定物体。神经网络的训练集由数百张手动挑选的图像组成，每个训练类的图像数量增加到数千张。使用单个GPU进行训练所需时间不到10分钟，并且单个哺乳动物细胞识别的准确率超过90%。至关重要的是，相同的模型可以用于对不同类型的物体进行分类，例如聚苯乙烯球、聚丙烯酰胺珠和MCF-7细胞。我们将开发的方法应用于选择用包封在琼脂糖凝胶珠中的Hek 293 FT细胞生成的3D细胞培养物，突出了该技术用于选择具有高度多样性的视觉外观的对象的潜力。单个液滴的实时分选与液滴生成一致，并且以高达每秒40个的速率发生，与高达480 X 480像素的图像尺寸无关。所提出的微流体装置还能够存储分选的液滴以允许下游分析。
Inherited or produced? Inferring protein production kinetics when protein counts are shaped by a cell's division history
Center for Biological Physics, Arizona State University（亚利桑那州立大学生物物理中心）；Department of Physics, Arizona State University（亚利桑那州立大学物理系）；School of Molecular Sciences, Arizona State University（亚利桑那州立大学分子科学学院）
