“What is this good for,” was the comment of our clinical cooperation partner, when we showed him the first imaging flow cytometry results using a microfluidic system and a customized quantitative phase microscope for high‐throughput blood cell analysis. What went wrong? We have shown the hematology‐oncology expert phase images of various leukocyte types. Our eyes were trained on these images showing round‐shaped cells of different sizes with rather low‐contrast but we could interpret the image quality and even differentiate leukocyte types. The clinician, however, looked at blood cell images he was never trained on and expected contrast similar to stained blood smear. Our first disappointment vanished when we showed him dot plots from image analysis, which pleased him as he started to become familiar due to his own expertise in flow cytometry. Showing him additional results on the discrimination potential of leukemia samples finally gave us a mark that we are looking at something with clinical relevance (1, 2). This meeting was a really good experience for the team working on phase imaging flow cytometry for next‐generation hematology analyzer and showed us again that expectations can differ dramatically between an engineer and a clinical user. The reader might ask himself, why is label‐free cell imaging of interest in hematology offering only low‐contrast images.

When we look into the field of flow cytometry, hematology analyzer are somewhat the automated high‐throughput beasts for blood cell analysis requiring only nonspecific staining to discriminate leukocyte or more challenging analytes, such as reticulocytes. The high‐throughput for a complete blood count (CBC) and a leukocyte differential (Diff) as well as the deep hematological information make these tools unique in the world of in vitro diagnostics (IVD). With low costs per test and high statistical power for blood cell biomarkers, a CBC/Diff is today one of the most requested clinical tests in the world and literally every patient receives a CBC/Diff. This achievement was the result of a few elegant technical solutions in the past starting with the Coulter counter principle in the early days and later light scatter analysis, which allows discriminating monocytes, lymphocytes, and different granulocytes by means of size and morphology without requiring manual microscopy of blood smears. Contrary to fluorescence flow cytometry (FCM) no specific and costly antibody labeling is required, which is key for integration, standardization, and robustness of the flow cytometry workflow. Even today, the unique automation level in flow cytometry and the wealth of biomarker information from a single IVD instrument are outstanding in a very mature central laboratory market. The defined CBC/Diff is also a much easier task for IVD certification compared to the numerous FCM assays. However, the hidden champion is—as it is very often in the IVD space—the magic chemistry of the reagents for hematology analyzer. Only with the right chemistry on board analyzer can achieve standardized sample preparation in seconds, such as the sphering of blood cells, erythrocyte lysis, the granulocyte differentiation, the extremely accurate high‐throughput counting, and even the hemoglobin concentration measurement on single erythrocyte level (3, 4). These developments make hematology analyzer after decades of usage in the clinical routine even today an amazing piece of engineering and still an active field of research (5, 6).

The caveat of the success story is the sample preparation limiting the morphological information for instance due to sphering, leukocyte artifacts from staining. With increasing number of reagents on board, the fluidics become complex as well as the maintenance effort to operate the analyzer. Such systems work effectively in central laboratory environments but are not suitable for point‐of‐care testing (POCT). In addition, the image information is missing and only indirect cell morphology information is obtained. Today, new blood cell biomarker classes, such as highly dilute circulating tumor cells (CTCs, 7) or blood cell aggregates (8), covering a large dynamic concentration range became of clinical interest. However, due to the indirect cell analysis, sample preparation limitations, and fluidic conditions, many new cellular biomarkers cannot be covered by conventional hematology analyzer depending on fixed sample volumes requiring medium to high target cell concentrations and rapid processing cycles to achieve economical high‐throughput. Moreover, in the clinical routine, only a few of the provided biomarker parameters of a CBC/Diff are used for routine analysis which is in reality too complex for routine diagnosis. Such market conditions are generally unfavorable to develop new technologies. Due to the plethora of data provided by these systems, multivariate data analysis and artificial intelligence algorithms are applied to reduce the effort for data interpretation and to fully exploit the analytical power of analyzer for differential diagnosis. This data mining can lead to unexpected results as we have shown for malaria testing (9).

Most research on new blood cell biomarkers, such as CTCs or cell aggregates, is performed today with FCM requiring multidimensional immunophenotyping for preclinical research, which is not matching the requirements for economical cell analysis for the clinical routine. Imaging flow cytometer, such as the famous Amnis system, has only found niche applications in the life sciences and preclinical research (10). One could assume that the current flow cytometry trends are representative for a “retreat” from morphology information due to the dominating high‐dimensional immunophenotyping methods. But none of these methods is compatible with the brutal economics of the IVD market and in particular for hematology analysis. This is also true for automated blood smear analysis, which still is considered the gold standard in hematology analysis. Even sophisticated automated microscopy systems with integrated sample preparation, such as the Bloodhound, can barely compete against the efficiency of hematology analyzer (11). Most recently, the integration of advanced microfluidic solutions with viscoelastic focusing and image analysis allows the Hemoscreen from Pixcell Medical to achieve clinically comparable CBC/Diff quality for POCT (12). The main advantage of using imager is the parallelized imaging flow cytometry of blood cells in a highly defined focal height, which allows researcher to decouple from serial cell analysis and compensates for usually low flow rates to achieve sufficient statistical power. However, the sample preparation and reagents remain essentially the same as in today's analyzer of the central laboratories.

Quantitative phase imaging (QPI) is a label‐free imaging opportunity to replace sample preparation in hematology analysis (13). Instead of measuring refractive index changes within and around the suspended blood cells by scatter analysis, phase images are acquired, which allow to visualize the individual cell with integrated phase information on an imager. Knowing the phase conditions quantitatively allows to reconstruct whole cells or respective slices depending on the optical measurement conditions (Fig. 1A). In this way, the morphological information of cells is preserved. With appropriate imaging conditions hematology analysis can now be performed in principle without any sample preparation reducing the fluidic complexity of automated analyzer but gaining access to nonstable biomarkers, which are not accessible with today's workflow, such as blood cell aggregates (Fig. 1B). In other words, in silico image analysis replaces chemical sample preparation, which distorts the cellular morphology (14). One example is the opportunity to resolve label‐free the Plasmodium falciparum life cycle in erythrocytes including the clinically relevant ring‐stage (15). With high statistical power and access to the life cycle of parasites, one could imagine that future hematology analyzer can even support to monitor the efficacy of treatments and identify resistances beyond the limits of costly molecular analysis. However, we have to admit that QPI will be insufficient to replace a regular CBC/Diff. We need to find and integrate additional optical solutions for reticulocytes and measuring effectively the hemoglobin concentration in an erythrocyte to match the state‐of‐the‐art biomarker panel to claim a CBC/Diff by physical methods only. The less elegant alternatives would be a mix of partial sample preparation and phase imaging solution or an analyzer for dedicated new hematological biomarkers. One critical piece of the puzzle for a next‐generation hematology analyzer is microfluidics. Only with appropriate solutions to enrich rare cellular analytes at low shearing conditions and to deplete erythrocytes, we can cover the dynamic concentration range of new biomarkers and minimize image analysis effort for redundant information. Storing Gigabytes of images data for offline analysis is only an option for research. The last piece of the puzzle is, therefore, real‐time image analysis and the respective image analysis algorithms. Due to this dependence, future imaging‐based hematology analyzer will be a partial digital healthcare product.

Digital holographic microscopy (DHM) is an interferometry‐based variant of QPI that typically uses the classic holographic principle, with the difference that the hologram recording is performed by a digital image sensor using a coherent light source (16). Although a rather old methodology, DHM has taken off only recently due to the computational power for image reconstruction and robust “off‐axis” imaging tools replacing classical interferometer setups. With tomographic information, one can resolve amazing details of cellular compartments (17-19), but for clinical high‐throughput operations a single image acquisition of blood cells is the only option to match today's hematology analyzer. With DHM, we have an attractive platform technology for the field of hematology analysis and parallelized analysis of >3,000 cell/s at 100 frames per second (1). With full access to undistorted morphological information and low shear stress, we can visualize and quantify the morphology of erythrocytes and platelets in plasma and we can resolve the differences in granularity and nuclei of leukocytes replacing forward and side scatter analysis. Second, we can visualize fragile megakaryocytes and observe cell–cell interactions in blood without depending on harsh sample preparation conditions, which interfere with these logistically nonstable biomarkers. Third, being potentially independent from sample preparation, we are not bound to a defined blood sample volume and the respective statistical power for a given cell concentration. Clinical users could run hematology analyzer by simply defining the statistical power they want to achieve rather than looking at a Diff from a few microliters of blood only. Last, with the appropriate cell enrichment tools, we can potentially look at any concentration range covering even very dilute biomarkers, such as CTCs, from even several milliliters of blood. The group of Natan Shaked at Tel Aviv University (Cytometry A. 2020 Sep 10. doi: 10.1002/cyto.a.24227. Online ahead of print) reports in this issue a proof‐of‐concept of rare CTCs detected with DHM. Future clinical studies will show if the cellular phase contrast allows sufficient sensitivity and specificity for CTCs. In this way, a new generation of imaging‐based hematology analyzer could cover even liquid biopsy marker.

Much more work is ahead of us to create smart and integratable workflows to match the accuracy of conventional hematology analyzer and to add new, derisked hematology biomarkers for the clinical routine. To achieve this goal, we need interdisciplinary work between engineers and clinicians requiring a deep understanding of biomarkers, clinical workflows, and technology to avoid surprises for translation. With new biomarkers of clinical relevance, DHM and additional optical solutions could outperform today's analyzer and disrupt both the clinical and POC market in a similar way as was the introduction of the platelet count in hematology analyzers in the past.

“什么帽子这对我们有好处吗？”这是我们的临床合作伙伴的评论，当时我们向他展示了使用微流体系统和定制的定量相位显微镜进行高通量血细胞分析的第一个成像流式细胞术结果。什么地方出了错？我们已经展示了各种白细胞类型的血液肿瘤学专家阶段图像。我们的眼睛接受了这些图像的训练，这些图像显示了不同大小的圆形细胞，对比度相当低，但我们可以解释图像质量，甚至区分白细胞类型。然而，临床医生查看了他从未接受过训练的血细胞图像，并期望对比度类似于染色的血涂片。当我们向他展示图像分析的点阵图时，我们的第一个失望消失了，这让他很高兴，因为他自己在流式细胞术方面的专业知识开始变得熟悉。1、2）。这次会议对于下一代血液分析仪相位成像流式细胞术的团队来说是一次非常好的经历，并再次向我们展示了工程师和临床用户之间的期望可能会有很大差异。读者可能会问自己，为什么血液学中感兴趣的无标记细胞成像仅提供低对比度图像。

当我们研究流式细胞术领域时，血液分析仪在某种程度上是用于血细胞分析的自动化高通量野兽，只需要非特异性染色即可区分白细胞或更具挑战性的分析物，例如网织红细胞。全血细胞计数 (CBC) 和白细胞分类 (Diff) 的高通量以及深入的血液学信息使这些工具在体外诊断 (IVD) 领域独树一帜。由于每次测试的成本低，血细胞生物标志物的统计能力高，CBC/Diff 是当今世界上最受欢迎的临床测试之一，实际上每个患者都会收到 CBC/Diff。这一成就是过去一些优雅的技术解决方案的结果，从早期的库尔特计数器原理到后来的光散射分析，它允许通过大小和形态区分单核细胞、淋巴细胞和不同的粒细胞，而无需手动显微镜检查血涂片。与荧光流式细胞术 (FCM) 不同，不需要特定且昂贵的抗体标记，这是流式细胞术工作流程的集成、标准化和稳健性的关键。即使在今天，流式细胞术独特的自动化水平和来自单个 IVD 仪器的丰富生物标志物信息在非常成熟的中心实验室市场中也非常出色。与众多 FCM 检测相比，定义的 CBC/Diff 也是 IVD 认证的一项简单得多的任务。然而，隐形冠军是血液分析仪试剂的神奇化学——正如它在体外诊断领域中经常出现的那样。3、4）。这些发展使血液学分析仪在临床常规中使用了数十年，即使在今天也是一项了不起的工程，并且仍然是一个活跃的研究领域 ( 5, 6 )。

成功故事的警告是样品制备限制了形态信息，例如由于球状、染色造成的白细胞伪影。随着船上试剂数量的增加，射流变得复杂，操作分析仪的维护工作也变得复杂。此类系统在中心实验室环境中有效工作，但不适用于即时检测 (POCT)。此外，缺少图像信息，只能获得间接的细胞形态信息。今天，新的血细胞生物标志物类别，例如高度稀释的循环肿瘤细胞（CTC，7）或血细胞聚集体（8)，覆盖大的动态浓度范围成为临床兴趣。然而，由于间接细胞分析、样品制备限制和流体条件，许多新的细胞生物标志物不能被传统的血液学分析仪覆盖，这取决于需要中高靶细胞浓度和快速处理周期的固定样品体积，以实现经济的高通量. 此外，在临床常规中，只有少数提供的 CBC/Diff 生物标志物参数用于常规分析，这实际上对于常规诊断来说过于复杂。这种市场条件一般不利于开发新技术。由于这些系统提供的大量数据，应用多元数据分析和人工智能算法，减少数据解释的工作量，充分发挥分析仪的分析能力进行鉴别诊断。这种数据挖掘可能会导致意想不到的结果，正如我们在疟疾测试中所展示的（9）。

目前，大多数关于新血细胞生物标志物（如 CTC 或细胞聚集体）的研究都是使用 FCM 进行的，需要多维免疫表型进行临床前研究，这不符合临床常规经济细胞分析的要求。成像流式细胞仪，例如著名的 Amnis 系统，只在生命科学和临床前研究中找到了利基应用（10）。由于高维免疫表型方法占主导地位，人们可以假设当前的流式细胞术趋势代表了形态学信息的“撤退”。但这些方法都与 IVD 市场的残酷经济学不相容，尤其是血液学分析。对于自动血液涂片分析也是如此，它仍然被认为是血液学分析的金标准。即使是具有集成样品制备功能的复杂自动显微镜系统，例如 Bloodhound，也几乎无法与血液分析仪的效率相媲美 ( 11）。最近，先进的微流体解决方案与粘弹性聚焦和图像分析的集成使 Pixcell Medical 的 Hemoscreen 能够实现临床上可比的 POCT CBC/Diff 质量 ( 12 )。使用成像仪的主要优势是在高度定义的焦高处对血细胞进行并行成像流式细胞术，这使研究人员能够从串行细胞分析中分离出来，并补偿通常的低流速以获得足够的统计能力。然而，样品制备和试剂与今天的中央实验室分析仪基本相同。

定量相位成像 (QPI) 是一种无标记成像机会，可替代血液学分析中的样品制备 ( 13）。不是通过散射分析测量悬浮血细胞内部和周围的折射率变化，而是获取相位图像，这允许在成像仪上显示具有集成相位信息的单个细胞。定量地了解相位条件允许根据光学测量条件重建整个细胞或相应的切片（图 1A）。这样，细胞的形态信息得以保留。在适当的成像条件下，原则上现在可以进行血液学分析，无需任何样品制备，降低了自动化分析仪的流体复杂性，但可以获得不稳定的生物标志物，这些生物标志物在当今的工作流程中是无法获得的，例如血细胞聚集体（图 1B）。换句话说，计算机图像分析取代了化学样品制备，14 )。一个例子是解决红细胞中无标记恶性疟原虫生命周期的机会，包括临床相关的环形阶段（15）。凭借强大的统计能力和对寄生虫生命周期的访问，人们可以想象未来的血液分析仪甚至可以支持监测治疗效果并识别超出昂贵分子分析限制的耐药性。但是，我们不得不承认 QPI 不足以取代常规的 CBC/Diff。我们需要为网织红细胞找到并整合额外的光学解决方案，并有效测量红细胞中的血红蛋白浓度，以匹配最先进的生物标志物面板，以便仅通过物理方法声明 CBC/Diff。不太优雅的替代方案是部分样品制备和相位成像解决方案的组合，或用于专用新血液学生物标志物的分析仪。下一代血液分析仪的一个关键难题是微流体。只有采用适当的解决方案在低剪切条件下丰富稀有细胞分析物并消耗红细胞，我们才能涵盖新生物标志物的动态浓度范围，并最大限度地减少冗余信息的图像分析工作。为离线分析存储千兆字节的图像数据只是研究的一种选择。因此，最后一块拼图是实时图像分析和相应的图像分析算法。由于这种依赖性，未来基于成像的血液分析仪将成为部分数字化的医疗保健产品。为离线分析存储千兆字节的图像数据只是研究的一种选择。因此，最后一块拼图是实时图像分析和相应的图像分析算法。由于这种依赖性，未来基于成像的血液分析仪将成为部分数字化的医疗保健产品。为离线分析存储千兆字节的图像数据只是研究的一种选择。因此，最后一块拼图是实时图像分析和相应的图像分析算法。由于这种依赖性，未来基于成像的血液分析仪将成为部分数字化的医疗保健产品。

数字全息显微镜 (DHM) 是一种基于干涉测量的 QPI 变体，它通常使用经典的全息原理，不同之处在于全息记录是由数字图像传感器使用相干光源执行的 ( 16 )。尽管是一种相当古老的方法，但由于图像重建的计算能力和强大的“离轴”成像工具取代了经典的干涉仪设置，DHM 直到最近才开始流行。借助断层扫描信息，人们可以解析细胞室的惊人细节 ( 17-19)，但对于临床高通量操作，血细胞的单个图像采集是匹配当今血液学分析仪的唯一选择。借助 DHM，我们为血液学分析和并行分析领域提供了极具吸引力的平台技术，以每秒 100 帧 ( 1）。通过完全访问未失真的形态信息和低剪切应力，我们可以可视化和量化血浆中红细胞和血小板的形态，我们可以解决白细胞粒度和细胞核的差异，取代前向和侧向散射分析。其次，我们可以可视化脆弱的巨核细胞并观察血液中的细胞间相互作用，而无需依赖苛刻的样品制备条件，这会干扰这些逻辑不稳定的生物标志物。第三，由于可能独立于样品制备，我们不受限定的血样体积和给定细胞浓度的相应统计功效的约束。临床用户可以通过简单地定义他们想要达到的统计功效来运行血液分析仪，而不是仅仅从几微升的血液中查看 Diff。最后，使用适当的细胞富集工具，我们可以从甚至几毫升的血液中查看任何浓度范围，甚至涵盖非常稀的生物标志物，例如 CTC。特拉维夫大学的 Natan Shaked 小组（Cytometry A. 2020 Sep 10. doi: 10.1002/cyto.a.24227. Online before print）在本期中报告了用 DHM 检测到的罕见 CTC 的概念验证。未来的临床研究将显示细胞相位对比是否允许对 CTC 具有足够的敏感性和特异性。这样，新一代基于成像的血液分析仪甚至可以覆盖液体活检标志物。特拉维夫大学的 Natan Shaked 小组（Cytometry A. 2020 Sep 10. doi: 10.1002/cyto.a.24227. Online before print）在本期中报告了用 DHM 检测到的罕见 CTC 的概念验证。未来的临床研究将显示细胞相位对比是否允许对 CTC 具有足够的敏感性和特异性。这样，新一代基于成像的血液分析仪甚至可以覆盖液体活检标志物。特拉维夫大学的 Natan Shaked 小组（Cytometry A. 2020 Sep 10. doi: 10.1002/cyto.a.24227. Online before print）在本期中报告了用 DHM 检测到的罕见 CTC 的概念验证。未来的临床研究将显示细胞相位对比是否允许对 CTC 具有足够的敏感性和特异性。这样，新一代基于成像的血液分析仪甚至可以覆盖液体活检标志物。

我们还有更多的工作要做，以创建智能和可集成的工作流程，以匹配传统血液分析仪的准确性，并为临床常规添加新的、去风险的血液生物标志物。为了实现这一目标，我们需要工程师和临床医生之间的跨学科合作，需要深入了解生物标志物、临床工作流程和技术，以避免翻译出现意外。凭借具有临床相关性的新生物标志物，DHM 和其他光学解决方案可以胜过当今的分析仪，并以类似于过去在血液分析仪中引入血小板计数的方式扰乱临床和 POC 市场。