Antibody-Based Multiplex Image Analysis: Standard Analytical Workflows and Artificial Intelligence Tools for Pathologists.

Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Lab Invest . Author manuscript; available in PMC: 2026 Apr 14. Published in final edited form as: Lab Invest. 2025 Jul 29;105(10):104220. doi: 10.1016/j.labinv.2025.104220 Search in PMC Search in PubMed View in NLM Catalog Add to search Antibody-Based Multiplex Image Analysis: Standard Analytical Workflows and Artificial Intelligence Tools for Pathologists Mohamed Omar Mohamed Omar a Department of Computational Biomedicine, Cedars-Sinai Medical Center, Los Angeles, California b Cancer Therapeutics Program, Samuel Oschin Comprehensive Cancer Institute, Cedars-Sinai Medical Center, Los Angeles, California Find articles by Mohamed Omar a, b , Giuseppe Nicolo’ Fanelli Giuseppe Nicolo’ Fanelli c Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York, New York d Department of Translational Research and New Technologies in Medicine and Surgery, University of Pisa, Pisa, Italy Find articles by Giuseppe Nicolo’ Fanelli c, d , Fabio Socciarelli Fabio Socciarelli c Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York, New York Find articles by Fabio Socciarelli c , Varun Ullanat Varun Ullanat e Dana-Farber Cancer Institute, Boston, Massachusetts Find articles by Varun Ullanat e , Sreekar Reddy Puchala Sreekar Reddy Puchala e Dana-Farber Cancer Institute, Boston, Massachusetts Find articles by Sreekar Reddy Puchala e , James Wen James Wen e Dana-Farber Cancer Institute, Boston, Massachusetts Find articles by James Wen e , Alex Chowdhury Alex Chowdhury c Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York, New York e Dana-Farber Cancer Institute, Boston, Massachusetts Find articles by Alex Chowdhury c, e , Itzel Valencia Itzel Valencia c Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York, New York Find articles by Itzel Valencia c , Cristian Scatena Cristian Scatena d Department of Translational Research and New Technologies in Medicine and Surgery, University of Pisa, Pisa, Italy Find articles by Cristian Scatena d , Luigi Marchionni Luigi Marchionni c Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York, New York e Dana-Farber Cancer Institute, Boston, Massachusetts Find articles by Luigi Marchionni c, e , Renato Umeton Renato Umeton c Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York, New York e Dana-Farber Cancer Institute, Boston, Massachusetts f Harvard T.H. Chan School of Public Health, Boston, Massachusetts g Massachusetts Institute of Technology, Cambridge, Massachusetts Find articles by Renato Umeton c, e, f, g, * , Massimo Loda Massimo Loda c Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York, New York e Dana-Farber Cancer Institute, Boston, Massachusetts g Massachusetts Institute of Technology, Cambridge, Massachusetts h Broad Institute of MIT and Harvard, Cambridge, Massachusetts Find articles by Massimo Loda c, e, g, h, * Author information Article notes Copyright and License information a Department of Computational Biomedicine, Cedars-Sinai Medical Center, Los Angeles, California b Cancer Therapeutics Program, Samuel Oschin Comprehensive Cancer Institute, Cedars-Sinai Medical Center, Los Angeles, California c Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York, New York d Department of Translational Research and New Technologies in Medicine and Surgery, University of Pisa, Pisa, Italy e Dana-Farber Cancer Institute, Boston, Massachusetts f Harvard T.H. Chan School of Public Health, Boston, Massachusetts g Massachusetts Institute of Technology, Cambridge, Massachusetts h Broad Institute of MIT and Harvard, Cambridge, Massachusetts * Corresponding authors. [email protected] (R. Umeton); [email protected] (M. Loda). # These authors contributed equally: Mohamed Omar and Giuseppe Nicolo’ Fanelli. Author Contributions First draft: M.O., G.N.F., F.S., J.W., S.R., V.U., A.C., I.V., C.S., and R. U.; digital pathology analyses: M.O., G.N.F., and F.S.; software implementation: M.O., V.U., S.R., J.W., A.C., and R.U.; supervision: L.M., R.U., and M.L. All authors have read and agreed to the final draft of the manuscript. Issue date 2025 Oct. PMC Copyright notice PMCID: PMC13072558  NIHMSID: NIHMS2158863  PMID: 40744223 The publisher's version of this article is available at Lab Invest Abstract Conventional histopathology has traditionally been the cornerstone of disease diagnosis, relying on qualitative or semiquantitative visual inspection of tissue sections to detect pathological changes. Singleplex immunohistochemistry (IHC), although effective in detecting specific biomarkers, is often limited by its single-marker focus, which constrains its ability to capture the complexity of the tissue environment. The introduction of multiplexed imaging technologies, such as multiplex IHC and multiplex immunofluorescence, has been transformative, enabling the simultaneous visualization of multiple biomarkers within a single tissue section. These approaches complement morphology with quantitative multimarker data and spatial context, providing a more comprehensive view of cellular interactions and disease mechanisms. However, the rich data from multiplex IHC/multiplex immunofluorescence experiments come with significant analytical challenges, as large multichannel images require comprehensive processing to transform raw imaging data into quantitative and meaningful information. This review focuses on the standard digital image analysis workflow for multiplex imaging in pathology, covering each step from image acquisition and preprocessing to cell segmentation and biomarker quantification. We discuss the common open-source tools that support each step to guide users in selecting appropriate solutions. By outlining an end-to-end pipeline with concrete examples, this review is intended for practicing pathologists and researchers with limited computational expertise. It provides practical guidance and best practices to help integrate multiplex image analysis into routine pathology workflows and translational research, bridging the gap between advanced imaging technology and day-to-day diagnostic practice. Keywords: artificial intelligence, computational pathology, digital pathology, multiplexed imaging, PathML, spatial omics Introduction Pathology has long been a cornerstone for detecting, analyzing, and understanding disease processes, bridging basic science and clinical practice. Traditionally, pathology has relied on qualitative evaluation of tissues using hematoxylin and eosin (H and E) staining or single-marker immunohistochemistry (IHC), both of which continue to provide essential morphological and molecular insights. However, the inherently qualitative nature of these examinations highlights the need for more standardized, quantitative approaches that can capture the complexity of cellular composition and biomarker distribution across the tissue microenvironment. 1 , 2 This is particularly important for undifferentiated tumors, cancers of unknown primary, sarcomas, and most lymphoid neoplasms, which require the evaluation of multiple biomarkers and their coexpression for full characterization. 3 Currently, flow cytometry is the only method available in most clinical laboratories for multiantigen labeling of individual cells. Still, this method is unsuitable for many sample types, including formalin-fixed paraffin-embedded tissues, and lacks spatial information. 4 Unlike traditional imaging that provides a singular, often morphological or semiquantitative perspective, modern multiplexed imaging techniques–such as multiplex immunofluorescence (mIF) and multiplex IHC (mIHC)–together with high-plex IF/imaging mass cytometry (IMC), offer a comprehensive view of the tissue microenvironment by mapping the expressions of several protein markers in their native histological context. 4 - 10 This multiplexed approach enables the simultaneous assessment of multiple biomarkers, offering a more comprehensive view of the disease state that complements the insights gained from traditional H and E staining alone. The integration of such diverse data through computational analysis promises not only to enhance the accuracy of existing predictive and prognostic models but also to uncover novel insights into disease mechanisms. For instance, by simultaneously visualizing and quantifying the expression of multiple biomarkers within a single tissue section, pathologists can obtain a nuanced understanding of tumor biology, including, but not limited to, cellular composition, spatial interactions between different cell types, and patterns of immune infiltration. Investigating these dynamics across different disease stages or different treatment modalities can, in turn, identify features or biomarkers associated with cancer progression or treatment resistance. Despite the advances brought by multiplexing technologies in quantitative histopathology research, transitioning to a computational analysis-driven approach involves significant challenges. The sheer volume of data produced by multiplexed imaging requires extensive computational resources and sophisticated algorithms for tasks such as preprocessing, cell detection, and tissue or cellular features quantification. These processes are crucial for translating raw pixel data into actionable single-cell information that can be analyzed quantitatively. 11 However, beyond these technical hurdles lies substantial potential to aid both pathology practice and biomedical research, especially in fields like oncologic pathology. To fully realize this potential, a multidisciplinary approach is essential. Effective integration of skills from pathology, computer science, and medical research is required to develop and refine computational tools that can manage and interpret the complexity of multiplexed data. This interdisciplinary collaboration is not merely a necessity but an opportunity to synergize knowledge across fields, driving forward the capabilities of digital pathology. Building on this foundation, this review is intended as a practical guide for pathologists and translational researchers seeking to implement multiplex image analysis using established digital tools and workflows. We outline the spectrum of tasks pertinent to the computational analysis of these images ( Fig. 1 ) and highlight commonly used, open-source software tools that, at times, leverage artificial intelligence (AI) to streamline these tasks, ranging from preprocessing and quality control (QC)–including primary antibody validation and spectral unmixing–to advanced computational algorithms for nuclei and cell segmentation. The subsequent discussion delves into feature extraction, quantification techniques, and methods for automated cell annotation, employing both gating-based and unsupervised clustering-based approaches. Given the complexity of such analysis and the plethora of tools involved, we provide guidance on how these tools can be integrated into routine analysis tasks, from basic cell counting to complex spatial profiling. Figure 1. Open in a new tab Overview of common pathology tasks and digital tool implementation for the analysis of histopathology multiplexed imaging data. This figure illustrates the range of tasks commonly encountered in the computational analysis of this data, categorized by their relative ease of implementation. Multiparametric Imaging Modalities: One Size Does Not Fit All Histological imaging has seen an incredible increase in acquisition modalities associated with IHC in the last 10 years. Once limited to 1 or 2 markers and mostly relegated to chromogenic detection, IHC is now experiencing a substantial expansion in multiplexing capacity and detection techniques, with innovative technologies allowing for the detection of up to 40 markers on a single slide. 12 , 13 Keeping track of the myriad multiplexing technologies and detection techniques is challenging, given their rapid evolution. Therefore, we will describe the main antibody-based approaches present in pathology research. Table 1 14 , 15 , 8 , 16 - 20 provides a comparative overview of the various multiplexed imaging modalities present in the literature. Although our review mainly focuses on protein-based markers with subcellular level resolution in 2-dimensional (2D) formats, it is important to acknowledge additional imaging modalities. Notable modalities include mass spectrometry imaging, which facilitates the detection of lipids, metabolites, and drugs, 12 various modalities of spatial transcriptomics, such as from 10× Genomics Visium technology 13 and NanoString GeoMx region-of-interest-based sampling, 21 multiplex fluorescence in situ hybridization, 22 and 3-dimensional (3D) clearing microscopy. 23 Each of these techniques contributes unique insights into tissue structure and function, expanding the frontiers of quantitative multimarker tissue analysis. Table 1. Imaging modalities for multiplex staining Variable Brightfield multiplex IHC 14 , 15 Iterative chromogenic multiplex IHC 8 TSA multiplex IF 16 Cycling mIF and stripping 9 , 17 Imaging Mass Cytometry 18 - 20 Detection system Chromogens with narrow absorbance band and dedicated camera AEC or similar chromogens with iterative staining/stripping cycles Tyramide-conjugated fluorophores associated with HRP-conjugated secondary antibody Fluorophore-conjugated secondary antibodies, imaging and following stripping Primary antibodies conjugated with metals Image dimensions 2D 2D 2D and 3D 2D and 3D 2D Channel multiplexing Low (6) Moderate (6-10) Low (max 8) Very high (up to 100) High (up to 40) Key advantages High-resolution, not expensive, can be observed at the brightfield microscope No autofluorescence, inexpensive imaging, excellent visual quality control, brightfield scanners are used, no bleed-through Robust protocol, very sensitive, fast High-plex, good resolution, very well automated Fast unmixing, useful if N fluorophores > N detection channels Potential pitfalls Chromogens not resistant to time, need for a dedicated camera, low-plex Labor intensive, requires accurate alignment across rounds, limited automation Traumatic for tissue, low-plex, preservation of antigenicity Not always sensitive, lengthy protocol, expensive, some approaches have short expiration date reagents, preservation of antigenicity after repeated cycle Very long protocol, small areas acquired, very expensive, low resolution, traumatic for the tissue, antibody conjugation an issue Commercial solutions Roche Diagnostics, Enzo Life Sciences, Leica Biosystems Manual implementations Akoya Biosciences, Thermo Scientific, Biotium Lunaphore, Akoya biosciences, Cell Signaling Technology, Ultivue Standard BioTools, IonPath Open in a new tab AEC, alcohol-soluble peroxidase substrate 3-amino-9-ethylcarbazole; HRP, horseradish peroxidase; IF, immunofluorescence; IHC, immunohistochemistry; mIF, multiplex immunofluorescence; TSA, tyramide signal amplification, 2D, 2-dimensional; 3D, 3-dimensional. Primary Antibody Validation: A Prerequisite for Reliable Multiplex Analysis After the choice of the proper multiplex imaging technique, the second step is the careful validation of each biomarker performance, singularly and in the context of a multiplex panel. Addressing the complexities of antibody validation is a task often underestimated. Numerous instances of the detrimental consequences of using poorly validated reagents underscore the importance of rigorous performance checks before their use. 24 , 25 Although this review primarily focuses on the analytical workflow for multiplexed images, it is also crucial to address the experimental work that precedes image creation, which can substantially impact image quality and subsequent analyses. Singleplex IHC assays are designed and often optimized directly by vendors to detect the presence or absence of specific proteins. In contrast, multiplex assays need strong validation to use the wide range of protein expression levels, provide continuous measurements, and integrate spatial information. The design and validation of multiplex panels require expertise in immunology, histopathology, and tissue staining techniques. 26 Although robust standard operating procedures and commercial solutions are readily available for various imaging modalities, preanalytical variables and primary antibody validation remain a critical bottleneck despite the presence of numerous guidelines. 27 - 32 Cold ischemia time should be minimized to preserve enzyme activity, protein integrity, and cytomorphology and to prevent autolysis. Fixation time and extent must adhere to strict recommendations, as delayed fixation prolongs ischemic time, and overfixation can harm antigenicity. Additional critical variables include tissue processing steps, slide drying, and storage conditions. 33 Notably, the storage of precut, unstained slides has been shown to significantly impact immunoreactivity over time, with studies showing that variables such as storage temperature, duration, and sealing methods contribute to a time-dependent decline in antigenicity. 34 , 35 For instance, the slides stored at room temperature or coated with paraffin exhibit a marked decrease in immunoreactivity for markers like p53, Ki-67, and androgen receptor over a year, whereas storage at −20° C without paraffin coating better preserves antigenicity. 34 Similar degradation patterns have also been observed in in situ hybridization assays. 35 When selecting antibodies, monoclonal antibodies are preferred over polyclonal due to higher specificity and consistency, 36 and notable initiatives like Human Protein Atlas, 37 HubMap, 38 Genecards, 39 and others 40 can be helpful. Well-validated and reproducible antibodies, along with properly validated controls, are widely accessible. However, many other targets lack readily available antibodies, needing extensive validation using different tissues and techniques such as western blotting of control cell lines. 28 Importantly, even vendor-validated antibodies require careful end-user validation due to variability in staining conditions, tissue preparation, and potential antibody specificity issues. As highlighted by Sfanos et al, 41 common pitfalls in antibody validation include overreliance on vendor-provided validation data, lack of standardized protocols across laboratories, improper use of positive and negative controls, and failure to account for tissue-specific variability in antibody performance. Optimizing multiplex assays requires figuring out the best antibody dilution, incubation time, and staining order. This starts with singleplex IHC for each antibody; then, each marker is paired with a suitable chromogen/fluorophore, considering its expression level and its characteristics. 33 Upon completing panel optimization, each stain should perform equivalently to singleplex IHC in terms of control performance, expression intensity, and cell-type-specific labeling. 42 Troubleshooting may involve adjusting staining order, antibody concentration, fluorophore concentration, or fluorophore-marker pairing to address issues like steric hindrance or signal overlap. 36 Finally, image acquisition and analysis also require thorough validation, with standardized field selection to avoid bias and comparison with H and E-stained slides at critical steps. Fluorophore Selection and Image Acquisition In immunofluorescence, signal detection is made by acquiring different wavelengths across the visible light and infrared spectrum. Multiple specialized filters and fluorophores, each designed to emit light at specific wavelengths when excited, are available for different uses. Employing multiple fluorophores with non/partial-overlapping emission spectra allows for the concurrent detection and differentiation of multiple targets within a single sample, providing a detailed and comprehensive view of the biological processes and interactions occurring in real time. 3 Each fluorophore used in fluorescent imaging has its own set of advantages and limitations, primarily defined by the wavelength at which it emits light. Understanding these characteristics is crucial for selecting the proper one for specific imaging needs and improving the quality and accuracy of the data obtained. Fluorophores emitting in the blue-to-green spectral ranges (eg, DAPI, green fluorescent protein) have shorter wavelengths, which provide higher spatial resolution due to their smaller diffraction limit. This makes them particularly effective for detecting fine details and observing smaller structures within cells, such as nuclei or cytoskeletal elements. For fixed samples, these fluorophores are ideal, as they allow for detailed imaging at a cellular and even subcellular level. However, these shorter wavelengths carry higher energy, which can be problematic in “live/fresh” imaging applications. High-energy light can cause photobleaching, where the fluorophore loses its ability to fluoresce after prolonged exposure, leading to signal loss over time. Furthermore, these fluorophores are more likely to cause photodamage to the cells or tissues being observed, potentially altering biological processes or causing cell death. Therefore, although short-wavelength fluorophores offer precision, they must be used cautiously, particularly in experiments involving live samples or repeated imaging. 43 Fluorophores emitting in the red or near-infrared regions (eg, Cy5, Alexa Fluor 647) have longer wavelengths and several distinct features. These fluorophores are associated with lower energy levels, making them less likely to cause photodamage or photobleaching, thus extending their utility in long-term or live imaging experiments. Additionally, longer wavelengths penetrate deeper into tissues compared with shorter wavelengths, making them highly suitable for imaging thick tissue samples, 3D cultures, or in vivo studies where deeper tissue structures must be visualized. This deep penetration is particularly helpful for applications such as whole-organ imaging, brain imaging in live animals, or imaging through complex tissue layers like skin. However, there are trade-offs. The longer wavelengths used by these fluorophores inherently produce lower-resolution images because of the larger diffraction limit, making it challenging to capture fine cellular details with the same precision as short-wavelength fluorophores. Additionally, the fluorescence signals from these wavelengths can be weaker, requiring the use of more sensitive detectors or more powerful light sources to achieve adequate signal strength. This can add complexity and cost to the imaging setup, and in some cases, this may still result in a lower signal-to-noise ratio compared with shorter-wavelength options. Moreover, certain biological tissues can absorb or scatter light at longer wavelengths, which may reduce image clarity or contrast if not properly accounted for during imaging. 44 Overall, the choice between short- and long-wavelength fluorophores depends on the specific application requirements, including the depth of imaging needed, the resolution desired, and the nature of the sample (fresh or fixed). Combining fluorophores across the spectrum while managing their specific limitations can optimize imaging performance and provide comprehensive insights into complex biological structures and processes. However, exposure times need to be set up carefully to maintain a balance of the signal intensity across markers in the panel. Finally, autofluorescence should be considered during image acquisition. It refers to the natural emission of light by certain tissues or cellular components, such as collagen or nicotinamide adenine dinucleotide), which can interfere with the detection of fluorophores. Autofluorescence can create background noise, reducing the clarity and contrast of the images. To mitigate this, careful selection of fluorophores with emission wavelengths distinct from those typically associated with autofluorescent molecules is essential. Advanced filtering, spectral unmixing (see below), and processing techniques are also employed to minimize the impact of autofluorescence, enhancing the overall image quality. 26 After the staining, the slides need to be acquired, and when opting for a scanner system for image acquisition, several factors must be considered, including the spectral range, fluorescence throughput, automation features, multiplexing capabilities, and camera resolution, among others, to ensure the capture of high-quality images. PhenoImager HT, one of the most used multispectral digital slide imaging systems, employs proprietary multispectral imaging technology to mitigate optical spectral bleed-through between channels and effectively distinguish signal from background autofluorescence. In an internal evaluation, the average optical bleed-through was 8.7% for a 6-plex assay and 13% for an 8-plex assay, and multispectral unmixing reduced residual bleed-through to <1% in both cases. 42 Spectral Unmixing Approaches and Considerations Spectral unmixing is a fundamental technique widely used in confocal microscopy to separate fluorescent signals within lambda stacks, which involve capturing a broad range of spectra, typically far exceeding the number of fluorophores used in the sample. This approach allows for the differentiation of overlapping emission spectra, providing a clear separation of signals even when multiple fluorophores are present. However, in a multiplex TSA (tyramide signal amplification)-based imaging technique, the spectral unmixing approach is different: the number of spectra acquired by the camera exactly matches the number of fluorophores being detected. This precise correlation optimizes signal detection and minimizes spectral overlap, which is critical when using TSA to achieve high sensitivity and specificity in detecting multiple targets within the same tissue section. 45 By ensuring that the captured spectra align with the fluorophores used, this approach enhances the accuracy and efficiency of multiplex imaging. However, even with this approach, a partial overlap of absorption and emission spectra among the fluorophores used is still present, leading to channel bleed-through. 46 To address this challenge, various computational algorithms have been developed, and the most commonly used are summarized below. 47 Linear unmixing (or linear decomposition) is the most common algorithm used for spectral unmixing. This method can calculate the different contributions of each fluorophore to every channel of the image using as a reference a library of fluorescent spectra acquired individually. 29 In case an autofluorescence-dedicated channel is acquired, the linear unmixing is also able to eliminate the autofluorescence from the tissue. PICASSO is an algorithm that iteratively minimizes mutual information (the statistical dependence between channels) by subtracting one scaled channel image from another. 48 This unmixing approach has been used on multiplex TSA immunofluorescence with success from the authors, reporting a superior performance of this algorithm than linear unmixing. Additionally, Maric et al 49 developed an algorithm based on linear unmixing in which the bleed-through between channels is estimated using the LASSO regression, and a semisupervised model is used to separate the different channels. Although library-free, this technique requires the user to indicate which pairs of channels are expected to show bleed-through. LUMoS 50 is an algorithm based on clustering machine learning and has been developed for images in which the number of fluorophores is equal to or greater than the number of detectors. The drawback of this method is that it has only been used on 2-photon microscopy and never applied to widefield fluorescent microscopy, limiting its application in the pathology field. Finally, an algorithm based on nonnegative matrix factorization 51 has been applied to widefield microscopy on tissue, but its efficacy has been questioned by other authors. 52 , 53 Image Preprocessing and Quality Control Image Preprocessing Operations Image preprocessing involves several essential techniques to prepare images for the following steps of segmentation, feature extraction, and quantification. Resizing/tiling ensures images are of a uniform size, which is crucial for the effective functioning of machine learning algorithms. Grayscaling simplifies the image data by converting color images to grayscale, reducing computational requirements for certain algorithms. Binarization converts grayscale images to black and white through thresholding, whereas contrast enhancement uses methods like histogram equalization to enhance contrast or normalize the dynamic range of intensities across channels. 54 Noise reduction is commonly performed using smoothing filters, such as Gaussian or median blurs, which help reduce high-frequency variations in intensity caused by background artifacts or imaging inconsistencies. 55 Normalization, such as scaling pixel values to a 0 to 1 range or adjusting per-channel intensity distributions, is used to standardize input across images and improve the consistency of downstream analytical outputs. Although these preprocessing techniques may not directly enhance diagnostic interpretability, they are critical for preparing data that can be reliably interpreted by computational models and automated pipelines. Quality Control Measures For the technical performance of the assay, QC metrics like those used in singleplex IHC, such as batch-to-batch differences in antibody performance and antibody performance in decalcified specimens, must be considered. 36 Additionally, scan quality variability should be minimized. 42 QC is a fundamental but often under-discussed component of the digital image analysis pipeline. Proper QC ensures that the data being analyzed accurately reflect the underlying biology and are not confounded by technical artifacts or inconsistencies in staining. This is especially true for complex multiplex staining procedures, which involve multiple rounds of staining, fluorophore application, and imaging, each of which introduces potential sources of variability or error. The accuracy and reproducibility of subsequent segmentation and quantification steps depend heavily on the quality of the raw image input. Quality assessment of slides is crucial to prevent the incorporation of low-quality data into the analytical pipelines. Common artifacts such as tissue folding, air bubbles, out-of-focus areas, tissue detachment, foreign bodies, and poor staining quality are prevalent in virtual slides 39 and can significantly undermine downstream analysis. 56 , 57 For brightfield and singleplex digital slides, several digital and automated QC tools or pipelines are readily available 58 ; however, for multiplexed images, these options are limited. This is partly because mIF introduces unique challenges, such as interchannel signal bleed-through, accumulating background artifacts from multiple staining rounds, halo effect from out-of-focus foreign objects, and higher complexity in image data, which singleplex QC tools were not originally designed to handle. Consequently, many automated QC pipelines struggle to accommodate multiplex data, and manual quality checks by experienced pathologists remain the gold standard for ensuring staining quality. 59 Several methods have been developed for QC in multiplex imaging data. For instance, Jiang et al 60 created a tool based on DAPI staining, which is able to recognize some artifacts like blurring, foreign bodies, halo artifacts, and folding. Another example is StainV and QC, a TissuUmaps3 plugin for assessing staining quality and background artifacts, which requires a preliminary step of cell segmentation. 61 A third example is MxIF Q-score, 62 a DAPI-based tool for evaluating the quality of image registration, tissue microarray cores, and cell segmentation on fluorescent digital slides. CyLinter 63 is another bio-informatic tool based on single-cell analysis that is able to recognize a large array of artifacts and that can be streamlined with the MCMICRO analytical pipeline. 64 Finally, CellProfiler is an open-source, widely used image analysis software primarily designed for cell-based immunofluorescence . 65 One of its available modules, MeasureImageQuality , assesses several quality metrics, mostly related to blurring but also correlated to signal saturation (often the result of folding or out-of-focus areas) and too low estimation of the exposure time. In our experience, certain blurring metrics were dependent on the cellular content present in the images, indicating that variations in cell density or structure could impact the results. This observation aligns with guidance from the CellProfiler developers emphasizing the importance of selecting an appropriate spatial scale when using the power log-log slope method. 31 Choosing the correct spatial scale is critical, since it ensures the metric accurately reflects image quality without being disproportionately affected by the variability in cellular features. This consideration is particularly important when applying automated image analysis techniques to heterogeneous samples, where differences in cell morphology or distribution may otherwise confound the interpretation of image-blurring metrics. A notable limitation of CellProfiler is its inability to read large images, requiring a preliminary tiling step. We anticipate that more robust, generalizable automated QC tools will be further developed to provide a more comprehensive quality assessment of multiplex imaging. In the meantime, human review, ideally performed on raw and spectrally unmixed images before analysis, remains indispensable. Cell Segmentation and Annotation Following robust preprocessing and QC, cell segmentation (isolation of individual cells from a complex arrangement of tissue) and annotation (cell categorization based on unique phenotypes determined by marker’s expression) are required steps before extracting meaningful insights from multiplexed images. Accurate cell segmentation is crucial for the reliability of quantitative analyses of cell morphology, spatial distribution, and the relationships between cellular components and their molecular expressions. Advanced Computational Algorithms for Automated Cell Segmentation in Multiplexed Imaging Cell segmentation techniques range from basic thresholding methods to sophisticated machine learning algorithms. Traditional methods like thresholding and edge detection were initial tools used to distinguish cells from the background based on intensity values or gradients. Although these methods are straightforward and computationally efficient, their effectiveness is reduced in heterogeneous tissue samples and variable staining quality. In multiplexed images, where multiple markers are analyzed simultaneously, the overlap of fluorescent signals can complicate the segmentation process, making these traditional techniques less suitable for complex analyses. Currently, most cell segmentation methods begin by identifying individual nuclei, typically using a nuclear marker such as DAPI. 4 Following nuclei segmentation, cell boundaries are delineated either by detecting the cytoplasm within a specified radius around each nucleus or, more accurately, by using a cytoplasmic or cell membrane marker. 4 , 59 Here, we summarize the latest array of techniques employed in cell segmentation, highlighting their applications, advantages, and limitations within the context of multiplexed image analysis ( Table 2 ). 66 - 73 Table 2. List of popular cell and nucleus segmentation tools Variable Mesmer 68 Cellpose 69 StarDist 70 , 71 CellSeg 72 Ilastik 73 UnMICST 74 Algorithm/method Deep learning-based, with ResNet50 backbone and Feature Pyramid Network Deep learning-based, utilizing U-Net architecture Deep learning, with a U-Net-like architecture designed for star-convex shape prediction Deep Learning, utilizing R-CNN architecture Machine learning-based, utilizing interactive learning and classification algorithms Deep learning, utilizing a suite of CNN architectures with real augmentation (intentionally defocused and over-saturated images) Input image type 2D and 3D 2D (extended to 3D but without 3D training data) 2D and 3D 2D and 3D 2D and 3D 2D and 3D Training data Trained using TissueNet, a comprehensive image dataset featuring >1 million paired whole-cell and nuclear annotations from 9 organs and captured using 6 different imaging platforms Trained on a diverse dataset comprising >70,000 segmented objects from a variety of cell images A dataset of 497 manually annotated reat microscopy images of cell nuclei from the 2018 Data Science Bowl 75 A dataset from the 2018 Kaggle Data Science Bowl, containing 29,464 ground truth segmented nuclei 75 Based on sparse user-provided training annotations; no extensive training dataset required Trained on manually curated data from 7 tissue types with ~10,400 nuclei labeled for nuclear contours, centers, and background. Also includes training with defocused and saturated images for real augmentation Cytoplasmic marker needed Yes Not required; can perform segmentation based on cell morphology and machine learning inference Optional; model is designed to predict cell shapes, can use cytoplasmic markers for enhanced segmentation Recommended for optimal segmentation performance, especially in dense tissue samples Not specifically required; utilizes machine learning on user-defined annotations to segment and classify objects Not specified Usability Model weights can be used in Python-based pipelines such as DeepCell; Intuitive web-based interface Python package with GUI for ease of use; some coding possible for advanced tasks Python API, with some GUI elements available via plugins or extensions Primarily command-line interface with some GUI elements User-friendly GUI designed for biologists or researchers without in-depth computational background Command-line interface with Python API, may require proficiency in programming and computational image analysis Integration Can integrate with common bioimage analysis workflows, including PathML 11 Standalone Python package, integrates with common Python data science tools Integrates, with Fiji/ImageJ and QuPath ecosystems Python-based, can be integrated with standard scientific Python stack and image analysis tools like Fiji/ ImageJ Can be used as a standalone application or integrated with Python or Fiji for automated workflows. Can process data larger than RAM and integrate with existing workflows via command-line for batch processing Integrated into MCMICRO 66 workflow Customizability Pretrained models available, can be further trained on user-provided datasets Models can be trained with user data for customized segmentation tasks, designed for continuous improvement by periodically re-training the model using community-contributed data Pretrained models available (for 2D only), can be further trained on user-provided datasets Optimized for fluorescence and brightfield biological microscopy images User annotations guide the learning process; predefined workflows are adaptable to various biological image analysis problems Not specified Open in a new tab API, application programming interface; GUI, graphic user interface; RAM, random-access memory; R-CNN, region-based convolutional neural network; 2D, 2-dimensional; 3D, 3-dimensional. A significant challenge for segmentation tasks is the scarcity of large, expert-annotated data sets for tissue structures. These data sets are instrumental for training models to accurately recognize and classify cellular components in histopathological images. Notably, the process of creating such data sets involves extensive manual annotation by expert pathologists, who identify different tissue and cellular structures within large gigapixel whole slide images. Typically, this includes annotating thousands of nuclei or cell types per slide, ideally across various tissue types obtained through diverse preparation, preprocessing, and scanning techniques. Although conventional H and E-stained whole slide images are commonly available in many institutions, slides prepared for multiplexed imaging modalities are more costly and require specialized equipment and expertise, thereby limiting their availability. As a result, building comprehensive annotation data sets for multiplexed images not only demands substantial financial investment for image generation but also significant resources for manual annotation efforts. Despite these challenges, data sets of annotated multiplexed imaging have been increasingly available, greatly aiding the development of robust cell segmentation pipelines. 66 , 74 - 76 These data sets often encompass a variety of multiplexed imaging modalities and tissue types, which is critical for developing models that are robust against technical and biological variability. Feature Extraction and Quantification Following cell segmentation, the next critical step before annotation in the analysis pipeline is quantification. This process involves extracting quantitative data from segmented cells or nuclei and translating visual information into numerical values suitable for statistical analysis. Typically, this results in the creation of a feature count matrix, where each row corresponds to a single-cell or nucleus, and columns represent various features and metrics of interest. Features often include marker expression levels within individual cells as measured by the fluorescence intensity, which can include the minimum and maximum intensity and the average intensity across the segmented cell area. Additionally, quantification algorithms record the spatial location of each cell, typically recorded as x and y coordinates on the imaging plane–as well as morphometric features such as eccentricity, which measures the elongation of a cell–and size, often reported as the area or volume of a cell or nucleus. Feature extraction methods can be broadly grouped into 2 main categories: nondeep learning and deep learning applications. Nondeep Learning Applications (Pixel- and Cell-Based Feature Extraction) Nondeep learning algorithms for feature quantification primarily use segmentation masks as inputs. Several tools exemplify this approach. For instance, PathML , 11 a Python-based package, uses Skimage’s “ regionprops” functionality to calculate features such as volume, bounding boxes, and intensity. MCMICRO offers similar functionality through its MCQUANT method for Nextflow users. 64 Eng et al created a Python-based package, cycIF_Validation , which focuses on improving processes for antibody specificity, signal removal, and batch normalization in cyclic mIF images. 77 Ilastik uses nondeep learning machine learning, such as random forests, to count and track objects in addition to classification. 71 Windhager et al 78 utilize this latter and CellProfiler in an end-to-end pipeline for multiplex image analysis in R. 78 In addition, off-the-shelf software products also play a significant role in analyzing multiplex images, such as Inform (Akoya Biosciences), which has traditionally been used for multispectral image analysis, fluorescent intensity quantitation, and rule-based phenotyping. Deep Learning Applications (Pixel-Based Feature Extraction) In tissue-based pathology specifically, deep learning approaches have been broadly utilized not only for segmentation and classification but also extensively for image quantification (eg, automated IHC scoring 79 - 81 and morphometric analysis 82 ), representation learning supporting content-based image retrieval, 83 , 84 and more recently visual-language models integrating imaging with text. 85 - 87 A significant feature of these models is their ability to process entire images as input, allowing for direct pixel-level feature extraction and reducing the need for manual identification of individual cell locations. Broadly, feature extraction models can be grouped into 2 main categories: (1) models that create latent space features for downstream analysis, and (2) models that identify pertinent biological features. Both groups of models often employ encoder-decoder network architectures, which are widely used in image segmentation tasks but are broadly applicable to other image analysis tasks. In digital pathology, these models have been mostly applied to H and E-stained images, rather than mIHC/mIF; however, their utility to multiplexed image analysis remains promising, especially with the increasing availability of large multiplex pathology imaging data sets. Similar to feature extraction models, quantification deep learning methods also build on encoder-decoder architectures. For example, Silina et al 88 built on a typical encoder-decoder network (HookNet) to create Hooknet-tertiary lymphoid structures to quantify tertiary lymphoid structures in H&E images. Liu et al 89 built off StarNet to create a model that quantifies myocardial inflammatory infiltration in H and E images. In Haghighi et al, 90 researchers used ResNet50 as the backbone of an encoder-decoder model to quantify dopamine neurons in Parkinson’s Disease. Although much of the literature discussed in this section focuses on H and E images, we provide them as representative demonstrations of successful deep learning applications in histopathology tasks, which can be feasibly adapted for mIHC/mIF image analysis. Cell Annotation Methods in Multiplexed Imaging Cell-type annotation is a crucial step in understanding the cellular composition and the intricate dynamics within the tumor microenvironment. In multiplexed imaging data, cell-type annotation has traditionally depended on the expert knowledge of pathologists who manually identify and label cell types based on morphological features and staining patterns. Although this manual approach is accurate, it can be labor intensive and subject to variability between observers. To improve the efficiency and reliability of cell phenotyping in multiplexed imaging data, automated or semiautomated annotation methods have gained prominence. These methods utilize well-characterized feature sets derived from known biomarker expressions and spatial distributions to identify cell types and states. Among the approaches employed are gating, unsupervised clustering, and the integration of deep learning techniques, all of which are increasingly acknowledged for their potential to transform cell phenotyping in multiplexed imaging. Gating-Based Approaches Gating is a technique adopted from flow cytometry and is among the most popular approaches for automated phenotyping in multiplexed imaging data. This approach employs a multidimensional space where each axis represents the intensity of a different fluorescent marker. Cells are selected or “gated” based on predefined intensity thresholds that correspond to known phenotypes. The spatial information can also be combined with the quantified fluorescence signals to enhance the accuracy of cell phenotyping. For instance, gating might be used to differentiate populations of CD8+ T cells from regulatory T cells by their respective marker expression profiles. This method’s success hinges on the clear definition of phenotypes and the availability of markers that can reliably distinguish between them. However, gating’s effectiveness can be reduced by technical noise factors, such as image processing artifacts or imperfect cell segmentation. Although gating-based cell phenotyping does not require specific computational tools, since it is entirely dependent on setting expert-defined decision rules, some publicly available tools offer gating-based functionality and intuitive user interfaces for cell-type annotation. For instance, Cytomapper, an R toolkit for spatial data analysis, allows hierarchical gating-based on the expression levels of up to 24 markers using a shiny interface. 91 Unsupervised Clustering-Based Approaches Unsupervised clustering is another technique used to group cells based on phenotypic similarity, which is particularly useful for identifying novel cell types without prior bias. 92 - 94 This method is commonly employed in single-cell RNA sequencing experiments where the cells are grouped into distinct clusters based on their gene expression profiles, with each cluster potentially representing a different cell-type or state. Unlike gating, which relies heavily on the analyst’s expertise and can introduce bias, unsupervised clustering offers a data-driven approach to cell-type identification. However, a benchmarking study by Hickey et al, 95 which compared hand gating with unsupervised clustering in annotating cell types in co-detection by indexing data, has shown that as the granularity of cell-type identification increases, the accuracy of labeling can decrease. This poses a challenge for clustering algorithms, which must balance the granularity of cell types with the confidence in the accuracy of their identification. The researchers managed this by avoiding overly subtle phenotype annotations during the clustering step, which can often lead to misclassification due to the continuous nature of marker expression levels. Moreover, the study recommended the use of overclustering, followed by spatial verification to refine cell-type identification. 95 In practice, both standalone clustering algorithms and those used in popular single-cell transcriptomics toolkits like Seurat 96 , 97 or Scanpy 98 can be employed for analyzing single-cell data from multiplexed imaging. Additionally, several publicly available tools have been developed specifically for analyzing data generated from spatially resolved technologies, incorporating clustering-based approaches for cell-type annotation. For example, Giotto, an R package, allows the use of various clustering algorithms, such as Louvain or Leiden clustering, on the single-cell data obtained from spatial transcriptomics or proteomics. 99 Another tool, ImaCytE, utilizes conventional dimensionality reduction and clustering methods for cell phenotyping in IMC data. 100 Semisupervised Clustering Approaches Semisupervised clustering has been used in the past for flow cytometry and mass cytometry data, 101 but scarcely applied to mIF/IMC data. Recently, Seal et al 102 compared the performance of several algorithms, including Random Forest (RF), linear discriminant analysis, and quadratic discriminant analysis, on different cohorts of mIF and IMC data and found that RF showed better performance, compared with linear discriminant analysis and quadratic discriminant analysis. Downstream Applications: From Research to Translational Implementation Traditional methods, such as singleplex IHC, fail to fully capture the complexity of the tumor microenvironment. On the other hand, multiplex technologies offer the ability to analyze multiple biomarkers simultaneously in situ, which could potentially have diagnostic and clinical benefits. In cancer research, multiplex imaging platforms are increasingly being utilized to investigate the tumor microenvironment and uncover previously unknown disease mechanisms, therapeutic targets, or potential biomarkers for risk stratification and outcome prediction. In non—small cell lung cancer, mIF has been used to simultaneously profile PD-L1 expression and immune cell populations, enhancing immunotherapy stratification beyond standard PD-L1 IHC staining. 103 In lymphomas, mIHC has revealed the spatial organization of immune cells and identified microenvironmental features associated with patient prognosis and treatment stratification. 104 In breast cancer, the spatial distribution of tumor-infiltrating lymphocytes has been shown to correlate with better patient outcomes, potentially serving as a valuable prognostic marker. 105 Furthermore, in ovarian cancer, studies using multiplex imaging have demonstrated that a higher density of intratumoral CD3+ T cells is predictive of improved survival. 106 Understanding tumor-stromal interactions is another key area of investigation enhanced with multiplex imaging. These interactions were shown to drive resistance to chemotherapy in pancreatic cancer 107 , 108 and predict metastatic progression in localized prostate cancer. 109 More broadly, preclinical studies across several tumor types demonstrate that multiplex imaging can uncover spatial biomarkers, predict therapeutic response, and define immune niches in ways not achievable with H and E-based analysis alone. 4 , 110 , 111 Despite the numerous research applications of multiplex assays, many pathologists and investigators are understandably cautious regarding their translational value, given practical barriers such as the need for specialized equipment, high cost, and the lack of regulatory-approved, standardized workflows. Indeed, for some applications, such as assessing immune cell infiltration, simpler approaches using routine H and E or single-plex IHC slides are already well-established and more readily deployable. However, as the field continues to evolve, multiplex antibody-based imaging techniques are expected to play a potentially significant role in cancer diagnostics and therapeutics. Notably, quantitative computational analysis workflows leveraging AI models can further support the discovery of novel biomarkers, refinement of predictive and prognostic models, and development of more personalized treatment strategies, potentially improving clinical outcomes. End-to-End Workflows for Multiplexed Image Analysis Analyzing multiplex imaging data involves a complex, multistep process that requires the integration of various computational techniques to fully exploit the richness of the data. Establishing an end-to-end workflow presents significant challenges, chiefly the integration of multiple, often disparate, processing steps into a seamless pipeline ( Fig. 2 ). Each step–from image preprocessing and cell detection to feature extraction and phenotyping–requires a specialized approach. These steps must not only be effective in isolation but also harmonize with the entire pipeline to ensure the integrity and relevance of the data. The complexity of these workflows is further amplified by the diversity of multiplexed imaging modalities. Technologies such as IMC, MIBI, CODEX, mIF, and mIHC produce data with unique characteristics, necessitating workflow flexibility and adaptability. This is no small feat, considering that the workflow must maintain robustness across various file formats, imaging resolutions, and staining protocols. Furthermore, the sheer data volume generated by high-resolution multiplexed imaging poses a considerable challenge. The workflow must not only handle large datasets efficiently but also apply rigorous QC measures to detect and correct artifacts and inconsistencies, which could otherwise lead to inaccurate biological interpretations. Figure 2. Open in a new tab Overview of the analysis workflow for histopathology multiplex imaging data. As the field of digital pathology evolves, an increasing number of tools have been developed to provide complete modular pipelines. These end-to-end workflows guide users from the acquisition of raw image data to the derivation of interpretable biological insights by streamlining the analysis through the consolidation of essential steps, including image preprocessing, segmentation, feature extraction, and phenotyping, culminating in a structured output format ( Table 3 ). 105 - 121 This format is designed to be readily utilized for further statistical analysis or machine learning applications, facilitating a smoother transition from data to discovery. 112 , 113 Table 3. List of publicly available end-to-end workflows for the analysis of multiplexed imaging data Variable PathML 11 Squidpy 114 CytoKit 115 MCMICRO 64 SIMPLI 116 Steinbock 78 Imaging modalities Various technologies including IMC, MIBI, CODEX, mIF, and mIHC Various technologies including IMC, MIBI, CODEX, mIF, and mIHC mIF Various technologies including IMC, MIBI, CODEX, mIF, and mIHC Various technologies including IMC, MIBI, CODEX, mIF, and mIHC Mainly for IMC but can be modified to handle other modalities Programming language/platform Python Python Python Galaxy 117 and Nextflow 118 R Input Raw images Raw images Raw images Raw images Raw images Raw images Stitching Supported using Groovy-based functionality Not supported Not supported ASHLAR 119 Not supported Not supported Image preprocessing Various preprocessing transforms including Gaussian, median, and box blur, normalization, superpixel interpolation, and morphological operations Grayscale conversion and smoothing using Gaussian filter Cycle registration and deconvolution Illumination correction using BaSiC 120 Includes normalization and background noise reduction capabilities IMC preprocessing, including hot pixel filtering, denoising, and channel-to-channel spillover. Preprocessing other modalities is not supported Nuclei and cells segmentation Mesmer 66 or custom user-trained models Watershed, StarDist 68 , 69 and Cellpose 69 ClassifyPixels-Unet 21 S3segmenter (watershed segmentation), UnMICST, 74 Ilastic, 73 Cypository Deterministic and deep learning models (CellProfiler 67 , 121 and custom models) Mesmer, 68 Ilastic, 73 CellProfiler, 67 , 121 and Cellpose 69 Feature extraction Segmentation feature extraction using Scikit-image Segmentation feature extraction using Scikit-image Segmentation feature extraction Segmentation feature extraction using MCQuant 122 Cell- and pixel-level feature extraction Segmentation feature extraction using Scikit-image Cell phenotyping Gating and unsupervised clustering Unsupervised clustering Gating FastPG 123 : unsupervised clustering approach derived from PhenoGraph Gating and unsupervised clustering Gating and unsupervised clustering using external packages Output format for feature matrix AnnData object AnnData object FCS or CSV formats AnnData object Tabular files AnnData object Open in a new tab CODEX, codetection by indexing; CSV, comma-separated values; FCS, flow cytometry standard; ; IMC, imaging mass cytometry; MIBI, multiplexed ion beam imaging; mIF, multiplex immunofluorescence; mIHC, multiplex immunohistochemistry. In a nutshell, multiplex antibody-based imaging technologies, such as mIF and mIHC, offer powerful opportunities to explore the spatial complexity of tissue architecture, cellular phenotypes, and biomarker coexpression within the same specimen. As these platforms continue to mature, quantitative digital analysis workflows are becoming increasingly critical to extract meaningful biological and clinical insights from highly complex image data sets. However, the path toward clinical adoption remains challenging, and practical barriers, including data size, image complexity, standardization of analytical pipelines, and integration into existing clinical workflows, must be addressed systematically to enable meaningful clinical implementation. Although innovations in cloud computing, distributed storage, and AI-based spatial modeling are beginning to emerge, further development of accessible, standardized, and validated analysis workflows is crucial to realize the full potential of multiplex imaging. In this review, we have outlined the standard analytical steps, available tools, and common workflows that pathologists and translational scientists can implement today for the analysis of multiplexed antibody-based images. Although technical challenges and translational hurdles remain, ongoing technological innovations hold promise for integrating multiplex imaging more fully into precision pathology and personalized medicine strategies in the future. Funding L.M. is supported by the National Cancer Institute (NCI) grant U54CA273956. G.N.F. is supported by a fellowship from the American-Italian Cancer Foundation (AICF) and by the Italian Ministry of University and Research–PON “Research and Innovation” 2014-2020 (PON R and I) Actions IV. 4 “Doctorates and research contracts on innovation topics”. M.L. is supported by the National Cancer Institute (NCI) grants P50CA211024 and P01CA265768, the USA Department of Defense (DoD) grant DoD PC160357, as well as the Prostate Cancer Foundation. Footnotes Declaration of Competing Interest None reported. Declaration of Generative AI and AI-Assisted Technologies in the Writing Process During the preparation of this work the author(s) used GPT4DFCI–a private and secure generative AI tool based on GPT-4 models and deployed at Dana-Farber Cancer Institute for nonclinical use–to improve this manuscript. Specifically, the tool was used to reword certain passages and further critique our work. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication. References 1. Penney KL, Tyekucheva S, Rosenthal J, et al. Metabolomics of prostate cancer gleason score in tumor tissue and serum. Mol Cancer Res. 2021;19 (3):475–484. 10.1158/1541-7786.MCR-20-0548 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Kaul KL. Preparing pathology for precision medicine: challenges and opportunities. Virchows Arch. 2017;471(2):141–146. 10.1007/s00428-017-2141-z [ DOI ] [ PubMed ] [ Google Scholar ] 3. Wharton KA, Wood D, Manesse M, Maclean KH, Leiss F, Zuraw A. Tissue multiplex analyte detection in anatomic pathology–pathways to clinical implementation. Front Mol Biosci. 2021;8:672531. 10.3389/fmolb.2021.672531 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Harms PW, Frankel TL, Moutafi M, et al. Multiplex Immunohistochemistry and immunofluorescence: a practical update for pathologists. Mod Pathol. 2023;36(7):100197. 10.1016/j.modpat.2023.100197 [ DOI ] [ PubMed ] [ Google Scholar ] 5. Lin JR, Izar B, Wang S, et al. Highly multiplexed immunofluorescence imaging of human tissues and tumors using t-CyCIF and conventional optical microscopes Chakraborty AK, Raj A, Marr C, Horváth P, eds. eLife. 2018;7: e31657. 10.7554/eLife.31657 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Schürch CM, Bhate SS, Barlow GL, et al. Coordinated cellular neighborhoods orchestrate antitumoral immunity at the colorectal cancer invasive front. Cell. 2020;182(5):1341–1359.e19. 10.1016/j.cell.2020.07.005 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Angelo M, Bendall SC, Finck R, et al. Multiplexed ion beam imaging of human breast tumors. Nat Med. 2014;20(4):436–442. 10.1038/nm.3488 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Tsujikawa T, Kumar S, Borkar RN, et al. Quantitative multiplex immunohistochemistry reveals myeloid-inflamed tumor-immune complexity associated with poor prognosis. Cell Reports. 2017;19(1):203–217. 10.1016/j.celrep.2017.03.037 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Lin JR, Fallahi-Sichani M, Chen JY, Sorger PK. Cyclic immunofluorescence (CycIF), a highly multiplexed method for single-cell imaging. Curr Protoc Chem Biol. 2016;8(4):251–264. 10.1002/cpch.14 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Lewis SM, Asselin-Labat ML, Nguyen Q, et al. Spatial omics and multiplexed imaging to explore cancer biology. Nat Methods. 2021;18(9):997–1012. 10.1038/s41592-021-01203-6 [ DOI ] [ PubMed ] [ Google Scholar ] 11. Rosenthal J, Carelli R, Omar M, et al. Building tools for machine learning and artificial intelligence in cancer research: best practices and a case study with the PathML toolkit for computational pathology. Mol Cancer Res. 2022;20(2):202–206. 10.1158/1541-7786.MCR-21-0665 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Shimma S. Mass Spectrometry Imaging. Mass Spectrom (Tokyo). 2022;11(1):A0102. 10.5702/massspectrometry.A0102 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Williams CG, Lee HJ, Asatsuma T, Vento-Tormo R, Haque A. An introduction to spatial transcriptomics for biomedical research. Genome Med. 2022;14 (1):68. 10.1186/s13073-022-01075-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Fassler DJ, Abousamra S, Gupta R, et al. Deep learning-based image analysis methods for brightfield-acquired multiplex immunohistochemistry images. Diagn Pathol. 2020;15(1):100. 10.1186/s13000-020-01003-0 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Morrison LE, Lefever MR, Behman LJ, et al. Brightfield multiplex immunohistochemistry with multispectral imaging. Lab Invest. 2020;100(8):1124–1136. 10.1038/s41374-020-0429-0 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Viratham Pulsawatdi A, Craig SG, Bingham V, et al. A robust multiplex immunofluorescence and digital pathology workflow for the characterisation of the tumour immune microenvironment. Mol Oncol. 2020;14(10):2384–2402. 10.1002/1878-0261.12764 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Rivest F, Eroglu D, Pelz B, et al. Fully automated sequential immunofluorescence (seqIF) for hyperplex spatial proteomics. Sci Rep. 2023;13(1):16994. 10.1038/s41598-023-43435-w [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Rost S, Giltnane J, Bordeaux JM, Hitzman C, Koeppen H, Liu SD. Multiplexed ion beam imaging analysis for quantitation of protein expresssion in cancer tissue sections. Lab Invest. 2017;97(8):992–1003. 10.1038/labinvest.2017.50 [ DOI ] [ PubMed ] [ Google Scholar ] 19. Chang Q, Ornatsky OI, Siddiqui I, Loboda A, Baranov VI, Hedley DW. Imaging mass cytometry. Cytometry A. 2017;91(2):160–169. 10.1002/cyto.a.23053 [ DOI ] [ PubMed ] [ Google Scholar ] 20. Kim EN, Chen PZ, Bressan D, et al. Dual-modality imaging of immunofluorescence and imaging mass cytometry for whole slide imaging with accurate single-cell segmentation. bioRxiv. 2023;2023.02.23.529718 10.1101/2023.02.23.529718 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Hernandez S, Lazcano R, Serrano A, et al. Challenges and opportunities for immunoprofiling using a spatial high-plex technology: the nanostring geomx ® digital spatial profiler. Front Oncol. 2022;12:890410. 10.3389/fonc.2022.890410 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Chen KH, Boettiger AN, Moffitt JR, Wang S, Zhuang X. RNA imaging. Spatially resolved, highly multiplexed RNA profiling in single cells. Science. 2015;348(6233):aaa6090. 10.1126/science.aaa6090 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Sabdyusheva Litschauer I, Becker K, Saghafi S, et al. 3D histopathology of human tumours by fast clearing and ultramicroscopy. Sci Rep. 2020;10(1):17619. 10.1038/s41598-020-71737-w [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Nelson AW, Groen AJ, Miller JL, et al. Comprehensive assessment of estrogen receptor beta antibodies in cancer cell line models and tissue reveals critical limitations in reagent specificity. Mol Cell Endocrinol. 2017;440:138–150. 10.1016/j.mce.2016.11.016 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Andersson S, Sundberg M, Pristovsek N, et al. Insufficient antibody validation challenges oestrogen receptor beta research. Nat Commun. 2017;8: 15840. 10.1038/ncomms15840 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Hernandez S, Rojas F, Laberiano C, Lazcano R, Wistuba I, Parra ER. Multiplex immunofluorescence tyramide signal amplification for immune cell profiling of paraffin-embedded tumor tissues. Front Mol Biosci. 2021;8: 667067. 10.3389/fmolb.2021.667067 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Fusco N, Ivanova M, Frascarelli C, et al. Advancing the PD-L1 CPS test in metastatic TNBC: Insights from pathologists and findings from a nationwide survey. Crit Rev Oncol Hematol. 2023;190:104103. 10.1016/j.critrevonc.2023.104103 [ DOI ] [ PubMed ] [ Google Scholar ] 28. Howat WJ, Lewis A, Jones P, et al. Antibody validation of immunohistochemistry for biomarker discovery: recommendations of a consortium of academic and pharmaceutical based histopathology researchers. Methods. 2014;70(1):34–38. 10.1016/j.ymeth.2014.01.018 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 29. Uhlen M, Bandrowski A, Carr S, et al. A proposal for validation of antibodies. Nat Methods. 2016;13(10):823–827. 10.1038/nmeth.3995 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 30. Bordeaux J, Welsh AW, Agarwal S, et al. Antibody validation. BioTechniques. 2010;48(3):197–209. 10.2144/000113382 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Marchiò C, Criscitiello C, Scatena C, et al. Think “HER2” different: integrative diagnostic approaches for HER2-low breast cancer. Pathologica. 2023;115(6):292–301. 10.32074/1591-951X-942 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Dameri M, Cirmena G, Ravera F, et al. Standard operating procedures (SOPs) for non-invasive multiple biomarkers detection in an academic setting: a critical review of the literature for the RENOVATE study protocol. Crit Rev Oncol Hematol. 2023;185:103963. 10.1016/j.critrevonc.2023.103963 [ DOI ] [ PubMed ] [ Google Scholar ] 33. Parra ER, Jiang M, Solis L, et al. Procedural requirements and recommendations for multiplex immunofluorescence tyramide signal amplification assays to support translational oncology studies. Cancers. 2020;12(2):255. 10.3390/cancers12020255 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. Andeen NK, Bowman R, Baullinger T, Brooks JM, Tretiakova MS. Epitope preservation methods for tissue microarrays: longitudinal prospective study. Am J Clin Pathol. 2017;148(5):380–389. 10.1093/ajcp/aqx062 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 35. Baena-Del Valle JA, Zheng Q, Hicks JL, et al. Rapid loss of RNA detection by in situ hybridization in stored tissue blocks and preservation by cold storage of unstained slides. Am J Clin Pathol. 2017;148(5):398–415. 10.1093/ajcp/aqx094 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Taube JM, Akturk G, Angelo M, et al. The Society for Immunotherapy of Cancer statement on best practices for multiplex immunohistochemistry (IHC) and immunofluorescence (IF) staining and validation. J Immunother Cancer. 2020;8(1):e000155. 10.1136/jitc-2019-000155 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Nilsson P, Paavilainen L, Larsson K, et al. Towards a human proteome atlas: high-throughput generation of mono-specific antibodies for tissue profiling. Proteomics. 2005;5(17):4327–4337. 10.1002/pmic.200500072 [ DOI ] [ PubMed ] [ Google Scholar ] 38. Quardokus EM, Saunders DC, McDonough E, et al. Organ mapping antibody panels: a community resource for standardized multiplexed tissue imaging. Nat Methods. 2023;20(8):1174–1178. 10.1038/s41592-023-01846-7 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Stelzer G, Rosen N, Plaschkes I, et al. The genecards suite: from gene data mining to disease genome sequence analyses. Curr Protoc Bioinformatics. 2016;54(1):1.30.1–1.30.33. 10.1002/cpbi.5 [ DOI ] [ PubMed ] [ Google Scholar ] 40. Biddle M, Stylianou P, Rekas M, et al. Improving the integrity and reproducibility of research that uses antibodies: a technical, data sharing, behavioral and policy challenge. MAbs. 2024;16(1):2323706. 10.1080/19420862.2024.2323706 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Sfanos KS, Yegnasubramanian S, Nelson WG, et al. If this is true, what does it imply? How end-user antibody validation facilitates insights into biology and disease. Asian J Urol. 2019;6(1):10–25. 10.1016/j.ajur.2018.11.006 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 42. Hoyt CC. Multiplex immunofluorescence and multispectral imaging: forming the basis of a clinical test platform for immuno-oncology. Front Mol Biosci. 2021;8:674747. 10.3389/fmolb.2021.674747 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Grimm JB, Lavis LD. Caveat fluorophore: an insiders’ guide to small-molecule fluorescent labels. Nat Methods. 2022;19(2):149–158. 10.1038/s41592-021-01338-6 [ DOI ] [ PubMed ] [ Google Scholar ] 44. Hong G, Antaris AL, Dai H. Near-infrared fluorophores for biomedical imaging. Nat Biomed Eng. 2017;1(1):1–22. 10.1038/s41551-016-0010 [ DOI ] [ Google Scholar ] 45. Clutter MR, Heffner GC, Krutzik PO, Sachen KL, Nolan GP. Tyramide signal amplification for analysis of kinase activity by intracellular flow cytometry. Cytometry A. 2010;77(11):1020–1031. 10.1002/cyto.a.20970 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 46. Lim JCT, Yeong JPS, Lim CJ, et al. An automated staining protocol for seven-colour immunofluorescence of human tissue sections for diagnostic and prognostic use. Pathology. 2018;50(3):333–341. 10.1016/j.pathol.2017.11.087 [ DOI ] [ PubMed ] [ Google Scholar ] 47. Tsurui H, Nishimura H, Hattori S, Hirose S, Okumura K, Shirai T. Seven-color fluorescence imaging of tissue samples based on Fourier spectroscopy and singular value decomposition. J Histochem Cytochem. 2000;48(5):653–662. 10.1177/002215540004800509 [ DOI ] [ PubMed ] [ Google Scholar ] 48. Seo J, Sim Y, Kim J, et al. PICASSO allows ultra-multiplexed fluorescence imaging of spatially overlapping proteins without reference spectra measurements. Nat Commun. 2022;13(1):2475. 10.1038/s41467-022-30168-z [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 49. Maric D, Jahanipour J, Li XR, et al. Whole-brain tissue mapping toolkit using large-scale highly multiplexed immunofluorescence imaging and deep neural networks. Nat Commun. 2021;12(1):1550. 10.1038/s41467-021-21735-x [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 50. McRae TD, Oleksyn D, Miller J, Gao YR. Robust blind spectral unmixing for fluorescence microscopy using unsupervised learning. PLoS One. 2019;14 (12):e02254. 10.10.1371/journal.pone.0225410 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 51. Jiménez-Sánchez D, Ariz M, Morgado JM, Cortés-Domínguez I, Ortiz-de-Solórzano C. NMF-RI: blind spectral unmixing of highly mixed multispectral flow and image cytometry data. Bioinformatics. 2020;36(5):1590–1598. 10.1093/bioinformatics/btz751 [ DOI ] [ PubMed ] [ Google Scholar ] 52. Geng X rui, yan Ji L, Sun K. Non-negative matrix factorization based unmixing for principal component transformed hyperspectral data. Frontiers Inf Technol Electronic Eng. 2016;17(5):403–412. 10.1631/FITEE.1600028 [ DOI ] [ Google Scholar ] 53. Wei J, Wang X. An overview on linear unmixing of hyperspectral data. Math Probl Eng. 2020;2020:1–12. 10.1155/2020/3735403 [ DOI ] [ Google Scholar ] 54. Hum YC, Lai KW, Mohamad Salim MI. Multiobjectives bihistogram equalization for image contrast enhancement. Complexity. 2014;20(2):22–36. 10.1002/cplx.21499 [ DOI ] [ Google Scholar ] 55. Haddad RA, Akansu AN. A class of fast Gaussian binomial filters for speech and image processing. IEEE Trans Signal Process. 1991;39(3):723–727. 10.1109/78.80892 [ DOI ] [ Google Scholar ] 56. Laberiano-Fernández C, Hernández-Ruiz S, Rojas F, Parra ER. Best practices for technical reproducibility assessment of multiplex immunofluorescence. Front Mol Biosci. 2021;8:660202. 10.3389/fmolb.2021.660202 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 57. Marrero A, Lawrence S, Wilsker D, Voth AR, Kinders R. Translating pharmacodynamic biomarkers from bench to bedside: analytical validation and fit-for-purpose studies to qualify multiplex immunofluorescent assays for use on clinical core biopsy specimens. Semin Oncol. 2016;43(4):453–463. 10.1053/j.seminoncol.2016.06.003 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 58. Janowczyk A, Zuo R, Gilmore H, Feldman M, Madabhushi A. HistoQC: an open-source quality control tool for digital pathology slides. JCO Clin Cancer Inform. 2019;3:1–7. 10.1200/CCI.18.00157 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 59. Lazcano R, Rojas F, Laberiano C, Hernandez S, Parra ER. Pathology quality control for multiplex immunofluorescence and image analysis assessment in longitudinal studies. Front Mol Biosci. 2021;8. 10.3389/fmolb.2021.661222 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 60. Jiang J, Moore R, Jordan CE, et al. Multiplex immunofluorescence image quality checking using dapi channel–referenced evaluation. J Histochem Cytochem. 2023;71(3):121. 10.1369/00221554231161693 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 61. Behanova A, Avenel C, Andersson A, et al. Visualization and quality control tools for large-scale multiplex tissue analysis in TissUUmaps3. Biol Imaging. 2023;3:e6. 10.1017/S2633903X23000053 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 62. Bao S, Li J, Cui C, et al. MxIF Q-score: Biology-informed quality assurance for multiplexed immunofluorescence imaging. In: Huo Y, Millis BA, Zhou Y, Wang X, Harrison AP, Xu Z, eds. Medical Optical Imaging and Virtual Microscopy Image Analysis. Switzerland: Springer Nature; 2022:42–52. 10.1007/978-3-031-16961-8_5 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 63. Baker GJ, Novikov E, Zhao Z, et al. Quality control for single-cell analysis of high-plex tissue profiles using CyLinter. Nat Methods. 2024;21(12):2248–2259. 10.1038/s41592-024-02328-0 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 64. Schapiro D, Sokolov A, Yapp C, et al. MCMICRO: a scalable, modular image-processing pipeline for multiplexed tissue imaging. Nat Methods. 2022;19 (3):311–315. 10.1038/s41592-021-01308-y [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 65. Carpenter AE, Jones TR, Lamprecht MR, et al. CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol. 2006;7(10):R100. 10.1186/gb-2006-7-10-r100 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 66. Greenwald NF, Miller G, Moen E, et al. Whole-cell segmentation of tissue images with human-level performance using large-scale data annotation and deep learning. Nat Biotechnol. 2022;40(4):555–565. 10.1038/s41587-021-01094-0 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 67. Stringer C, Wang T, Michaelos M, Pachitariu M. Cellpose: a generalist algorithm for cellular segmentation. Nat Methods. 2021;18(1):100–106. 10.1038/s41592-020-01018-x [ DOI ] [ PubMed ] [ Google Scholar ] 68. Schmidt U, Weigert M, Broaddus C, Myers G. Cell detection with star-convex polygons. In: Frangi AF, Schnabel JA, Davatzikos C, Alberola-López C, Fichtinger G, eds. Medical Image Computing and Computer Assisted Intervention–MICCAI. Vol 11071. Springer; 2018:265–273. 10.1007/978-3-030-00934-2_30 [ DOI ] [ Google Scholar ] 69. Weigert M, Schmidt U, Haase R, Sugawara K, Myers G. Star-convex polyhedra for 3D object detection and segmentation in microscopy. In: 2020 IEEE Winter Conference on Applications of Computer Vision (WACV). 2020: 3655–3662. 10.1109/WACV45572.2020.9093435 [ DOI ] [ Google Scholar ] 70. Lee MY, Bedia JS, Bhate SS, et al. CellSeg: a robust, pre-trained nucleus segmentation and pixel quantification software for highly multiplexed fluorescence images. BMC Bioinformatics. 2022;23(1):46. 10.1186/s12859-022-04570-9 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 71. Berg S, Kutra D, Kroeger T, et al. Ilastik: interactive machine learning for (bio)image analysis. Nat Methods. 2019;16(12):1226–1232. 10.1038/s41592-019-0582-9 [ DOI ] [ PubMed ] [ Google Scholar ] 72. Yapp C, Novikov E, Jang WD, et al. UnMICST: Deep learning with real augmentation for robust segmentation of highly multiplexed images of human tissues. Commun Biol. 2022;5(1):1–13. 10.1038/s42003-022-04076-3 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 73. Caicedo JC, Goodman A, Karhohs KW, et al. Nucleus segmentation across imaging experiments: the 2018 Data Science Bowl. Nat Methods. 2019;16 (12):1247–1253. 10.1038/s41592-019-0612-7 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 74. Kromp F, Bozsaky E, Rifatbegovic F, et al. An annotated fluorescence image dataset for training nuclear segmentation methods. Sci Data. 2020;7:262. 10.1038/s41597-020-00608-w [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 75. Ghahremani P, Li Y, Kaufman A, et al. Deep learning-inferred multiplex immunofluorescence for immunohistochemical image quantification. Nat Mach Intell. 2022;4(4):401–412. 10.1038/s42256-022-00471-x [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 76. Han W, Cheung AM, Yaffe MJ, Martel AL. Cell segmentation for immunofluorescence multiplexed images using two-stage domain adaptation and weakly labeled data for pre-training. Sci Rep. 2022;12(1):4399. 10.1038/s41598-022-08355-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 77. Eng J, Bucher E, Hu Z, et al. A framework for multiplex imaging optimization and reproducible analysis. Commun Biol. 2022;5(1):438. 10.1038/s42003-022-03368-y [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 78. Windhager J, Zanotelli VRT, Schulz D, et al. An end-to-end workflow for multiplexed image processing and analysis. Nat Protoc. 2023;18(11):3565–3613. 10.1038/s41596-023-00881-0 [ DOI ] [ PubMed ] [ Google Scholar ] 79. Jha SK, Mishra P, Mathur S, et al. Development and validation of fully automatic deep learning-based algorithms for immunohistochemistry reporting of invasive breast ductal carcinoma. arXiv:2406.10893. 2024. 10.48550/arXiv.2406.10893 [ DOI ] [ Google Scholar ] 80. Kildal W, Cyll K, Kalsnes J, et al. Deep learning for automated scoring of immunohistochemically stained tumour tissue sections – Validation across tumour types based on patient outcomes. Heliyon. 2024;10(13):e32529. 10.1016/j.heliyon.2024.e32529 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 81. Wen Z, Luo D, Wang S, et al. Deep learning-based H-score quantification of immunohistochemistry-stained images. Mod Pathol. 2024;37(2):100398. 10.1016/j.modpat.2023.100398 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 82. Hoölscher DL, Bouteldja N, Joodaki M, et al. Next-generation morphometry for pathomics-data mining in histopathology. Nat Commun. 2023;14(1):470. 10.1038/s41467-023-36173-0 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 83. Tommasino C, Merolla F, Russo C, Staibano S, Rinaldi AM. Histopathological image deep feature representation for CBIR in smart PACS. J Digit Imaging. 2023;36(5):2194–2209. 10.1007/s10278-023-00832-x [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 84. Yang P, Zhai Y, Li L, et al. A deep metric learning approach for histopathological image retrieval. Methods. 2020;179:14–25. 10.1016/j.ymeth.2020.05.015 [ DOI ] [ PubMed ] [ Google Scholar ] 85. Lu MY, Chen B, Williamson DFK, et al. A visual-language foundation model for computational pathology. Nat Med. 2024;30(3):863–874. 10.1038/s41591-024-02856-4 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 86. Xiang J, Wang X, Zhang X, et al. A vision–language foundation model for precision oncology. Nature. 2025;638(8051):769–778. 10.1038/s41586-024-08378-w [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 87. Huang Z, Bianchi F, Yuksekgonul M, Montine TJ, Zou J. A visual–language foundation model for pathology image analysis using medical Twitter. Nat Med. 2023;29(9):2307–2316. 10.1038/s41591-023-02504-3 [ DOI ] [ PubMed ] [ Google Scholar ] 88. Silina K, Ciompi F. Hitchhiker’s guide to cancer-associated lymphoid aggregates in histology images: manual and deep learning-based quantification approaches. arXiv e-prints. 2024. 10.48550/arXiv.2403.04142 [ DOI ] [ PubMed ] [ Google Scholar ] 89. Liu Y, Hua X, Zhu S, et al. Automated identification and quantification of myocardial inflammatory infiltration in digital histological images to diagnose myocarditis. arXiv:2307.01098. 2023. 10.48550/arXiv.2307.01098 [ DOI ] [ Google Scholar ] 90. Haghighi F, Ghosh S, Ngu H, et al. Self-supervised learning for segmentation and quantification of dopamine neurons in Parkinson’s disease. arXiv. 2301:08141. 2023. 10.48550/arXiv.2301.08141 [ DOI ] [ Google Scholar ] 91. Eling N, Damond N, Hoch T, Bodenmiller B. Cytomapper: an R/bioconductor package for visualization of highly multiplexed imaging data. Bioinformatics. 2021;36(24):5706–5708. 10.1093/bioinformatics/btaa1061 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 92. Samusik N, Good Z, Spitzer MH, Davis KL, Nolan GP. Automated mapping of phenotype space with single-cell data. Nat Methods. 2016;13(6):493–496. 10.1038/nmeth.3863 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 93. DiGiuseppe JA, Cardinali JL, Rezuke WN, Pe’er D. PhenoGraph and viSNE facilitate the identification of abnormal T-cell populations in routine clinical flow cytometric data. Cytometry B Clin Cytom. 2018;94(5):744–757. 10.1002/cyto.b.21588 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 94. Brummelman J, Haftmann C, Núñez NG, et al. Development, application and computational analysis of high-dimensional fluorescent antibody panels for single-cell flow cytometry. Nat Protoc. 2019;14(7):1946–1969. 10.1038/s41596-019-0166-2 [ DOI ] [ PubMed ] [ Google Scholar ] 95. Hickey JW, Tan Y, Nolan GP, Goltsev Y. Strategies for accurate cell type identification in CODEX multiplexed imaging data. Front Immunol. 2021;12. 10.3389/fimmu.2021.727626 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 96. Hao Y, Stuart T, Kowalski MH, et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat Biotechnol. 2024;42(2):293–304. 10.1038/s41587-023-01767-y [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 97. Stuart T, Butler A, Hoffman P, et al. Comprehensive integration of single-cell data. Cell. 2019;177(7):1888–1902.e21. 10.1016/j.cell.2019.05.031 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 98. Wolf FA, Angerer P, Theis FJ. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol. 2018;19(1):15. 10.1186/s13059-017-1382-0 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 99. Dries R, Zhu Q, Dong R, et al. Giotto: a toolbox for integrative analysis and visualization of spatial expression data. Genome Biol. 2021;22(1):78. 10.1186/s13059-021-02286-2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 100. Somarakis A, Van Unen V, Koning F, Lelieveldt B, Hoöllt T. ImaCytE: visual exploration of cellular micro-environments for imaging mass cytometry data. IEEE Trans Vis Comput Graph. 2021;27(1):98–110. 10.1109/TVCG.2019.2931299 [ DOI ] [ PubMed ] [ Google Scholar ] 101. Liu X, Song W, Wong BY, et al. A comparison framework and guideline of clustering methods for mass cytometry data. Genome Biol. 2019;20(1):297. 10.1186/s13059-019-1917-7 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 102. Seal S, Wrobel J, Johnson AM, et al. On clustering for cell-phenotyping in multiplex immunohistochemistry (mIHC) and multiplexed ion beam imaging (MIBI) data. BMC Res Notes. 2022;15(1):215. 10.1186/s13104-022-06097-x [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 103. Abdullahi Sidi F, Bingham V, Craig SG, et al. PD-L1 multiplex and quantitative image analysis for molecular diagnostics. Cancers. 2021;13(1):29. 10.3390/cancers13010029 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 104. Pugh M, Akarka AU, Hunter K, Dojcinov S, Marafioti T. Multiplex immunohistochemistry in lymphoma pathology: a research tool for study of the immune microenvironment. Diagn Histopathol. 2020;26(9):407–420. 10.1016/j.mpdhp.2020.07.002 [ DOI ] [ Google Scholar ] 105. Fassler DJ, Torre-Healy LA, Gupta R, et al. Spatial characterization of tumor-infiltrating lymphocytes and breast cancer progression. Cancers (Basel). 2022;14(9):2148. 10.3390/cancers14092148 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 106. Zhang L, Conejo-Garcia JR, Katsaros D, et al. Intratumoral T cells, recurrence, and survival in epithelial ovarian cancer. N Engl J Med. 2003;348(3):203–213. 10.1056/NEJMoa020177 [ DOI ] [ PubMed ] [ Google Scholar ] 107. Gorchs L, Kaipe H. Interactions between cancer-associated fibroblasts and T cells in the pancreatic tumor microenvironment and the role of chemokines. Cancers (Basel). 2021;13(12):2995. 10.3390/cancers13122995 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 108. Öhlund D, Handly-Santana A, Biffi G, et al. Distinct populations of inflammatory fibroblasts and myofibroblasts in pancreatic cancer. J Exp Med. 2017;214(3):579–596. 10.1084/jem.20162024 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 109. Pakula H, Omar M, Carelli R, et al. Distinct mesenchymal cell states mediate prostate cancer progression. Nat Commun. 2024;15(1):363. 10.1038/s41467-023-44210-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 110. Tan WCC, Nerurkar SN, Cai HY, et al. Overview of multiplex immunohistochemistry/immunofluorescence techniques in the era of cancer immunotherapy. Cancer Commun. 2020;40(4):135–153. 10.1002/cac2.12023 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 111. Locke D, Hoyt CC. Companion diagnostic requirements for spatial biology using multiplex immunofluorescence and multispectral imaging. Front Mol Biosci. 2023;10. 10.3389/fmolb.2023.1051491 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 112. Umeton R, Kwok A, Maurya R, et al. GPT-4 in a cancer center–institute-wide deployment challenges and lessons learned. NEJM AI. 2024;1(4):AIcs2300191. 10.1056/AIcs2300191 [ DOI ] [ Google Scholar ] 113. Omar M, Ullanat V, Loda M, Marchionni L, Umeton R. ChatGPT for digital pathology research. Lancet Digit Health. 2024;6(8):e595–e600. 10.1016/S2589-7500(24)00114-6 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 114. Palla G, Spitzer H, Klein M, et al. Squidpy: a scalable framework for spatial omics analysis. Nat Methods. 2022;19(2):171–178. 10.1038/s41592-021-01358-2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 115. Czech E, Aksoy BA, Aksoy P, Hammerbacher J. Cytokit: a single-cell analysis toolkit for high dimensional fluorescent microscopy imaging. BMC Bioinformatics. 2019;20(1):448. 10.1186/s12859-019-3055-3 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 116. Bortolomeazzi M, Montorsi L, Temelkovski D, et al. A SIMPLI (single-cell identification from MultiPLexed images) approach for spatially-resolved tissue phenotyping at single-cell resolution. Nat Commun. 2022;13(1):781. 10.1038/s41467-022-28470-x [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 117. Jalili V, Afgan E, Gu Q, et al. The Galaxy platform for accessible, reproducible and collaborative biomedical analyses: 2020 update. Nucleic Acids Research. 2020;48(W1):W395–W402. 10.1093/nar/gkaa434 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 118. Di Tommaso P, Chatzou M, Floden EW, Barja PP, Palumbo E, Notredame C. Nextflow enables reproducible computational workflows. Nat Biotechnol. 2017;35(4):316–319. 10.1038/nbt.3820 [ DOI ] [ PubMed ] [ Google Scholar ] 119. Muhlich JL, Chen YA, Yapp C, Russell D, Santagata S, Sorger PK. Stitching and registering highly multiplexed whole-slide images of tissues and tumors using ASHLAR. Bioinformatics. 2022;38(19):4613–4621. 10.1093/bioinformatics/btac544 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 120. Peng T, Thorn K, Schroeder T, et al. A BaSiC tool for background and shading correction of optical microscopy images. Nat Commun. 2017;8(1):14836. 10.1038/ncomms14836 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 121. McQuin C, Goodman A, Chernyshev V, et al. CellProfiler 3.0: Next-generation image processing for biology. PLOS Biology. 2018;16(7):e2005970. 10.1371/journal.pbio.2005970 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 122. Schapiro D, Jackson HW, Raghuraman S, et al. histoCAT: analysis of cell phenotypes and interactions in multiplex image cytometry data. Nat Methods. 2017;14(9):873–876. 10.1038/nmeth.4391 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 123. Bodenheimer T, Halappanavar M, Jefferys S, et al. FastPG: Fast clustering of millions of single cells. bioRxiv. 2020:2020.06.19.159749 10.1101/2020.06.19.159749 [ DOI ] [ Google Scholar ] ACTIONS View on publisher site PDF (531.4 KB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top