Reconstructing cell cycle and disease progression using deep learning

Eulenberg, Philipp; Köhler, Niklas; Blasi, Thomas; Filby, Andrew; Carpenter, Anne E.; Rees, Paul; Theis, Fabian J.; Wolf, F. Alexander

doi:10.1038/s41467-017-00623-3

Cited by 243 publications

(206 citation statements)

References 24 publications

Supporting

Mentioning

188

Contrasting

Unclassified

Order By: Relevance

“…DL approaches are able to autonomously extract relevant information from stain‐free imagery, a conclusion that is supported by previous work . However, they do not outperform the classical approaches, which achieve the best results for both data sets.…”

Section: Discussionsupporting

confidence: 67%

Classification of Human White Blood Cells Using Machine Learning for Stain‐Free Imaging Flow Cytometry

et al. 2019

Self Cite

View full text Add to dashboard Cite

Imaging flow cytometry (IFC) produces up to 12 spectrally distinct, information‐rich images of single cells at a throughput of 5,000 cells per second. Yet often, cell populations are still studied using manual gating, a technique that has several drawbacks, hence it would be advantageous to replace manual gating with an automated process. Ideally, this automated process would be based on stain‐free measurements, as the currently used staining techniques are expensive and potentially confounding. These stain‐free measurements originate from the brightfield and darkfield image channels, which capture transmitted and scattered light, respectively. To realize this automated, stain‐free approach, advanced machine learning (ML) methods are required. Previous works have successfully tested this approach on cell cycle phase classification with both a classical ML approach based on manually engineered features, and a deep learning (DL) approach. In this work, we compare both approaches extensively on the problem of white blood cell classification. Four human whole blood samples were assayed on an ImageStream‐X MK II imaging flow cytometer. Two samples were stained for the identification of eight white blood cell types, while two other sample sets were stained for the identification of resting and active eosinophils. For both data sets, four ML classifiers were evaluated on stain‐free imagery with stratified 5‐fold cross‐validation. On the white blood cell data set, the best obtained results were 0.778 and 0.703 balanced accuracy for classical ML and DL, respectively. On the eosinophil data set, this was 0.871 and 0.856 balanced accuracy. We conclude that classifying cell types based on only stain‐free images is possible with all four classifiers. Noteworthy, we also find that the DL approaches tested in this work do not outperform the approaches based on manually engineered features. © 2019 International Society for Advancement of Cytometry

show abstract

Section: Discussionsupporting

confidence: 67%

Classification of Human White Blood Cells Using Machine Learning for Stain‐Free Imaging Flow Cytometry

et al. 2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…Previous studies have established the capabilities of such techniques in label-free cell cycle analysis (50,51) and combined with the application of powerful IFC software platforms such as CellProfiler and CellProfiler Analyst (52). Approaches such as machine learning and deep learning techniques can potentially be used to explore all intuitive and non-intuitive measures of classifying RBC images.…”

Section: Practical Implementationmentioning

confidence: 99%

Label‐Free Analysis of Red Blood Cell Storage Lesions Using Imaging Flow Cytometry

et al. 2019

View full text Add to dashboard Cite

Deleterious changes, collectively termed as storage lesions, alter the characteristics of red blood cell (RBC) morphology during in vitro storage. Due to gradual loss of cellular membrane, RBCs lose their original biconcave disk shape and begin a process of spherical deformation that is characterized by well-defined morphological criteria. At the spheroechinocyte stage, the structure of RBC is irreversibly damaged and lacks the elasticity necessary to efficiently deliver oxygen. Quantifying the prevalence of spheroechinocytes could provide an important morphological measure of the quality of stored blood products. Unlike the conventional RBC morphology characterization assay involving light microscopy, we introduce a label-free assay using imaging flow cytometry (IFC). The technique captures 100,000 images of a sample and calculates a relative measure of spheroechinocyte population in a fraction of the time required by the conventional method. A comparative method study, measuring a morphological index for 11 RCC units through storage, found that the two techniques measured similar trends with IFC reporting the metric at an average of 3.9% higher. We monitored 18 RCC units between Weeks 1 and 6 of storage and found that the spheroechinocyte population increased by an average of 26.2%. The large (3.5-64.1%) variation between the units' spheroechinocyte population percentage at Week 1 suggests a possible dependence of blood product quality on donor characteristics. Given our method's ability to rapidly monitor large samples and refine morphological characterization beyond conventional methods, we believe our technique offers good potential for studying the underlying relationships between RBC morphology and blood storage lesions.Key terms deformability; erythrocyte; imaging flow cytometry; red blood cell morphology; storage lesion CURRENTLY, blood banks and hospital laboratories are devoid of convenient, label-free techniques to measure the morphological distribution of red blood cells (RBCs) stored in vitro. Parameters such as membrane shape and deformability are critical for understanding an RBC's functional efficacy as an oxygen transporter (1). Microscopy-related assays, considered the gold standard for RBC morphology analysis, stain and smear samples prior to manual characterization. High-throughput assays that can measure RBCs within environments closely resembling in vivo conditions could obtain a more insightful assessment of the morphology distribution.RBCs are stored in vitro primarily to enhance the oxygen carrying capacity of vulnerable patients via transfusions (2). Blood banks extract RBCs from donated blood using customized separation techniques and suspend the cells in a preservative fluid to increase its shelf-life (3). During storage of this red cell concentrate (RCC), typically set at a maximum of 6 weeks, RBCs undergo a sequence of significant biochemical and biomechanical changes, which have been collectively termed hypothermic storage lesions (4,5). RBC storage lesions cause the formation of m...

show abstract

“…These approaches have shown impressive results in object detection and quantification, even in complex biomedical images (5,6). Deep learning can also be utilized as an unsupervised feature extractor or phenotypic classifier.…”

Section: The Use Of Machine Learning In Object Detection and Quantifimentioning

confidence: 99%

Zooming in on adipocytes: High and deep

Doan

2017

Cytometry Pt A

View full text Add to dashboard Cite

show abstract

Reconstructing cell cycle and disease progression using deep learning

Cited by 243 publications

References 24 publications

Classification of Human White Blood Cells Using Machine Learning for Stain‐Free Imaging Flow Cytometry

Classification of Human White Blood Cells Using Machine Learning for Stain‐Free Imaging Flow Cytometry

Label‐Free Analysis of Red Blood Cell Storage Lesions Using Imaging Flow Cytometry

Zooming in on adipocytes: High and deep

Contact Info

Product

Resources

About