This R/Bioconductor package supports interactive visualization of multi-channel images and segmentation masks generated by imaging mass cytometry and other highly multiplexed imaging techniques using shiny. The cytoviewer interface is divided into image-level (Composite and Channels) and cell-level visualization (Masks). It allows users to overlay individual images with segmentation masks, integrates well with SingleCellExperiment and SpatialExperiment objects for metadata visualization and supports image downloads.
cytoviewer 1.6.0
This vignette introduces the cytoviewer
package for interactive
multi-channel image visualization. Images as well as corresponding
segmentation masks generated by imaging mass cytometry (IMC) and other
highly multiplexed imaging techniques can be interactively visualized
and explored.
The cytoviewer
package builds on top of the cytomapper
Bioconductor package (Eling et al. 2020) and extends the static visualization strategies
provided by cytomapper
via an interactive Shiny application. The
cytoviewer
package leverages the image handling, analysis and
visualization strategies provided by the EBImage
Bioconductor package and offers interactive image visualization
strategies in a similar fashion as iSEE for single-cell
data. In addition, building up on SingleCellExperiment,
SpatialExperiment and cytomapper::CytoImageList
classes, the cytoviewer
package integrates into the Bioconductor
framework for single-cell and image analysis.
Read the pre-print here.
Highly multiplexed imaging allows simultaneous spatially and single-cell resolved detection of dozens of biological molecules (e.g. proteins) in their native tissue context. As a result, these technologies allow an in-depth analysis of complex systems and diseases such as the tumor microenvironment (Jackson et al. 2020) and type 1 diabetes progression (Damond et al. 2019).
Imaging-based spatial proteomics methods (Moffitt, Lundberg, and Heyn 2022) can be broadly divided into fluorescent cyclic approaches such as tissue-based cyclic immunofluorescence (t-CyCIF) (Lin et al. 2018) and one-step mass-tag based approaches such as multiplexed ion beam imaging (MIBI) (Angelo et al. 2014) and IMC (Giesen et al. 2014).
Of note, the instructions below will focus on the visualization and
exploration of IMC data as an example. However, data from other
technologies such as t-CyCIF or MIBI, which produce pixel-level
intensities and (optionally) segmentation masks, can be interactively
visualized with cytoviewer
as long as they have the appropriate input
format (see Section Data input format).
IMC, an advancement of CyTOF, combines antibodies tagged with isotopically pure rare earth metals with laser ablation and mass-spectrometry-based detection to produce high-dimensional images (Giesen et al. 2014). It captures the spatial expression of over 40 proteins in parallel at a sub-cellular resolution of 1 μm. Thus, IMC is able to detect cytoplasmic and nuclear localization of proteins.
To fully leverage the information contained in IMC and multiplexed imaging data in general, computational tools are of key importance.
The main analysis steps, irrespective of the biological question, include 1) Visual inspection of images for quality control, 2) Image pre-processing and segmentation and 3) Single-cell and spatial analysis (Windhager, Bodenmiller, and Eling 2021).
A comprehensive end-to-end workflow for multiplexed image processing and analysis with detailed information for every analysis step can be found here.
Importantly, the cytoviewer
package can support, simplify and improve
any of these analysis steps with its easy-to-use interactive
visualization interface in R.
Below we will showcase an example workflow that
highlights the different functionality and potential application fields
of cytoviewer
.
The cytoviewer
interface is broadly divided into
image-level (Composite and Channels) and
cell-level visualization (Masks). It allows users to
overlay individual images with segmentation masks, integrates well with
SingleCellExperiment
and SpatialExperiment
objects for metadata
visualization and supports image downloads (Figure 2B).
The cytoviewer
package combines objects of
SingleCellExperiment, SpatialExperiment
and cytomapper::CytoImageList
classes (from cytomapper)
to visualize image- and cell-level information.
The cytoviewer
function takes up to five arguments.
Firstly, image
refers to a CytoImageList
object containing one or
multiple multi-channel images where each channel represents the
pixel-intensities of one marker (proteins in IMC).
Secondly, mask
refers to a CytoImageList
object containing one or
multiple segmentation masks. Segmentation masks are defined as
one-channel images containing integer values, which represent the cell
ids or background.
Thirdly, the object
entry refers to a SingleCellExperiment
or
SpatialExperiment
class object that contains cell-specific metadata in
the colData
slot.
Lastly, to match information between the CytoImageList
objects and the
SingleCellExperiment
/SpatialExperiment
object, two additional spots
can be specified:
img_id
: a single character indicating the colData
(of the
SingleCellExperiment
/SpatialExperiment
object) and
elementMetadata
(of the CytoImageList
object) entry that
contains the image identifiers. These image ids have to match
between the SingleCellExperiment
/ SpatialExperiment
object and
the CytoImageList
objects.
cell_id
: a single character indicating the colData
entry that
contains the cell identifiers. These should be integer values
corresponding to pixel-values in the segmentation masks.
For more detailed information on the input objects, please refer to the respective documentation (e.g. the vignettes of the cytomapper or SingleCellExperiment/ SpatialExperiment packages).
In the Read in data section, we provide
example code to directly read in images and masks (e.g. in .tiff format)
into a CytoImageList
object and create a SingleCellExperiment
object
from them, which we can then visualize with cytoviewer
.
The functionality of cytoviewer
depends on which input objects are
user-provided. Below we describe the four use cases in respect to
input objects and functionality.
1. Usage of cytoviewer with images, masks and object
The full functionality of cytoviewer can be leveraged when image
,
mask
and object
are provided, which is the main intended use case.
This allows image-level visualization (Composite and Channels), cell-level visualization, overlaying images with segmentation masks as well as metadata visualization.
2. Usage of cytoviewer with images only
If only the image
object is specified, image-level
visualization (Composite and Channels) is possible.
3. Usage of cytoviewer with images and masks
Image-level visualization (Composite and Channels),
overlaying of images with segmentation masks and
cell-level visualization is feasible when image
and
mask
objects are provided.
4. Usage of cytoviewer with masks and object
If mask
and object
are specified, cell-level
visualization as well as metadata visualization is possible.
The cytoviewer
package can be installed from Bioconductor
via:
if (!requireNamespace("BiocManager", quietly = TRUE))
install.packages("BiocManager")
BiocManager::install("cytoviewer")
The development version of cytoviewer
can be installed from Github
via:
if (!requireNamespace("remotes", quietly = TRUE))
install.packages("remotes")
remotes::install_github("BodenmillerGroup/cytoviewer")
To load the package in your R session, type the following:
library(cytoviewer)
For visualization purposes, we will use a toy dataset provided by the cytomapper package.
The dataset contains 3 images of \(100\mu{m}\) x \(100\mu{m}\) dimensions with 362 segmented cells and pixel-intensities for 5 proteins: H3, CD99, PIN, CD8a, and CDH. It is a small subset from a Type 1 Diabetes dataset (Damond et al. 2019).
Pixel-level intensities for all 5 markers (5 channels) are stored in the
pancreasImages
object.
The corresponding segmentation masks are stored in the pancreasMasks
object.
All cell-specific metadata are stored in the colData
slot of the
corresponding SingleCellExperiment
object: pancreasSCE
.
For more detailed information on the dataset, please refer to the
respective documentation (e.g. via ?pancreasImages
or the vignette of
the cytomapper package).
We also provide example code to directly read in images and masks
(e.g. in .tiff format) into a CytoImageList
object and create a
SingleCellExperiment
object from them in the
Read in data section.
# Load example datasets
library(cytomapper)
data("pancreasImages")
data("pancreasMasks")
data("pancreasSCE")
pancreasImages
## CytoImageList containing 3 image(s)
## names(3): E34_imc G01_imc J02_imc
## Each image contains 5 channel(s)
## channelNames(5): H3 CD99 PIN CD8a CDH
pancreasMasks
## CytoImageList containing 3 image(s)
## names(3): E34_mask G01_mask J02_mask
## Each image contains 1 channel
pancreasSCE
## class: SingleCellExperiment
## dim: 5 362
## metadata(0):
## assays(2): counts exprs
## rownames(5): H3 CD99 PIN CD8a CDH
## rowData names(4): MetalTag Target clean_Target frame
## colnames(362): E34_824 E34_835 ... J02_4190 J02_4209
## colData names(9): ImageName Pos_X ... MaskName Pattern
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):
Here as an example, we call cytoviewer
with image
, mask
and
object
data to leverage all provided functionality.
This setting allows image-level visualization (Composite and Channels), cell-level visualization, overlaying images with segmentation masks as well as metadata visualization.
For further details, please refer to the ?cytoviewer
manual or the
Help page
within the shiny application.
# Use cytoviewer with images, masks and object
app <- cytoviewer(image = pancreasImages,
mask = pancreasMasks,
object = pancreasSCE,
img_id = "ImageNb",
cell_id = "CellNb")
if (interactive()) {
shiny::runApp(app, launch.browser = TRUE)
}
The cytoviewer
interface is divided into a Header, Sidebar and
Body section (see Figure below).
The Header includes package version information, access to session information and the help page as well as a dropdown-menu for image downloads.
The Body features a Tabset-Panel layout allowing the user to switch between three image modes: Image-level (Composite and Channels) and Cell-level (Mask). Furthermore, the Composite and Mask tabs have zoom controls.
The Sidebar panel is subdivided into four sections: Sample selection, Image-level, Cell-level and General controls.
Image visualization control is split into basic and advanced controls.
Basic controls supports the selection of up to six markers/channels
for image
display. Each marker has color control settings that allow
the user to set contrast, brightness, gamma and select a channel color.
In the advanced controls part, the user can choose to overlay the
displayed images with provided segmentation masks
. Outline color and
mask thickness can be adjusted by the user. Moreover, the masks can be
outlined by cell-specific metadata provided in colData
slot of the
object
.
Of note, for categorical and continuous metadata entries the user can choose between discrete colors and continuous color palettes (viridis, inferno, plasma), respectively.
Cell visualization has basic controls.
Here, the user can choose to display the provided segmentation masks
.
If an object
is provided, the masks can be colored by cell-specific
metadata.
Please note again that for categorical and continuous metadata entries the user can choose between discrete colors and continuous color palettes (viridis, inferno, plasma), respectively.
General controls is subdivided into an Image appearance and Image filters part.
In the Image appearance section, the user can adjust the scale bar length and include legend/image titles, while the Image filters section allows to control pixel-wise interpolation (default) and apply a Gaussian filter.
The cytoviewer
package supports fast and uncomplicated image
downloads.
Download controls are part of the Header (see Section Interface).
The user can specify a file name, select the image of interest (Composite, Channels, Mask) and the file format (pdf, png). Upon clicking the download button, a pop-window should appear where the user can specify the download location.
To conveniently read in images and segmentation masks into a CytoImageList
object and then visualize these using cytoviewer
,
the cytomapper
package provides the loadImages
function.
The loadImages
function returns a CytoImageList
object containing the
multi-channel images or segmentation masks. Refer to the ?loadImages
function
to see the full functionality.
As an example, we will read in multi-channel images and segmentation masks
provided by the cytomapper
package.
To correctly scale pixel values of the segmentation masks when reading them in,
we will need to set as.is = TRUE
. Users needs to take care that pixel values
are scaled correctly in more complex cases.
library(cytomapper)
# Data directory that stores images and masks in tiff format
data_path <- system.file("extdata", package = "cytomapper")
# Read in images
cur_images <- loadImages(data_path, pattern = "_imc.tiff")
cur_images
## CytoImageList containing 3 image(s)
## names(3): E34_imc G01_imc J02_imc
## Each image contains 5 channel(s)
# Read in masks
cur_masks <- loadImages(data_path, pattern = "_mask.tiff", as.is = TRUE)
cur_masks
## CytoImageList containing 3 image(s)
## names(3): E34_mask G01_mask J02_mask
## Each image contains 1 channel
To link images between the two CytoImageList
objects and the corresponding
SingleCellExperiment
object, the image ids need to be added to the
elementMetadata
slot of the CytoImageList
objects.
names(cur_images)
## [1] "E34_imc" "G01_imc" "J02_imc"
names(cur_masks)
## [1] "E34_mask" "G01_mask" "J02_mask"
mcols(cur_masks)$ImageNb <- mcols(cur_images)$ImageNb <- c("E34", "G01", "J02")
To access the correct images in the multi-channel CytoImageList
object, the
user needs to set the correct channel names. For this, the cytomapper
package
provides the ?channelNames
getter and setter function:
channelNames(cur_images) <- c("H3", "CD99", "PIN", "CD8a", "CDH")
Based on the processed segmentation masks and multi-channel images,
cytomapper
can be used to measure cell-specific intensities and morphological features.
Here, these features are stored in form of a SingleCellExperiment
object.
cur_sce <- measureObjects(image = cur_images,
mask = cur_masks,
img_id = "ImageNb")
cur_sce
## class: SingleCellExperiment
## dim: 5 362
## metadata(0):
## assays(1): counts
## rownames(5): H3 CD99 PIN CD8a CDH
## rowData names(0):
## colnames: NULL
## colData names(8): ImageNb object_id ... m.majoraxis m.eccentricity
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):
Next, we can again call cytoviewer
with the generated image
, mask
and
object
data and leverage all provided functionality.
# Use cytoviewer with images, masks and object
app_1 <- cytoviewer(image = cur_images,
mask = cur_masks,
object = cur_sce,
img_id = "ImageNb",
cell_id = "object_id")
if (interactive()) {
shiny::runApp(app_1, launch.browser = TRUE)
}
For more detailed information on the input objects, please refer to the respective documentation (the vignettes of the cytomapper or SingleCellExperiment/ SpatialExperiment packages).
## R version 4.4.1 (2024-06-14)
## Platform: x86_64-pc-linux-gnu
## Running under: Ubuntu 24.04.1 LTS
##
## Matrix products: default
## BLAS: /home/biocbuild/bbs-3.20-bioc/R/lib/libRblas.so
## LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.12.0
##
## locale:
## [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
## [3] LC_TIME=en_GB LC_COLLATE=C
## [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
## [7] LC_PAPER=en_US.UTF-8 LC_NAME=C
## [9] LC_ADDRESS=C LC_TELEPHONE=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## time zone: America/New_York
## tzcode source: system (glibc)
##
## attached base packages:
## [1] stats4 stats graphics grDevices utils datasets methods
## [8] base
##
## other attached packages:
## [1] cytomapper_1.18.0 SingleCellExperiment_1.28.0
## [3] SummarizedExperiment_1.36.0 Biobase_2.66.0
## [5] GenomicRanges_1.58.0 GenomeInfoDb_1.42.0
## [7] IRanges_2.40.0 S4Vectors_0.44.0
## [9] BiocGenerics_0.52.0 MatrixGenerics_1.18.0
## [11] matrixStats_1.4.1 EBImage_4.48.0
## [13] cytoviewer_1.6.0 BiocStyle_2.34.0
##
## loaded via a namespace (and not attached):
## [1] bitops_1.0-9 gridExtra_2.3 rlang_1.1.4
## [4] magrittr_2.0.3 svgPanZoom_0.3.4 shinydashboard_0.7.2
## [7] compiler_4.4.1 systemfonts_1.1.0 png_0.1-8
## [10] fftwtools_0.9-11 vctrs_0.6.5 pkgconfig_2.0.3
## [13] SpatialExperiment_1.16.0 crayon_1.5.3 fastmap_1.2.0
## [16] magick_2.8.5 XVector_0.46.0 fontawesome_0.5.2
## [19] utf8_1.2.4 promises_1.3.0 rmarkdown_2.28
## [22] UCSC.utils_1.2.0 ggbeeswarm_0.7.2 xfun_0.48
## [25] zlibbioc_1.52.0 cachem_1.1.0 jsonlite_1.8.9
## [28] later_1.3.2 rhdf5filters_1.18.0 DelayedArray_0.32.0
## [31] Rhdf5lib_1.28.0 BiocParallel_1.40.0 jpeg_0.1-10
## [34] tiff_0.1-12 terra_1.7-83 parallel_4.4.1
## [37] R6_2.5.1 bslib_0.8.0 RColorBrewer_1.1-3
## [40] jquerylib_0.1.4 Rcpp_1.0.13 bookdown_0.41
## [43] knitr_1.48 httpuv_1.6.15 Matrix_1.7-1
## [46] nnls_1.6 tidyselect_1.2.1 viridis_0.6.5
## [49] abind_1.4-8 yaml_2.3.10 codetools_0.2-20
## [52] miniUI_0.1.1.1 lattice_0.22-6 tibble_3.2.1
## [55] shiny_1.9.1 evaluate_1.0.1 archive_1.1.9
## [58] shinycssloaders_1.1.0 pillar_1.9.0 BiocManager_1.30.25
## [61] generics_0.1.3 sp_2.1-4 RCurl_1.98-1.16
## [64] ggplot2_3.5.1 munsell_0.5.1 scales_1.3.0
## [67] xtable_1.8-4 glue_1.8.0 tools_4.4.1
## [70] colourpicker_1.3.0 locfit_1.5-9.10 rhdf5_2.50.0
## [73] grid_4.4.1 colorspace_2.1-1 GenomeInfoDbData_1.2.13
## [76] raster_3.6-30 beeswarm_0.4.0 HDF5Array_1.34.0
## [79] vipor_0.4.7 cli_3.6.3 fansi_1.0.6
## [82] viridisLite_0.4.2 S4Arrays_1.6.0 svglite_2.1.3
## [85] dplyr_1.1.4 gtable_0.3.6 sass_0.4.9
## [88] digest_0.6.37 SparseArray_1.6.0 rjson_0.2.23
## [91] htmlwidgets_1.6.4 memoise_2.0.1 htmltools_0.5.8.1
## [94] lifecycle_1.0.4 httr_1.4.7 mime_0.12
Angelo, Michael, Sean C. Bendall, Rachel Finck, Matthew B. Hale, Chuck Hitzman, Alexander D. Borowsky, Richard M. Levenson, et al. 2014. “Multiplexed Ion Beam Imaging of Human Breast Tumors.” Nature Medicine 20 (4): 436–42.
Damond, Nicolas, Stefanie Engler, Vito R. T. Zanotelli, Denis Schapiro, Clive H. Wasserfall, Irina Kusmartseva, Harry S. Nick, et al. 2019. “A Map of Human Type 1 Diabetes Progression by Imaging Mass Cytometry.” Cell Metabolism 29 (3): 755–768.e5.
Eling, Nils, Nicolas Damond, Tobias Hoch, and Bernd Bodenmiller. 2020. “Cytomapper: An R/Bioconductor Package for Visualization of Highly Multiplexed Imaging Data.” Bioinformatics 36 (24): 5706–8. https://doi.org/10.1093/bioinformatics/btaa1061.
Giesen, Charlotte, Hao A. O. Wang, Denis Schapiro, Nevena Zivanovic, Andrea Jacobs, Bodo Hattendorf, Peter J. Schüffler, et al. 2014. “Highly Multiplexed Imaging of Tumor Tissues with Subcellular Resolution by Mass Cytometry.” Nature Methods 11 (4): 417–22.
Jackson, Hartland W., Jana R. Fischer, Vito R. T. Zanotelli, H. Raza Ali, Robert Mechera, Savas D. Soysal, Holger Moch, et al. 2020. “The Single-Cell Pathology Landscape of Breast Cancer.” Nature 578 (7796): 615–20.
Lin, Jia-Ren, Benjamin Izar, Shu Wang, Clarence Yapp, Shaolin Mei, Parin M. Shah, Sandro Santagata, and Peter K. Sorger. 2018. “Highly Multiplexed Immunofluorescence Imaging of Human Tissues and Tumors Using T-Cycif and Conventional Optical Microscopes.” eLife 7: 1–46.
Meyer, Lasse, Nils Eling, and Bernd Bodenmiller. 2023. “Cytoviewer: An R/Bioconductor Package for Interactive Visualization and Exploration of Highly Multiplexed Imaging Data.” https://doi.org/10.1101/2023.05.24.542115.
Moffitt, Jeffrey R., Emma Lundberg, and Holger Heyn. 2022. “The Emerging Landscape of Spatial Profiling Technologies.” Nature Reviews Genetics 23: 741–59.
Windhager, Jonas, Bernd Bodenmiller, and Nils Eling. 2021. “An End-to-End Workflow for Multiplexed Image Processing and Analysis.” https://doi.org/10.1101/2021.11.12.468357.