updated font sizes in readme.md figure

.dockerignore

@@ -0,0 +1,3 @@
+.venv
+__pycache__/
+*.pyc

.python-version

@@ -0,0 +1 @@
+3.12

00_index.md

@@ -1,39 +1,48 @@
 ---
-title: 'Unbiased representative images using embedded microscopy maps'
-short_title: Re-Imaging the Wheel
+title: 'Reimag(in)ing Representative Images'
+short_title: Virtual Microscopy
 numbering:
   heading_2: false
 ---
 
 +++ {"part": "abstract"}
 
-Add your abstract here.
-Avoid complicated equations / citations here, for crossref compatibility.
+Static figures are the standard interface for scientific discourse and peer review, so microscopists typically communicate experimental results through representative images and plots.
+This article focuses on virtual microscopy, a novel visualization strategy for researchers to share microscopy data in publications.
+Web-based microscopy maps improve the transparency, accessibility, and archival of bioimages as an unbiased alternative to representative images.
+These features make virtual microscopy useful for the critical appraisal of observational studies.
+Using this approach, an original claim of reovirus-induced endoplasmic reticulum remodeling was reexamined.
+Virtual microscopy subjects the results of qualitative experiments to unbiased peer review by giving readers direct access to the data.
+While data visualization can be refined to minimize subjective biases, the interpretation of data inevitably reflects the author's perspective.
+Virtual microscopy allows readers to interpret the same data as the author, which is a meaningful step forward, yet how authors determine the outcome of qualitative microscopy experiments remains unclear.
 
 +++
 
-+++{"part":"epigraph"}
++++ {"part":"epigraph"}
+
 :::{warning} Pre-print
 This article has not yet been peer-reviewed.  
 _Updated 2025 September 27_
 :::
 
 +++
 
-+++ {"part": "acknowledgements"}
++++ {"part":"acknowledgements"}
+
+We thank the [Baker Institute for Animal Health](https://www.vet.cornell.edu/departments-centers-and-institutes/baker-institute-animal-health) for supporting this unfunded project.
 
-Add your acknowledgments, if any, here.
+This work used the Open Storage Network through allocation BIO250450 from the Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS) program, which is supported by U.S. National Science Foundation grants #2138259, #2138286, #2138307, #2137603, and #2138296 [@doi:10.1145/3569951.3597559].
 
 +++
 
-+++ {"part": "competing interests"}
++++ {"part":"competing interests"}
 
 ## Competing Interests
 
-Add your competing interests, if any, here.
+All authors declared no conflicts of interest.
+
 +++
 
-## Data availability
+## Data Availability
 
-NucleusNet was [deposited](#huggingface-upload) to Hugging Face to facilitate data sharing and the repository was streamed to present images in interactive figures.
-The full collection of 250,000 microscopy fields, 1,600 stitched images and corresponding mask files, as well as the cropped images were zipped and archived at Zenodo.
+OME-Zarr data is hosted on the Open Storage Network (https://doi.org/10.6075/J0RR1ZCR), see [Metadata Availability](#metadata_page) for URLs.

01_introduction.md

@@ -1,71 +1,30 @@
 ---
 title: Introduction
 numbering:
-  enumerator: 1.%s
-label : introduction_page
+  enumerator: 0.%s
 ---
 
-Representative microscopy images are visual communication tools used by microscopists to present the results of research to other scientists.
-The earliest representative microscopy images were hand drawings in [Micrographia](<wiki:Micrographia>) in 1665 by Robert Hooke.
-Nowadays, roughly three-quarters of publications in biomedical journals report at least one microscopy image [@doi:10.7554/eLife.55133].
-In scientific journals, microscopists present stereotypical images of described phenomena in static figures.
-An un-spoken rule is that scientists trust that representative images are honest and accurate.
+Representative microscopy images are visual communication tools used by microscopists to present research results to other scientists.
+The earliest representative microscopy images were hand-drawn in Robert Hooke's [Micrographia](<wiki:Micrographia>) in 1665.
+Today, roughly three-quarters of publications in biomedical journals report at least one microscopy image [@doi:10.7554/eLife.55133].
+There is an unspoken expectation that scientists rely on these images to be honest and accurate.
 However, microscopy image selection and quantitative analyses thereof are vulnerable to bias and deception [@doi:10.1242/jcs.261567].
-Aside from issues with integrity, scientists sincerely misinterpret observations.
-The problem is that it is difficult for readers to critically evaluate representative images in static figures.
-Furthermore, archival and review of primary microscopy data is disorganized and time-consuming if at all available.
-
-Virtual nanoscopy was an innovative visualization strategy for unbiased presentation of transmission electron microscopy data [@doi:10.1083/jcb.201201140].
-It used digital map software developed for satellite images to present high-magnification tiled microscopy data in an interactive viewer, like Google Earth for cell biologists.
-In practice, it is easy to implement [@doi:10.1242/jcs.262198] but difficult to publish because scientific journals use static figures.
-This project introduces embedded microscopy maps as a proof of principle to share immunofluorescence bioimaging data without external links or software.
-
-Microscopists contribute to the [reproducibility crisis](<wiki:Replication_crisis>) in science, which is being adressed through protocol standardization [@doi:10.1111/jmi.13041].
-From the perspective of most biomedical researchers, the perceived cause is the pressure to publish [@doi:10.1371/journal.pbio.3002870].
-Large-scale replicability studies performed by scientists at Amgen and Bayer assessed that 11% [@doi:10.1038/483531a] and 25% [@doi:10.1038/nrd3439-c1] of key findings from preclinical cancer research papers were robust and replicable.
-Another reproducibility project evaluated replication success with several criteria, though the 'lowest bar' determined a replication rate of 79% for representative images [@doi:10.7554/eLife.71601]. 
-Notably, 26% of claimed effects were evidenced by a representative image without quantitative analysis.
-Consider that representative images cannot describe population variation and the uncertainty of an observation.
-
-A study critical of representative microscopy images focused on methods of objective image selection to limit bias (https://doi.org/10.1016/s0006-3495(99)77379-0).
-The authors implemented a web server that chose typical images from uploaded data, but it is now unsupported and there is no modern equivalent.
-This article started by revisiting the idea of objective image selection, as Barkley and Parker thought it was an interesting use case for artificial intelligence.
-Barkley automated the process of sample collection and image selection from a collection of one million single-cell images of the [cell nucleus](<wiki:Cell_nucleus>), called _NucleusNet_.
-Cropped images from NucleusNet were presented in interactive figures, including embedded microscopy maps using free and open-source technology [@doi:10.1242/jcs.262198].
-Representative images were selected from NucleusNet by an [autoencoder](<wiki:Autoencoder>) model.
-Autoencoders are unsupervised deep learning models that compress and reconstruct images through a vector bottleneck referred to as _latent space_.
-The structure of latent space is a [black box](<wiki:Black_box>), though it can be shaped to be more useful with the art of representation learning [@doi:10.1109/TPAMI.2013.50].
-Barkley determined that it would be irresponsible to use an uninterpretable embedding for the task of image selection. 
-To define what is representative, both the subjective determinations made by humans and the objective determinations made by autoencoders are opaque.
-Therefore, this project shifted towards demonstrations of virtual nanoscopy as an alternative to image selection.
-Barkley saw the value of using virtual nanoscopy to replicate microscopy experiments and challenged the claim of reovirus-induced endoplasmic reticulum (ER) organelle remodelling.
-[OrganelleBox](https://organellebox.sf.czbiohub.org/), was used for cross-validation and to expand the investigation to study sixteen organelles during reovirus infection to create an atlas of bioimaging data.
-
----
-
-## Contributions
-
-1. **A single-cell confocal microscopy image dataset of 1,061,277 unique cell nuclei.**
-
-2. **Demonstration of a novel use case for autoencoders: typical image selection.**
-
-3. **Embedded microscopy maps for unbiased data sharing with virtual nanoscopy**
-
-4. **An image atlas of cellular organelles during mammalian orthoreovirus infection.**
-
----
-
-## Definitions
-
-_Virtual nanoscopy_ was a term used to describe the process of automated image acquisition, followed by unbiased presentation in an interactive tile-based viewer. [@doi:10.1083/jcb.201201140]
-
-_Microscopy map_ is the interactive figure used to view virtual nanoscopy data. [@doi:10.1083/jcb.201201140]
-The original term '_electron_ microscopy map' was redefined to fit a broader use case because our study used confocal microscopy data.
-
-A _representative image_ is defined as a visual communication tool used by microscopists to communicate research findings to other scientists.
-This definition is agnostic to the research context and it emphasizes the use of pictures as scientific evidence.
-Parker suggested that _stereotypical image_ would be an apt name for the current use of representative images.
-
-A _grayscale image_ is defined as a two-dimensional matrix where each pixel is a single intensity value ranging from $0-1$ that represents the amount of light or intensity information at a specific point [@doi:10.1109/BIP60195.2023.10379342].
-
-A _latent space_ is a collection of vectors that form a reduced-dimensionality embedding of the data, fit by a machine learning model [@doi:10.1111/cgf.13672].
+Besides cherry-picking, scientists can sincerely misinterpret observations.
+The problem is that readers find it difficult to critically evaluate representative images.
+Furthermore, archival and review of primary microscopy data is disorganized and time-consuming, if at all available.
+These problems can be addressed with novel visualization strategies like [virtual microscopy](<wiki:Virtual_microscopy>) that transmit entire interactive datasets over the internet.
+An example was the unbiased presentation of electron microscopy data to readers as interactive digital maps — or microscopy maps [@doi:10.1083/jcb.201201140].
+In practice, microscopy maps are easy to share but are accessed via external links [@doi:10.1242/jcs.262198].
+
+[Reproducibility](<wiki:Reproducibility>) is an essential principle of the [scientific method](<wiki:Scientific_method>) for acquiring knowledge, yet there is a contentious [reproducibility crisis](<wiki:Replication_crisis>) in science.
+From the perspective of biomedical researchers, the perceived cause is the pressure to publish [@doi:10.1371/journal.pbio.3002870].
+Repeatability within the same study by the same researchers is distinguished from replicability by independent researchers using the same methodology.
+Large-scale replicability studies found that only 11% [@doi:10.1038/483531a] and 25% [@doi:10.1038/nrd3439-c1] of key findings from preclinical cancer research papers were replicable.
+The replication crisis is multifaceted, but the irreproducibility of representative images deserves scrutiny in its own right.
+According to one study, replication rates for representative images ranged from 27% to 90% [@doi:10.7554/eLife.71601].
+This wide range reflects how difficult it is to assess the replicability of representative images without dedicated tools.
+A feature of virtual microscopy is the unbiased sampling and transmission of multi-scale image data, making it useful as a method to replicate representative images. 
+If results are reproducible, original claims should be readily observable in microscopy maps, which capture a broader range of variation.
+Here, microscopy maps were used to reassess the claim that mammalian orthoreovirus remodels the endoplasmic reticulum (ER) of cells during infection [@doi:10.1128/mBio.01253-18].
+Using the protocols described in the original paper, hundreds of overlapping tiles were sampled at high-magnification from replicate immunofluorescence experiments on a fully motorized confocal microscope.
+Microscopy maps were embedded in the article using the next generation file format OME-Zarr [@doi:10.1007/s00418-023-02209-1] and custom viewers based on Viv [@doi:10.1038/s41592-022-01482-7].

02_approach.md

@@ -0,0 +1,97 @@
+---
+title: Approach
+numbering:
+  enumerator: 1.%s
+---
+
+Already adopted by some clinicians and educators, virtual microscopy could also benefit researchers as an interface for communicating and interpreting image data.
+Indeed, interactive microscopy maps were first proposed for precisely this purpose, as an unbiased way to present electron microscopy data [@doi:10.1083/jcb.201201140].
+Free and open-source solutions [@doi:10.1242/jcs.262198] and recent advances in file formats [@doi:10.1007/s00418-023-02209-1] now make this feasible, yet these tools have not been adopted in scientific publishing.
+Our approach combined automated image acquisition, multi-scale [pyramid](<wiki:Pyramid_(image_processing)>) formats and embedded interactive viewers ([](#abstract)).
+
+```{figure} ./figures/Virtual_microscopy.png
+:label: abstract
+Virtual microscopy. 1. Automated dataset collection with any suitable microscope. ZDC: Z-drift compensation. 2. Stitch tiles then convert the image to DZI or OME-Zarr, host online. 3. Image tiles or data chunks are retrieved from cloud storage as readers interact with microscopy maps.
+```
+
+Elemental Microscopy supports two interactive viewers; [Vizarr](https://github.com/hms-dbmi/vizarr) [@doi:10.1038/s41592-022-01482-7] and [OpenSeadragon](https://openseadragon.github.io/), enabling two methods of virtual microscopy based on OME-Zarr or Deep Zoom. 
+Authors can also build custom viewers. 
+Brodrick developed [microATLAS](https://github.com/LadInTheLab/microATLAS-widget), which was used to view OME-Zarr data in this article. 
+All three viewers are readily embeddable in Elemental Microscopy. 
+For Vizarr and OpenSeadragon, authors simply copy the markdown and replace the source URL with a link to their own data. 
+For microATLAS, authors are encouraged to use the [widget builder](https://ladinthelab.github.io/microATLAS-widget/builder.html) to generate code.
+
+## OME-Zarr
+
+OME-Zarr is a standardized and flexible image format developed for use cases like virtual microscopy [@doi:10.1007/s00418-023-02209-1].
+Most native microscopy file formats can be converted to OME-Zarr with preserved metadata using [Bio-Formats](https://www.openmicroscopy.org/bio-formats/).
+It is a chunked file type that can support multiple resolution levels.
+As an advanced implementation of microscopy maps [@doi:10.1242/jcs.262198], pyramidal OME-Zarr data can be viewed online [@doi:10.1038/s41592-022-01482-7].
+
+### Vizarr
+
+```{code}
+:label: vizarr
+:caption: Minimal markdown to embed Vizarr in Elemental Microscopy. Replace the source with your data.
+:::{any:bundle} https://curvenote.github.io/widgets/widgets/vizarr-viewer.js
+{
+  "source": "https://your-bucket.s3.amazonaws.com/sample.zarr/",
+  "height": "600px"
+}
+:::
+```
+
+### microATLAS
+
+```{code}
+:label: microatlas
+:caption: Example markdown to embed microATLAS in Elemental Microscopy. Create your own with the [builder](https://ladinthelab.github.io/microATLAS-widget/builder.html).
+:::{any:bundle} https://ladinthelab.github.io/microATLAS-widget/widget.js
+{
+  "source": "https://your-bucket.s3.amazonaws.com/sample.zarr/",
+  "views": [
+    {
+      "name": "Overview",
+      "zoom": -4,
+      "target": [4886, 4886, 0]
+    }
+  ],
+  "annotations": [
+    { "name": "Region A", "target": [2400, 2400], "color": [100, 200, 255] }
+  ],
+  "scaleBar": { "maxWidth": 120, "position": "bottom-right" }
+}
+:::
+```
+
+```{figure} ./thumbnail.gif
+:label: demo
+Interacting with microATLAS: 
+Use the menu (☰, top-left corner) to change the displayed channels, colours, or contrast.
+Scroll (↕) to zoom into cells of interest.
+```
+
+## Deep Zoom
+
+[Deep Zoom](<wiki:Deep_Zoom>) is a free and open-source solution, described elsewhere [@doi:10.1242/jcs.262198], now embeddable in Elemental Microscopy.
+Pyramidal Deep Zoom image (DZI) data hosted on Github Pages is streamed to the [OpenSeadragon](https://openseadragon.github.io/) viewer.
+By default, readers cannot control dimensions like channels or time, so it is best suited for single-channel or composite images.
+
+### OpenSeadragon
+
+```{code}
+:label: openseadragon
+:caption: Minimal markdown to embed OpenSeadragon in Elemental Microscopy. Replace the source with your DZI.
+:::{any:bundle} https://curvenote.github.io/widgets/widgets/openseadragon.mjs
+{
+    "tileSources": "https://your-github.github.io/your-repository/sample.dzi"
+}
+:::
+```
+
+:::{dropdown} Imaging
+Data collection was fully-motorized for unbiased sampling of large panoramas from replicate immunofluorescence experiments.
+Grids of overlapping tiles were acquired with a robotic XY stage (IX3-SSU) and Z-drift compensation (IX3-ZDC).
+Coverslips were imaged with an Olympus Fluoview FV3000 confocal microscope using a 100X objective lens (Olympus model UPLAPO100X, NA 1.50) in immersion oil. 
+Pinholes in each channel were maximized to capture the widest depth of field.
+:::