> For the complete documentation index, see [llms.txt](https://www.parapathology.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://www.parapathology.com/clippings/multiplex-immunofluorescence-image-analysis-with-qupath-part-1.md). # Multiplex Immunofluorescence Image Analysis with QuPath — Part 1: Understanding Digital Images This repository contains the training materials for Part 1 of the Multiplex Immunofluorescence Image Analysis with QuPath workshop. This introductory section provides the foundational knowledge requir * **Instructor:** Sebastian G. Gilbert, PhD — Image and Spatial Omics Computational Scientist, Molecular and Advanced Pathology Core (MAPcore), University of British Columbia ([ORCID 0000-0002-7502-8409](https://orcid.org/0000-0002-7502-8409)) * **Format:** Workshop, first delivered **30 June 2026**; recorded and released as public modules on **24 August 2026** for those unable to attend and for a wider audience * **Scope:** 16 modules across two parts. Part 1 (8 modules, ≈2 h 14 min of video) is published; Part 2 covers mIF analysis proper * **Module 1 video:** [Part 1, Module 1: Workshop Introduction](https://www.youtube.com/watch?v=mPBsHYbAZi0) (11 min) * **Channel:** [Sebastian Gilbert — videos](https://www.youtube.com/@sebmathbio/videos) * **Slides (Part 1, all 8 decks):** [Zenodo record 22084859](https://zenodo.org/records/22084859) — [doi:10.5281/zenodo.22084859](https://doi.org/10.5281/zenodo.22084859), CC BY 4.0, deposited 24 August 2026 * **Workshop files and exercises:** [GPEC/Multiplex-immunofluorescence-image-analysis-with-QuPath](https://github.com/GPEC/Multiplex-immunofluorescence-image-analysis-with-QuPath) (MIT, created 4 September 2026) * **Software:** [QuPath](https://qupath.github.io/) — no prior QuPath experience assumed *** ## Summary An introductory, open-access workshop that takes a complete beginner from installing QuPath to quantitatively inspecting multiplex immunofluorescence (mIF) images, then on to tissue and cell segmentation, phenotyping, and cell-composition analysis. Part 1 — the portion covered by the Zenodo deposit and the eight published videos — deliberately spends its whole length on *digital images as such*: pixels, bit depth, histograms, channels and look-up tables, dimensions and spatial calibration, file formats, annotation, and image quality. The mIF-specific work is Part 2. Two design choices stand out: terminology is anchored to three external bioimage-analysis references rather than to QuPath's own vocabulary, and image/channel quality assessment is the first step of the mIF workflow, ahead of segmentation. ## Stated learning objectives (Part 1) From the Zenodo record, participants completing Part 1 will: * Install and navigate the QuPath software environment. * Understand how fluorescence images are generated by microscopes and detectors. * Learn the fundamentals of digital images: pixels, image dimensions, data types, image histograms. * Understand visualisation concepts — brightness, contrast, channels, colours, look-up tables (LUTs). * Learn about image metadata, spatial calibration, multidimensional image data, and projections. * Compare common bioimaging file formats, including compression, OME-TIFF, and OME-Zarr. * Create annotations and extract measurements within QuPath. * Apply best practices for quantitative image inspection and scientific image presentation. ## Module map ### Part 1 — Understanding Digital Images (published) | # | Module | Runtime | Video | | - | -------------------------------------------- | ------- | ---------------------------------------------------- | | 1 | Workshop Introduction / Workshop Information | 11 min | [watch](https://www.youtube.com/watch?v=mPBsHYbAZi0) | | 2 | Install & Navigate QuPath | 23 min | [watch](https://www.youtube.com/watch?v=rmYZS_2928E) | | 3 | Digital Image Basics | 21 min | [watch](https://www.youtube.com/watch?v=wn-V2izoifk) | | 4 | Channels and Colors | 11 min | [watch](https://www.youtube.com/watch?v=ieeYEOgiI4Y) | | 5 | Dimensions | 14 min | [watch](https://www.youtube.com/watch?v=G77duX1mI20) | | 6 | Image Data Formats | 15 min | [watch](https://www.youtube.com/watch?v=hjjhKLO-C8s) | | 7 | Annotating Images | 24 min | [watch](https://www.youtube.com/watch?v=2duQPkGQEPM) | | 8 | Quantitative Inspection | 17 min | [watch](https://www.youtube.com/watch?v=FFt-lEUK0Jw) | Each module has a matching PowerPoint deck (dated 2026-06-30) in the Zenodo record. ### Part 2 — mIF Image Analysis (announced) 9. Standard Image Analysis Workflow 10. Image/Channel Quality Assessment 11. Tissue Segmentation 12. Artifact Removal 13. Cell Segmentation 14. Cell Classification 15. Quantitative Analyses 16. Outlook and Future Directions Part 2 videos had not been posted to the instructor's channel as of 2026-09-15, and the Zenodo record covers Part 1 slides only. *** ## Module 1 — Workshop Introduction *Digest below is written from the video's auto-generated captions plus the published module description; it is not a transcript.* ### Where the workshop starts and stops The stated arc is: set up and navigate QuPath and its surrounding open-source ecosystem → acquire the vocabulary of digital image analysis (channels, colours, dimensions) and see how each term maps onto a workflow step → view, interact with, and annotate images → quantitatively inspect them for quality → apply all of it to mIF images through tissue and cell segmentation, classification and phenotyping, and presentation of results. What is explicitly *out of scope* is as instructive as what is in. The workshop ends at **cell composition** — a table of phenotyped cells per tissue compartment. Everything downstream of that (spatial statistics, cell–cell networks, niches, published cellular-neighbourhood methods) is named, positioned as question-dependent, and deferred to advanced modules that are described as in preparation. The workshop's deliverable is framed as the *objects* that make such analyses possible rather than the analyses themselves. ### Terminology is borrowed, not invented Module names and definitions are taken from three external resources, with an explicit warning that other sources define the same terms slightly differently: 1. The **Network of European BioImage Analysts (NEUBIAS)** and its bioimage-analysis terminology. 2. **Pete Bankhead's** book on analysing fluorescence microscopy images with ImageJ — Bankhead being the developer of QuPath. 3. The **Carpentries** lesson *Image Processing with Python*. The caution given is practical: within the workshop these resources' definitions are the ones in force, and the same terms carry slightly different properties elsewhere in the literature. ### What "mIF image analysis" is taken to mean Analysing multiple fluorescent signals within a single digital image to extract meaningful data. Four aspects are called out as the things that make it a distinct problem: * **Spectral unmixing** — separating the wavelengths of overlapping fluorophores so that each marker can be quantified as the biology it is supposed to report. * **Intensity extraction** — reading fluorescence intensity as a proxy for underlying biology. * **Automation** — using software to visualise, identify, quantify, and measure at a scale humans cannot. * **Spatial relationships** — how the different signals are arranged relative to one another. ### The generic image analysis workflow (four steps) 1. **Pre-processing** — enhance marker signal and suppress background and noise; the signal-to-noise ratio is the quality currency. 2. **Segmentation** — identify the image features that correspond to the biology of interest. 3. **Object formation and post-processing** — merge features into objects that can be manipulated and measured, with quality control applied to the object set. 4. **Measurement and comparison** — quantify the final objects and compare across tissue types, conditions, or experimental arms. ### The mIF workflow used in this workshop (six steps) Demonstrated on a single tissue-microarray core: 1. **Quality checks** — is the image good enough to analyse at all, and did every marker work? This step refines the image down to the channels that can be pursued. 2. **Tissue segmentation** — define the analysable area and, where relevant, separate tissue compartments. Automatable. 3. **Artifact removal** — mostly a manual step, with partial automation; further refines the tissue area. 4. **Cell segmentation** — identify individual cells within the retained tissue or compartments. 5. **Cell classification / phenotyping**. 6. **Cell composition analysis** — quantify and compare composition between compartments or areas. Compared with the generic four-step workflow above, this one front-loads quality assessment: the channel set is settled before any segmentation is attempted. ### Instructor background and acknowledgements A PhD begun in 2020 at the University of Birmingham on complex spatial structures in biomedical images, co-funded by AstraZeneca with academic collaborators; a postdoc developing those methods into commercial tools; then core-facility image analysis in the UK and Canada, with mIF work for academic and industrial partners and publications on digital image analysis tools. The workshop is funded and supported by MAPcore at UBC and was first piloted at the Birmingham tissue-analytics core facility \[unverified — the facility name is taken from auto-generated captions]. The video closes with a land acknowledgement for the UBC Vancouver Point Grey campus, on the ancestral and unceded territory of the Musqueam, Squamish, and Tsleil-Waututh Nations. *** ## Why this is worth keeping * It is the rare mIF course that treats **image literacy as the prerequisite**, not the appendix: eight modules and over two hours before a single cell is segmented. Bit depth, LUTs, spatial calibration, and OME-TIFF/OME-Zarr are handled as analysis-determining facts rather than file-handling trivia. * **Quality assessment gates the pipeline.** Channels that failed are dropped before segmentation, which is the honest alternative to carrying a bad marker through to a phenotype table. * **Licensing is genuinely open** — CC BY 4.0 slides on Zenodo with a DOI, MIT-licensed exercise files, free software — so the material can be reused for local teaching with attribution. * It complements the LJI *From Samples to Knowledge* QuPath course: that one is an advanced two-day workshop assuming comfort with QuPath, this one is the on-ramp to it. ## Related notes in this vault * [From Samples to Knowledge 2025: QuPath Training Course](/clippings/from-samples-to-knowledge-2025-qupath-training-course.md) * [Solving Unpopular Problems: The QuPath Story](/clippings/solving-unpopular-problems-the-qupath-story-the-pathologist.md) * [Image Analysis](/computational-digital-and-mathematical-pathology/image-analysis.md) * [Digital Pathology Software](/computational-digital-and-mathematical-pathology/digital-pathology-software.md) * [Courses and MOOCs](/appendix/courses-and-moocs.md) ## Citation Gilbert, S. (2026). *Multiplex immunofluorescence image analysis with QuPath — Part 1: Understanding Digital Images* \[Lesson]. Zenodo. (CC BY 4.0)