Experiments · from SlideScope 1.22.0

Compare Treatments Across a Whole Microscopy Experiment

Organise a folder of images into treatments, samples and replicates. Run one counting or IHC scoring recipe on every image, then compare treatments with a dot plot, mean ± SD and the right statistical test, with n counted as samples, not images or cells. Outliers and disagreeing replicates go to a review queue, and every table, plot and Methods paragraph exports for your paper.

Download SlideScope See plans

Treatments, samples and replicates One recipe on every image n = samples, not images

Why comparing treatments by hand goes wrong

Different settings per image

Counting each image separately invites small changes in threshold or region from one image to the next, and the difference between treatments can come from the settings rather than the biology.

Inflated n

Treating every image, or every cell, as an independent observation makes n far larger than the number of animals or cultures, and p-values look more convincing than the experiment supports.

Outliers nobody looked at

One out-of-focus field or a fold in the tissue can move a whole sample. In a spreadsheet of numbers, nothing points you back to the image that caused it.

How an experiment works in SlideScope

Open the folder in SlideScope. It opens as a dataset of every image inside, in any format SlideScope reads: CZI, ND2, SVS, NDPI, OME-TIFF, LIF and more.
Organise it into an experiment. Give each image a treatment, a sample and a replicate, by importing a sample sheet or by accepting suggestions from the file names. Exclude an image with a reason: it stays on record and can be brought back. The experiment is a small file next to your images, so it moves with the folder.
Run one recipe on every image (Plus). The same Quantify settings count nuclei, cells or particles and score DAB on each image, so every number was produced the same way.
Compare treatments (Plus). Pick a measurement and SlideScope plots one dot per sample, with each treatment's mean and SD, and runs the matching tests.
Review and export. Open any flagged image from the review queue, correct an object's class or IHC grade, and export the image, sample and group tables, the plot and a Methods paragraph.

A dataset organised into vehicle and drug treatments, with a sample and replicate for each image and the Run recipe and Compare buttons — A folder organised as an experiment: 2 treatments, 6 samples, 13 images. Example data: EPFL BrdU IHC crops (CC BY 4.0), with treatment and sample names made up for the demonstration.

The statistics, and why n is the number of samples

Images and cells from the same animal, patient or culture are not independent. SlideScope averages each sample's images first, then runs the tests on one value per sample, so n is the number of biological replicates you actually have.

Treatments	Tests SlideScope runs	Shown with
Two	Welch's t-test (no equal-variance assumption) and Mann-Whitney U (rank-based)	A dot plot of every sample, mean ± SD, n per treatment
Three or more	One-way ANOVA and Kruskal-Wallis H (rank-based)	The same plot and table for every treatment

You can compare any value the recipe produced for every image: object count, density per mm², percent positive, H-score, Allred score, mean object area, or mean intensity per channel.

A dot plot of percent positive nuclei per sample for vehicle and drug, n = 3 each, with Welch's t-test and Mann-Whitney results and an outlier warning — Compare in SlideScope 1.22: one dot per sample, the mean and SD of each treatment, both tests, and a warning about an image far from the rest of its sample. Example data, as above.

Checks before you trust a difference

Outlier images: an image far from the other images of its sample (robust z above 3.5) is flagged and goes to the review queue.
Disagreeing replicates: a treatment whose samples vary widely (coefficient of variation above 0.5) is flagged.
Too few samples: a treatment with fewer than three samples is flagged, because no test can say much about it.
Image quality: results with too few objects, clipped pixels or a poor slide are flagged by the same checks Quantify uses on one image.
Corrections on record: objects you reclassify or regrade, and images you exclude with a reason, are written into the exported Methods text.

What it doesn't do

It does not run pairwise post-hoc tests after ANOVA, mixed models, or power calculations. Export the sample table to Prism, R or Python for those.
It compares the values one recipe produced; it does not choose the recipe or the measurement for you.
Ask Your Slide answers questions about the open image. It does not yet answer questions about a whole experiment.
It is for research. It is not a diagnostic system, and a statistically significant difference is not a clinical finding.

Questions about experiment analysis

How do I compare treatment groups in microscopy images?

Open the folder in SlideScope and organise its images into an experiment: treatments, samples and replicates, from a sample sheet or from suggestions based on the file names. With SlideScope Plus, run one Quantify recipe on every image, then choose Compare: SlideScope averages each sample's images, plots one dot per sample with the mean and SD of each treatment, and runs Welch's t-test and Mann-Whitney for two treatments or ANOVA and Kruskal-Wallis for more.

Why is n the number of samples and not the number of images or cells?

Images and cells from the same animal, patient or culture are not independent, so counting them as separate observations inflates n and makes differences look more certain than they are. SlideScope averages the images of each sample first and runs the tests on one value per sample.

Which measurements can I compare?

Any value the Quantify recipe produced for every image: object count, density per mm², percent positive, H-score, Allred score, mean object area and mean intensity per channel.

Is experiment analysis included in every plan?

Organising images into an experiment is included in every plan. Running one recipe across the experiment, comparing treatments and reviewing the run's results need SlideScope Plus. Quantify itself, on one image or a folder, is in every plan.

Does SlideScope replace a statistics package?

No. It runs the standard two-group and multi-group tests on one value per sample and reports them with the data behind them. It does not run pairwise post-hoc tests after ANOVA, mixed models or power calculations; export the sample table to Prism, R or Python for those.

Quick answers

Best fit: research groups comparing treatments, genotypes or time points across many images of stained sections, cultures or organoids, who want every image counted the same way and n reported correctly.

Key strengths: one recipe on every image, the sample as the statistical unit, outlier and replicate checks with a review queue, and exports with a Methods paragraph.

Important limitation: standard two-group and multi-group tests only; no post-hoc or mixed-model statistics.

Last reviewed: November 11, 2026.

Count a whole experiment the same way

Download SlideScope, try Quantify on the free sample library, and run your experiment with SlideScope Plus.

Download SlideScope