WEBKNOSSOS allows researchers and students to explore large 3D microscopy datasets directly in the browser. We spoke with Ahmed Elewa, assistant professor at Miami University, about how he used WEBKNOSSOS to teach cell biology through a hands-on course project.

Can you tell us about the course project you ran with your students?

I was teaching an introduction to cell biology for the first time last semester. While there’s a great standard textbook, I wanted students to also see real cell images, not just diagrams. At the same time, I was working on a project using FIB-SEM datasets and needed manual annotations for machine learning. I realized that combining the two made sense: each student could adopt a cell from a published embryo dataset and annotate its features throughout the semester.
The dataset I used was a C. elegans embryo with around 550 cells. Because this organism has an invariant developmental plan, each cell can be specifically identified. That meant I could assign every student a unique cell.
We had about 48 students, divided into groups of six or seven, each representing different cell types - neuron, gut, muscle, hypodermis, seam and pharyngeal cells. Throughout the semester, assignments matched the course topics. For example, when we discussed cell energetics, students annotated mitochondria; when we covered protein synthesis, they annotated nucleoli, which are responsible for ribosome biogenesis.

How did WEBKNOSSOS fit into this workflow?

Everything happened on WEBKNOSSOS. The students made accounts during the first week, and I posted tutorial videos on how to use the platform and segment different cell compartments. Each week, the tutorials guided them through a new assignment tied to the lecture topic.
Students could identify their cell, segment the nucleus, trace the cell membrane, annotate organelles like mitochondria, lipid droplets, endosomes, or even recognize asymmetries in the cell. For example, they noticed how mitochondria often cluster on one side depending on neighboring cells. This hands-on work helped them test their understanding and develop hypotheses about cellular organization.
By the end of the semester, all annotations were merged in WEBKNOSSOS, producing a comprehensive annotated dataset. Each student also submitted a final report summarizing their observations, counting organelles, and reflecting on unique features.

Combined annotations in WEBKNOSSOS across multiple cells of the embryo. 3D meshes show segmented nuclei, mitochondria, lipid droplets, and endomembrane compartments.

What were some key findings from the students’ work?

One exciting discovery was related to gut granules. Students identified unusual endomembrane compartments in gut cells, which were later manually curated and quantified. I found 314 of these granules, only in gut cells, revealing structural features that hadn’t been fully described in 3D before. These observations form the basis of our publication providing ultrastructural evidence for the organization of this organelle.

WEBKNOSSOS view of annotated gut granules in a C. elegans embryo. The 3D mesh reveals the characteristic tubular ring structure surrounding a central compartment, as identified in this study.

Some students also explored cell asymmetry, noting how the distribution of organelles differed depending on neighboring cells. These observations, coupled with the systematic student annotations, created a valuable manual dataset for future machine learning.

What challenges did you encounter with managing annotations in WEBKNOSSOS?

Merging student annotations was tricky. At first, I tried using the merge feature, but it confused the organization of the segment tab. Eventually, I reviewed and re-annotated the combined dataset myself, keeping everything tidy with consistent segment IDs.

Note: Thanks to Ahmed’s feedback, the WEBKNOSSOS team has since improved the merge annotation functionality, keeping the organization of the segment list.

Even so, students had a very positive experience. Most used their personal laptops without issues, and posting tutorials weekly helped guide them through more complicated segmentation tasks.

Which WEBKNOSSOS features were most useful for this project?

One of the most helpful tools was the quick select feature. It allowed students to rapidly segment, though for very fine structures like endoplasmic reticulum, manual annotation was still necessary.
Another key aspect was direct access to the published dataset via WEBKNOSSOS links. The dataset was already online, which made it easy to onboard students and start annotating immediately. Unfortunately, groups sometimes make data available in principle but not in a usable format - WEBKNOSSOS made this seamless.

How did you plan to use the student annotations for AI?

I’d like to use the curated annotations as training data for AI models. We want to automatically detect structures like the nucleus, mitochondria, lipid droplets, and endomembrane compartments. For instance, by segmenting nuclei, we could quantify heterochromatin vs euchromatin, potentially linking these data to single-cell RNA-seq for gene expression analysis. I’m doing this in collaboration with a colleague in the biomedical engineering department at Miami University. So her students are also working with WEBKNOSSOS. I’m also keen on trying out the new AI features WEBKNOSSOS has released.

How did students react to using WEBKNOSSOS?

They were very engaged. Students had a smooth technical experience, and most quickly became confident with the platform. Weekly tutorials of 20–30 minutes helped them navigate more advanced tools, like plasma membrane annotation or 3D navigation.
One of the most rewarding aspects was seeing students use their own descriptions for uncharacterized compartments, such as comparing vesicles to dice patterns. They weren’t just annotating - they were thinking critically about what they observed.

What’s next for this project?

I plan to run the course again in the fall, using a different stage of the embryo. I hope together we can discover and publish another cool feature of C. elegans cell biology.

A Student Perspective

One of the students in Ahmed Elewa’s class, Megan Cole, described the project as “unlike any other class” she had taken before.
Rather than only learning from slides, students immediately applied lecture concepts to real microscopy data. Every student adopted a single C. elegans cell and annotated it throughout the semester as new topics were introduced in class.
“We learned the content in lecture and then actually saw how it applied in real life,” she explained. “It made cell biology feel much more tangible.”
Using WEBKNOSSOS, students annotated nuclei, mitochondria, lipid droplets, and other cellular structures directly in the browser. Weekly tutorial videos guided them through the segmentation workflow and helped them organize annotations into different groups and folders.
“Once I got over the initial learning curve, it became very intuitive,” she said. “The tutorials made it really easy to understand how to create segments, organize annotations, and navigate through the dataset.”
She especially appreciated the ability to compute 3D meshes from annotations.“The 3D view was really helpful because it’s hard to fully understand structures from only one 2D plane. Being able to compute the mesh helped us analyze the shape of mitochondria and the overall structure of the cell.”
Students mostly worked independently on their assigned cells during the semester before comparing findings within groups of similar cell types near the end of the course. The collaborative aspect, she said, became one of the most valuable parts of the experience.
“It taught me a lot about biology, but also communication and collaboration. It felt much more like doing real science than just taking a class.”

Ahmed’s course demonstrates that WEBKNOSSOS is not just a research tool - it’s also a powerful platform for education and collaborative science, bridging the gap between real microscopy data and machine learning applications.