Etch a Cell - Correct a Cell

How are the images shown in this project taken?

You will predominantly see images taken using an electron microscope. These images comprise a subset of the data that you helped annotate previously in the Etch a Cell - Fat Checker and Etch a Cell - Fat Checker Round 2 projects, where the former are images of liver tissue and the latter are of breast tissue.

What are the preloaded annotations seen on the subject?

The preloaded fat droplet annotations are generated using an automated machine learning model trained on the volunteer provided annotations from the Etch a Cell - Fat Checker project. The same model was also applied directly to Etch a Cell - Fat Checker Round 2 data.

What machine model has been used to produce initial annotation guesses?

We used a model called "PatchGAN", which is an image-to-image translation framework. Think of this as a model that can learn to "convert" one image (e.g., black and white) to a target image (e.g., colored image). In our context, we use this model to predict the image regions corresponding to fat droplets. See Etch a Cell - Fat Checker Results for more details.

How can I edit the preloaded annotations?

What is an electron microscope?

A traditional microscope uses light to illuminate a specimen, whereas an electron microscope uses a beam of electrons. Using an electron microscope rather than a light (photon) microscope allows us to look at smaller objects. This is because electrons have a much shorter wavelength than photons, and so are able to distinguish smaller objects.

What can you look at using an electron microscope?

Electron microscopes can be used to study a huge range of biological specimens, from tiny molecules such as proteins, to larger structures such as cells or tissues, and everything in between. Electron microscopy can be used to advance our understanding of many different processes, such as how healthy cells grow and develop, how bacteria and viruses infect our cells, how our immune systems fight infectious diseases and cancer, how our brains work and how we age.

How do we study the feature of interest in the image?

An image taken with an electron microscope contains a huge amount of information. The information we want to extract from an image will depend on our particular research question. Until recently though, most electron microscopes could only give a 2D view from a very thin slice through a cell or tissue. This was a problem because humans aren’t very good at visualising 3D structures from a 2D image. Think of the difference between seeing a city on a 2D map, and actually visiting that city and moving through it and seeing how all the roads and buildings and rivers and people interact with each other in 3D.

How can we study specimens in 3D?

We can look at a cell in 3D using different types of microscope, the two we have used here are called a serial block face scanning electron microscope (SBF SEM) and a focused ion beam scanning electron microscope (FIB SEM). In these microscopes, the cell surface is imaged, and then a thin ‘section’ is cut away, and then the cell is imaged again. A diamond knife is used to cut the sections away in the SBF SEM and a highly focused beam of heavy ions is used in the FIB SEM. These microscopes can cut sections so thin that it would take 2,000-20,000 sections to cut through the width of a single human hair! This process of cutting and imaging creates a series of 2D images.