Scientists have developed a technique and software that enhance electron microscopy by integrating real-time machine learning into the imaging process, known as "smart microscopy." They integrate AI with electron microscopy to accelerate the process of creating intricate maps of brain networks. The goal is to improve studies on connectomics and clinical pathology. 

Alzheimer's disease

The area of study called connectomics is growing quickly. Its goal is to map the complex network of connections in animal brains. Over ten years, it has grown from a small field of study to one that is about to (hopefully) solve the mysteries of memory and the physical causes of neuropathologies like Alzheimer's disease. 

Researchers gave powerful electron microscopes the analytical power of machine learning and put them at the centre of this field. Unlike traditional electron microscopy, the built-in AI is a "brain" that learns about the specimen. At the same time, the images are being taken and wisely focus on the important pixels at the nanoscale level, similar to how animals look around them. 

SmartEM

"SmartEM" helps connectomics quickly look at and rebuild the brain's complicated network of neurons and synapses with millimetre accuracy. Unlike standard electron microscopy, it has built-in AI, which helps us learn more about the brain's workings.

To reconstruct a human brain segment of about 100,000 neurons, using a regular microscope would take ten years of constant imaging and cost too much to be practical. But with SmartEM, the job could be done in three months by buying four of these cutting-edge microscopes for less than $1 million each, according to researchers.

Brains of animals

From these early stages, the field has come a long way. For example, in the 1980s, work was done to map the relatively easy connectome of C. elegans, which are small worms. Now, scientists are studying the brains of animals like zebrafish and mice, which are much more complicated.

This evolution shows not only huge progress but also rising complexity and demands: the team says that just mapping the mouse brain requires managing a vast 1,000 petabytes of data, which is a job that is far beyond any university's storage capabilities. 

Testing the waters

For their study, the researchers looked at 30-nanometer-thick slices of octopus tissue stuck on tapes, then put on wafers, and finally put into electron microscopes. Scientists took pictures of each slice of an octopus brain, which made up billions of pixels. It lets them combine the images to make a three-dimensional cube with nanometer-level detail. It gave an obvious picture of synapses. The main goal? To give these pictures colour, find each neuron, and figure out how they link, we must make a detailed map, or "connectome" map of the brain's wiring.

Conclusion

The group foresees a future in which connectomics will be readily available and reasonably priced. They anticipate that when biopsies from living patients are accessible, a broader range of research institutions can contribute to neuroscience without relying on large partnerships. This method will quickly become the norm. In addition, they are enthusiastic about utilizing the technology to comprehend pathologies, expanding its utility beyond connectomics.

Image source: Unsplash