A team from the Gladstone Institutes, San Francisco, has used deep learning, a form of artificial intelligence, to teach a computer to identify and analyse brain cells.
The study, partially funded by the National Institute of Neurological Disorders and Stroke (NINDS), shows that deep learning can be used to speed up analysis and improve scientists’ understanding of brain cells.
“This approach has the potential to revolutionise biomedical research,” said PhD program director at the NINDS Margaret Sutherland.
“Researchers are now generating extraordinary amounts of data. For neuroscientists, this means that training machines to help analyse this information can help speed up our understanding of how the cells of the brain are put together and in applications related to drug development.”
The team, led by Dr Steven Finkbeiner, director and senior investigator at the Gladstone Institutes and professor of neurology and physiology at the University of California, trained the computer programme to analyse brain cells by showing it stained and unstained images. To test whether the programme had learned from this data, the researchers challenged it with new unlabelled images and found that with each round of training the programme was able to identify more complex features.
Additional experiments showed that the programme was able to successfully distinguish dead cells from living cells and identify specific types of brain cells.
“Deep Learning takes an algorithm, or a set of rules, and structures it in layers, identifying simple features from parts of the image, and then passes the information to other layers that recognise increasingly complex features, such as patterns and structures,” said Finkbeiner.
“This is reminiscent of how our brain processes visual information. Deep learning methods are able to uncover much more information than can be seen with the human eye.”
While the technology shows promise, it requires very large datasets of around 15,000 images to be effective. There is also the risk of overtraining the programme so that they become so specialised they can only identify structures in a particular set of images or in images generated in a particular way, and not make predictions about more general images.
Finkbeiner and his colleagues plan to apply these methods to disease-focused research, but further research is needed to refine the technology and make it more widely available.
“Now that we showed that this technology works, we can start using it in disease research. Deep learning may spot something in cells that could help predict clinical outcomes and can help us screen potential treatments,” said Finkbeiner.