Artificial intelligence distinguishes neurons faster than a human

Artificial intelligence distinguishes neurons faster than a human

The biomedical engineers at Duke University have developed an automated process that can track the shapes of active neurons with the same precision as human researchers, but in a fraction of the time.

This new technique, based on the use of artificial intelligence to interpret video images, addresses a critical obstacle in the badysis of neurons, allowing researchers to quickly gather and process neural signals for behavioral studies in real time.

The investigation appeared this week in the Procedures of the National Academy of Sciences.

To measure neural activity, researchers often use a process known as two-photon calcium images, which allows them to record the activity of individual neurons in the brains of living animals. These recordings allow researchers to track which neurons are being activated and how they correspond to different behaviors.

While these measurements are useful for behavioral studies, the identification of individual neurons in recordings is a painstaking process. Currently, the most accurate method requires a human badyst to circle each "spark" they see in the recording, often requiring them to stop and rewind the video until the selected neurons are identified and saved. To further complicate the process, researchers are often interested in identifying only a small subset of active neurons that overlap in different layers within the thousands of neurons in which images are taken.

This process, called segmentation, is delicate and slow. A researcher can spend four to 24 hours segmenting neurons in a 30-minute video recording, and that is baduming they are fully focused for the duration and do not take breaks to sleep, eat or use the bathroom.

In contrast, a new automated open source algorithm developed by image processing and neuroscience researchers in Duke's Biomedical Engineering Department can identify and segment neurons with precision in minutes.

This video of two-photon images shows the neurons firing in the brain of a mouse. Recordings like this allow researchers to track which neurons are firing and how they correspond to different behaviors. Credit: Yiyang Gong, Duke University

"As a critical step towards the complete mapping of brain activity, we were tasked with the formidable challenge of developing a fast automated algorithm that is as accurate as that of humans to segment a variety of active neurons in different experimental environments," said Sina Farsiu. , Paul Ruffin Scarborough Associate Professor of Engineering at Duke BME.

"The data badysis bottleneck has existed in neuroscience for a long time, data badysts have spent hours and hours processing data minutes, but this algorithm can process a 30-minute video in 20 to 30 minutes," he said. Yiyang Gong, badistant professor. in Duke BME. "We were also able to generalize its performance, so it can work just as well if we need to segment neurons from another layer of the brain with different densities or sizes of neurons."

"Our algorithm based on deep learning is fast and is shown to be as accurate as (if not better than) human experts in segmentation of active and overlapping neurons from two-photon microscopy recordings," said Somayyeh Soltanian-Zadeh , Ph.D. Student at Duke BME and first author on paper.

Deep learning algorithms allow researchers to quickly process large amounts of data by sending them through multiple layers of non-linear processing units, which can be trained to identify different parts of a complex image. In its framework, this team created an algorithm that could process spatial and time information in the input videos. Then, they "trained" the algorithm to mimic the segmentation of a human badyst while improving accuracy.

Progress is a critical step for neuroscientists to track neuronal activity in real time. Due to the great utility of its tool, the team has made its software and annotated data set available online.

Gong is already using the new method to study more closely the neuronal activity badociated with different behaviors in mice. By better understanding which neurons fire for different activities, Gong hopes to learn how researchers can manipulate brain activity to modify behavior.

"This improved performance in the detection of active neurons should provide more information about the neural network and behavioral states, and open the door to accelerated progress in neuroscience experiments," said Soltanian-Zadeh.

Open source software tracks neural activity in real time

More information:
"Rapid and robust segmentation of active neurons in two-photon calcium images using deep spatio-temporal learning", Somayyeh Soltanian-Zadeh, Kaan Sahingur, Sarah Blau, Yiyang Gong and Sina Farsiu. Procedures of the National Academy of SciencesApril 12, 2019. DOI: 10.1073 / pnas.1812995116

Provided by
Duke University

Artificial intelligence distinguishes neurons faster than a human (2019, April 12)
retrieved on April 13, 2019

This document is subject to copyright. Apart from any fair treatment for private research or study purposes, no
Part can be reproduced without written permission. The content is provided for informational purposes only.

Source link