Machine Learning Enhances Light-Field Imaging-Based Method

Facebook X LinkedIn Email
Researchers from The Rockefeller University have developed an imaging technique that could allow scientists to penetrate the depth and breadth of the cortical areas of the brain at peak resolution, scale, and speed. The technology, called MesoLF, is a modular, mesoscale, light-field imaging hardware and software solution.

In experiments, MesoLF recorded thousands of neurons in a mouse cortex at up to 350 µm depth. The optical design and computational approach of the system could help scientists achieve a more complete understanding of the mechanics of cognition, while using workstation-grade computing resources.

Although mesoscopes provide resolution fine enough to resolve single cells and fields of view large enough to capture neurons across broad swathes of the brain, it is difficult to use them to capture neuronal activity within broad fields of view both simultaneously and volumetrically.

“The challenge with using mesoscopes for visualizing the fast activity of single neurons in 3D is that high-resolution, point-scanning approaches are typically needed, for which the scanning timescales very unfavorably with the size of the imaged volume,” said professor Alipasha Vaziri.

Techniques for deep-tissue brain imaging typically are based on the sequential acquisition of images. Light-field microscopy (LFM), a 3D imaging technique known for providing fast, high-resolution imaging — and the technique on which MesoLF is based — performs poorly when imaging deep, dense tissue that scatters light.

In previous work, Vaziri and his team circumvented some of the limitations of LFM with a machine-learning algorithm that estimates the locations of active neurons to better detect brain cell activity in dense tissue. In the current work, the team expanded the reach of the system by adding software and hardware that enable MesoLF to investigate tissues of different shapes and degrees of rigidity.

“This is made possible through a custom optical design for maintaining high optical imaging resolution over mesoscopic volumes, in combination with a set of algorithmic innovations that scale our modular computational pipeline’s capacity and capabilities accordingly,” Vaziri said.

The team said that the computational costs that come with the processing of large quantities of raw data have been kept as low as possible in the new, scaled-up version of MesoLF.

In tests, MesoLF captured key interactions between 10,500 neurons distributed volumetrically across multiple cortical areas in mice. The system imaged cells buried at previously inaccessible depths that were firing from brain regions many millimeters apart, and with high resolution. The researchers used the system to record thousands of neurons at 18 volumes per second, at an effective voxel rate of about 40 megavoxels per second.

In animals and humans alike, cognition and behavior depend on complex networks of neurons communicating with each other. Historically, imaging tools have lacked the capability to trace how neurons fire in sync from deep within the brain’s cortex.

Given the relatively low cost of optical hardware, Vaziri hopes to make the MesoLF technique widely available. The team’s designs are now available under an open-source license.

The research was published in Nature Methods (

Published: March 2023
machine learning
Machine learning (ML) is a subset of artificial intelligence (AI) that focuses on the development of algorithms and statistical models that enable computers to improve their performance on a specific task through experience or training. Instead of being explicitly programmed to perform a task, a machine learning system learns from data and examples. The primary goal of machine learning is to develop models that can generalize patterns from data and make predictions or decisions without being...
Research & TechnologyeducationAmericasThe Rockefeller Universitybrainbrain activitybrain imagingvolumetricBiophotonicsmesoscopymesoscopicdevicesinstrumentationlight-field imagingLFMMicroscopymachine learninghardwareSoftwareAIBioScan

We use cookies to improve user experience and analyze our website traffic as stated in our Privacy Policy. By using this website, you agree to the use of cookies unless you have disabled them.