Brain Imaging, Image Reconstruction Combine in a Single Workflow

Facebook X LinkedIn Email
Scientists at Institute of Science and Technology Austria (ISTA) have developed an imaging technology to analyze live brain tissue at a scope and spatial resolution that, according to the researchers, was not previously possible.

The technology, called Live Information Optimized Nanoscopy Enabling Saturated Segmentation (LIONESS), integrates optical methods and deep learning to allow dense, 4D, nanoscale reconstruction of living brain tissue. LIONESS images and reconstructs the sample in a way that clarifies many dynamic structures and functions in the live tissue. The method unites live imaging with nanoscale reconstruction, extending tissue analysis with information on morphological dynamics, molecular identities, and neuronal activity.

By helping to decode tissue architecture in the living brain and potentially in other organs, LIONESS could provide insight into the extent and significance of plasticity in the central nervous system.

Existing brain imaging technologies are limited in their ability to reconstruct living brain tissue, in 3D, all the way down to the single-synapse level. Light microscopy is hindered by insufficient 3D resolution, inadequate signal-to-noise ratio, and a prohibitive light burden. Although electron microscopy can capture images at nanometer-scale resolution, it is inherently static, requiring the sample to be fixed in one biological state.

To remove these restraints, the researchers turned to stimulated emission depletion (STED) microscopy; they applied dye molecules to the spaces around sample cells and used STED to reveal the superresolved shadows of the cellular structures. Superresolution shadow imaging, where the use of extracellularly applied fluorophores reveals cellular structures and arrangements, enabled the researchers to visualize these structures in the tissue.

However, increased resolution comes with a high load of imaging light on the sample, which can damage living tissue. This factor, along with optical imperfections that can cause progressive signal loss at higher resolution, limited 3D resolution and signal-to-noise ratio and interfered with the researchers’ goal of imaging brain tissue at a resolution that matched the actual architecture of brain tissue.

By coupling STED with a two-stage deep learning strategy, the researchers enabled LIONESS to provide isotropic, superresolution imaging while keeping the sample alive. The first level of deep learning enhances the image quality, and the second identifies the different cellular structures in the dense neuronal environment.

The Danzl group and collaborators photographed under ISTA’s new Michael Gröller Bridge. Courtesy of Peter Rigaud/ISTA.
The Danzl group and collaborators photographed under ISTA’s new Michael Gröller Bridge. Courtesy of Peter Rigaud/ISTA.
During imaging, LIONESS collects only as much information from the sample as needed. This is followed by the first deep learning step, which serves to fill in additional information on the brain tissue’s structure using image restoration.

LIONESS leverages information on sample structure from separate, previous measurements to reduce the light burden and imaging time without sacrificing resolution. This allows LIONESS to achieve a resolution of about 130 nm, while remaining gentle enough for imaging living brain tissue in real time.

During the second deep learning step, LIONESS interprets the complex imaging data and identifies the neuronal structures. Volumetric, live-imaging data is translated into nanoscale-resolved instance segmentations. From live tissue STED, to restoration network training, to segmentation network training, to analysis, LIONESS provides a pipeline to reconstruct live brain tissue.

“With LIONESS, for the first time, it is possible to get comprehensive, dense reconstruction of living brain tissue,” researcher Philipp Velicky said. By imaging the tissue multiple times, he said, the newly developed method allows users to observe and measure the dynamic cellular biology in the brain take its course. The output is delivered as a reconstructed image of the cellular arrangements in three dimensions, with time making up the fourth dimension, since the sample can be imaged over minutes, hours, or days, according to Velicky.

By revealing subcellular structures and capturing how these may change over time, the tool developed at ISTA could provide valuable insight into the functional architecture of brain tissue — and potentially other organs as well.

The research was published in Nature Methods (

Published: July 2023
sted microscopy
STED microscopy, or stimulated emission depletion microscopy, is a superresolution imaging technique in fluorescence microscopy that surpasses the diffraction limit, enabling the visualization of structures at the nanoscale level. This technique was developed to overcome the limitations imposed by the diffraction of light, which traditionally hindered the resolution of optical microscopy to a few hundred nanometers. Key features and principles of STED microscopy: Superresolution: STED...
deep learning
Deep learning is a subset of machine learning that involves the use of artificial neural networks to model and solve complex problems. The term "deep" in deep learning refers to the use of deep neural networks, which are neural networks with multiple layers (deep architectures). These networks, often called deep neural networks or deep neural architectures, have the ability to automatically learn hierarchical representations of data. Key concepts and components of deep learning include: ...
In optical character recognition, the method of dividing a string of characters into separate, distinct characters.
Superresolution refers to the enhancement or improvement of the spatial resolution beyond the conventional limits imposed by the diffraction of light. In the context of imaging, it is a set of techniques and algorithms that aim to achieve higher resolution images than what is traditionally possible using standard imaging systems. In conventional optical microscopy, the resolution is limited by the diffraction of light, a phenomenon described by Ernst Abbe's diffraction limit. This limit sets a...
Imagingbrainbrain imaginglive brain imagingBiophotonicsendoscopynanoscopyMicroscopySTED microscopyAIdeep learningEuropeResearch & Technologyeducationsegmentationimage segmentationimage reconstructionimage reconstruction algorithmsInstitute of Science and Technology Austriasuperresolutionsuperresolution STED microscopysuperresolution imagingBioScan

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.