Light-Sheet Microscope Automatically Adjusts to Optimize Image Quality

A light-sheet microscope has been developed that can automatically adapt to the dynamic optical conditions of large living specimens. The smart microscope combines a novel hardware design with a software system that can analyze a specimen continuously and automatically adjust the settings to maximize image quality. Called the AutoPilot framework, this development could enable long-term adaptive imaging of entire developing embryos and improve the resolution of light-sheet microscopes up to five-fold.

Researchers from the Janelia Research Campus at Howard Hughes Medical Institute, the Max Planck Institute of Molecular Cell Biology and Genetics, and Coleman Technologies Inc. collaborated on the development of the microscope, which seeks to resolve one of the primary challenges of light-sheet microscopy: that the light sheet illuminating the specimen and the focal plane of the detection system must be perfectly co-planar.

“If they deviate in any way geometrically, it’s like taking an image out of focus,” said Philipp Keller, Janelia group leader for the research.

The AutoPilot framework enables the microscope to analyze and optimize the spatial relationships between light-sheets and detection planes across the specimen volume in real-time and automatically adapt to the dynamic optical conditions of living specimens.

“The microscope is smart in the sense that it controls the experiment itself,” said Keller. “It’s not the human who instructs the microscope exactly how to take images. The microscope figures out on its own what it needs to do to get a sharp image.”

To enable the microscope, flexibility was added to the hardware, such as giving the microscope more freedom to rotate the light sheet in space. Software for the Autopilot framework was developed for analyzing images in real-time, determining how to improve their quality, and adjusting the microscope parameters accordingly. Calculations, analysis and adjustments are performed during the microscope’s brief downtime between collecting images.

20-h imaging of a fruit fly embryo with the nervous system fluorescently labeled. AutoPilot automatically detects the onset of the expression of the marker and optimizes all parameters associated with this color channel in response to the emerging signal. Courtesy of MPI of Molecular Cell Biology and Genetics.
“The system understands the relationship between all the different variables and when it observes that something is off, it can figure out which knob to turn to correct things,” said Loïc Royer, developer of the software.

Researchers tested the microscope on various model systems, imaging the development of zebrafish and fly embryos over a 20-plus hour period. They also performed adaptive whole-brain functional imaging in larval zebrafish. These experiments demonstrated improvements in spatial resolution and signal strength by a factor of two to five and exhibited recovery of cellular and sub-cellular structures in many regions that could not be resolved by non-adaptive imaging.

With the AutoPilot’s high level of instrument automation, even users without a technical background or training in light-sheet microscopy can produce optimal image data.

“We wanted to make the microscope as powerful as possible but also as easy to use as possible,” said Keller. “The framework does this by helping the user ensure that the experiment is set up correctly, and also making sure the user can effortlessly and reproducibly produce optimal quality images in every single experiment.”

Making microscopes adaptive and autonomous could contribute to the future use of light-sheet microscopy for automated high-throughput drug screens, mutant screens, and the construction of anatomical and developmental atlases in various biological model systems.

“So far, researchers had to sit at their microscope and tweak things manually—our system puts an end to this. It is like a self-driving car: it functions autonomously,” said Royer.

The research was published in Nature Biotechnology (doi:10.1038/nbt.3708).