AI in Optical Microscopy Enables Quick, High-Quality Imaging of Thick 3D Samples

Optical microscopy has existed for centuries, but scientists are still discovering new methods, techniques, and technologies to help them observe the microscopic world. Most recently, researchers have combined a computational imaging approach, known as compressive imaging, with a faster scanning method.

The new method was used to acquire two-photon microscopy images of a pollen grain in less than a second. With traditional approaches, it would have taken five times longer. Some possible applications will be to visualize a neural network or monitor activity from hundreds of neurons simultaneously. Neurons transmit signals in about 10 milliseconds, far faster than what conventional microscopy systems can follow.

Making Improvements to Speedier Scanning Two-Photon Microscopy

The new form of optical microscopy delivers ultrafast pulses of infrared laser light to the sample. The light interacts with tissue or fluorescent labels that emit signals used to create an image. It’s used because of its ability to produce high-resolution, 3D images up to a depth of 1 mm. Because the low-light conditions call for point detectors that require point-by-point image acquisition and reconstruction, this iteration of optical microscopy has limited imaging speed.

To speed up the imaging, researches developed a multi-focus laser illumination method. The method uses a digital micromirror device (DMD), a type of low-cost scanner that’s typically used in projectors. Researchers were able to further increase the imaging speed by combining multi-focus scanning with compressive sensing. This approach allows a specimen to be reconstructed using between 70 and 90 percent fewer exposures.

Related Blog Articles

Innovative Technologies Emerge to Meet Evolving Machine Vision and Imaging Challenges

Read

Imaging Systems are Leading the Way for Warehouse Profitability

Read

AI and Deep Learning Changing the Game for Optical Microscopy

Faster microscopy techniques mean that more images can be captured in less time. Analysis of this vast number of images becomes a bottleneck, but artificial intelligence, specifically the use of deep learning to process huge volumes of imagery with superhuman accuracy, can speed it. Advanced AI systems can then use those processed images to build high definition composite images — all in a fraction of the time required by human scientists.

Though there is still room for improvement, it’s exciting to be in a defining period where traditional scientific disciplines are coming together with very different disciplines, like AI and machine vision, to create something more than the sum of their parts.