Categories: world

Improvement of molecular imaging using deep learning

Credit: Rensselaer Polytechnic InstituteGenerating extensive molecular images of organs and tumors in living organisms can be performed at ultra-fast rate using a new deep learning approach to image reconstruction developed by researchers at the Rensselaer Polytechnic Institute. The research group's new technology has the potential to significantly improve the quality and speed of imaging in living matter and was the focus of an article recently published in Light: Science and Applications in a Nature Journal. 1 9659005] Compressed sensing based image processing is a signal processing technique that can be used to create images based on a limited set of point measurements. Recently, a Rensselaer research group proposed a new instrumental approach to utilize this methodology to acquire extensive molecular data sets reported in Nature Photonics . While this approach provided more complete images, data processing and image formation could take hours. The latest methodology developed at Rensselaer is based on previous development and has the potential to produce images in real time while improving the quality and usability of the images formed. This can facilitate the development of personal drugs, improve clinical diagnostics or identify tissue to be cut. In addition to providing an overall snapshot of the subject being investigated, including those organs or tumors that scientists have visually directed by florescence, this image process may reveal information about successful intracellular delivery of drugs by measuring the rate of disintegration of the fluorescence. To enable almost real-time visualization of molecular events, the research group has utilized the latest…



Credit: Rensselaer Polytechnic Institute

Generating extensive molecular images of organs and tumors in living organisms can be performed at ultra-fast rate using a new deep learning approach to image reconstruction developed by researchers at the Rensselaer Polytechnic Institute.

The research group’s new technology has the potential to significantly improve the quality and speed of imaging in living matter and was the focus of an article recently published in Light: Science and Applications in a Nature Journal. 1

9659005] Compressed sensing based image processing is a signal processing technique that can be used to create images based on a limited set of point measurements. Recently, a Rensselaer research group proposed a new instrumental approach to utilize this methodology to acquire extensive molecular data sets reported in Nature Photonics . While this approach provided more complete images, data processing and image formation could take hours.

The latest methodology developed at Rensselaer is based on previous development and has the potential to produce images in real time while improving the quality and usability of the images formed. This can facilitate the development of personal drugs, improve clinical diagnostics or identify tissue to be cut. In addition to providing an overall snapshot of the subject being investigated, including those organs or tumors that scientists have visually directed by florescence, this image process may reveal information about successful intracellular delivery of drugs by measuring the rate of disintegration of the fluorescence.

To enable almost real-time visualization of molecular events, the research group has utilized the latest developments in artificial intelligence. The very improved image reconstruction is achieved by means of deep learning. In-depth learning is a complex set of algorithms designed to teach a computer to recognize and classify data. Specifically, this team developed a folding network architecture that the Rensselaer researchers call Net-FLICS, which stands for fluorescence lifetime images with compressed sense.

“This technology is very promising for a more accurate diagnosis and treatment,” said Pingkun Yan, co-director of the Biomedical Imaging Center in Rensselaer. “This technique can help a doctor to better visualize where a tumor is and its exact size. They can just cut off the tumor instead of cutting a larger portion and saving the healthy, normal tissue.”

Yan developed this method with corresponding author Xavier Intes, the second co-director of the Biomedical Imaging Center in Rensselaer, who is part of the Rensselaer Center for Biotechnology and Interdisciplinary Studies. PhD student Marien Ochoa and Ruoyang Yao supported the research.

“At the end, the goal is to translate these into a clinical environment. Usually when you have clinical systems you want to be as fast as possible, Ochoa says, which she reflected on how quickly this new technology allows researchers to capture these

Further development is needed before this pioneering new technology can be used in clinical setting, but its progress has been accelerated by introducing simulated data based on modeling, a specialty for Intes and his lab.

“For deep learning needs You usually have a very large amount of data for training, but for this system, it doesn’t have the luxury yet, because it’s a very new system, Yan says.

He said the team’s research shows that modeling can be used innovatively in image processing, which exactly extends the model to the actual experimental data.


Explore further:
New bioimaging technology is fast and economical

More information:
Ruoyang Yao et al. Net-FLICS: fast quantitative fluorescence longevity art with compressed sense – deep learning, Light: Science & Applications (2019). DOI: 10,1038 / s41377-019-0138-x [QiPianetalCompressivehyperspectraltime-resolvedwide-anglefluorescencelifetimeimages Nature Photonics (2017). DOI: 10,1038 / nphoton.2017.82

Share
Published by
Faela