Photoacoustic microscopy (PAM) allows researchers to see the smallest blood vessels in the body, but it can generate unwanted signals and noise. A team of researchers at the McKelby Institute of Technology at Washington University in St. Louis have found a way to significantly reduce noise and maintain image quality while reducing the laser energy required to generate images by 80%.
Song Hu, an associate professor of biomedical engineering, and members of his laboratory Machine learningA base image processing technique called sparse coding to remove noise from PAM images of vascular structure, oxygen saturation, and oxygen saturation. Blood flow In the mouse brain.The results of the work have been published online IEEE transactions for medical imaging..
To obtain such images, researchers need high-density sampling of the data. This requires a high laser pulse repetition rate, which can raise safety concerns.However, reducing the laser pulse energy will cause a failure. image quality Inaccurate measurement of blood oxygenation and blood flow. So Zhuoying Wang, a doctoral student in Hu’s lab and the first author of the dissertation, is a type of machine commonly used in image processing that does not require ground truth to train to improve images. Introduced sparse coding, which is learning. Quality and quantitative accuracy while using low laser dose.
The team applied this technique to images of blood hemoglobin concentration, oxygenation, and flow in the brain of mice at both normal and reduced energy levels. Their two-step approach worked very well, significantly reducing noise and achieving similar image quality that was previously only possible with five times higher laser energy.
“In the first step of our approach, the noise is less sparse than the signal, so sparse coding separated the vascular signal from the noise in a cross-sectional scan taken at various tissue locations called the B scan.” Wang said. “Then we applied the same Sparse coding The strategy of the projected image formed by the denoised B scan in the second step to further suppress the background noise. “
Hu said that machine learning was previously used to denoise photoacoustic images, but the two-step method is one step ahead.
“Our approach allows us to remove noise and leave the signal untouched,” Hu said. “It not only enhances the visibility of microvessels, but also provides the opportunity to maintain signal presentation and perform quantitative imaging.”
This is the first demonstration of what these machine learning tools can do, but Hu said it demonstrates the importance of advanced computational tools in imaging in general, especially in photoacoustic microscopy.
“We hope to reduce laser energy by one-fifth, but we believe that advances in follow-up can do more than just reduce it. Laser energy But also to improve temporal resolution, or how fast an image can be taken without losing resolution and spatial coverage, “he said.
Zhuoying Wang et al, Low fluence multiparametric photoacoustic microscope with sparse coding, IEEE transactions for medical imaging (2021). DOI: 10.1109 / TMI.2021.3124124
Washington University in St. Louis
Quote: Https: //medicalxpress.com/news/2021-11-energy-quality-photoacoustic-microscopy-images.html low energy by machine learning (November 22, 2021) acquired on November 22, 2021 , High quality photoacoustic microscopic image
This document is subject to copyright. No part may be reproduced without written permission, except for fair transactions for personal investigation or research purposes. Content is provided for informational purposes only.
Less energy, better quality photoacoustic microscopy images with machine learning Source link Less energy, better quality photoacoustic microscopy images with machine learning
The post Less energy, better quality photoacoustic microscopy images with machine learning appeared first on California News Times.