Deep learning improves cancer diagnostic tools

0

Yi “Edwin” Sun, Ph.D. Electrical and Computer Engineering candidate at the University of Illinois Urbana-Champaign and a member of the Beckman Institute Biophotonic Imaging Lab led by Stephen Boppart, explored how deep learning methods can render optical coherence tomography polarization sensitive, or PS-OCT, more expensive. efficient and better equipped to diagnose cancer in biological tissues.

The article, titled “Synthetic polarization-sensitive Optical Coherence Tomography by Deep Learning”, was published in npj Digital Medicine.

OCT systems are common in the clinic and are used to generate high resolution cross sectional images of areas of the human body. Sun and his team have developed a revolutionary method of software application to the OCT tool to provide polarization-sensitive capabilities, without the cost and complexity that come with hardware-based PS-OCT imaging systems.

“We are trying to replace the hardware associated with the PS-OCT,” Sun said. “Nevertheless, [it] is still in the development and research stage. By adding a deep learning model on top of an OCT system, we suddenly come to PS-OCT capabilities without the traditional hardware added.

OCT is a non-invasive imaging test that uses light waves to determine the properties of a biological sample. However, by allowing the tool to use polarization sensitivity, scientists can detect relevant information that OCT cannot capture on its own. For example: OCT can differentiate tissues in a precise way and when the broader characteristics are clear; PS-OCT can detect abnormalities at a deeper level, differentiating microstructural features such as the orientations of collagen fibers that change in a cancer-infected area compared to a normal area.

“We have proven that applying our method to other systems can generate PS-OCT contrast, and that this model can be used on many OCT systems to help us differentiate between cancerous tissue and other types of tissue. tissues much better than OCT systems alone, ”Sun said. . “It’s a huge improvement, which makes this system better for cancer diagnoses.”

Deep learning, a subset of machine learning, has enabled the Sun team to create software that can be combined with OCT systems to provide polarization sensitivity.

“Deep learning has enabled a more advanced method of detecting subtle features in images, which can be used for more precise segmentation and classification. It also allows the imaging tool to use multiple layers to capture spatial features in an image, ”Sun said.

By applying historical data, deep learning methods aid in accurate diagnoses and even medical predictions. Sun’s team tested their model by predicting what a photo of a lush summer forest in December might look like: barren, gray, maybe a bit of ice and snow in the trees. With this concept in mind, images from OCT systems, coupled with this deep learning approach, can even predict the PS-OCT images that would come from more complex and expensive PS-OCT systems.

“Edwin’s study really highlights the power and potential of AI and deep learning approaches to predict and generate synthetic PS-OCT images from standard OCT images, a type of translation from image to image. With the increasing use of OCT in medical fields, this advancement is likely to have a broad impact and ultimately help improve disease detection and diagnosis, ”said Boppart, Ph.D. of Sun. thesis director who is both a doctor and a professor of engineering at the UIUC.

This research was funded in part by grants from the National Cancer Institute and the National Institutes of Health.

Reference: Sun Y, Wang J, Shi J, Boppart SA. Deep learning polarization sensitive synthetic optical coherence tomography. npj Average number. 2021; 4 (1): 1-7. doi: 10.1038 / s41746-021-00475-8

This article was republished from the following materials. Note: The material may have been modified for its length and content. For more information, please contact the cited source.


Source link

Leave A Reply

Your email address will not be published.