Deep Learning Applications in Biomedical Imaging
Abstract
Recently developed deep learning techniques revolutionized image analysis methods in the last decade. Classification, segmentation, quantitative prediction, and generating new data can be performed by the training of deep learning models. These tasks can directly be applied to biomedical imaging and successful applications will provide strong advantages to researchers and physicians in terms of efficiency for their studies and improvement in diagnosis.
In this talk, I will discuss my recent studies that using deep learning techniques to the following topics: glaucoma screening from fundus photographs based on regional retinal nerve fiber layer (RNFL) thickness estimation using deep learning, label-free digital histopathology with QPI imaging based on virtual staining and image classification techniques, and high-throughput phenotype screening platform using office scanner. For these studies, we have employed the convolution neural network (CNN) network architectures and trained them to perform image-to-number regression, image classification, image segmentation, and image-to-image generation. The details of each approach will be also discussed.
Biography
Dr. Hyunmo Yang earned his Ph.D. in physics from Ulsan National Institute of Science of Technology (UNIST) in Korea. After his degree in 2019, he joined as a postdoc researcher to the translational biophotonics laboratory in department of biomedical engineering at UNIST. He is currently working on developing machine learning and deep learning applications for biomedical imaging. His research interests are digital medicine, digital screening and digital histopathology using A.I. technology.
Click here to register for in-person attendance (lunch will be served)
