Event Date
Event Date
We introduce a super-resolved virtual staining (VS) approach that employs a diffusion model to significantly enhance spatial resolution and VS fidelity. We validated our approach through rigorous blind testing on low-resolution autofluorescence images acquired from label-free human lung tissue samples. Our model produced virtually stained images closely resembling corresponding high-resolution bright-field images of H&E-stained tissue sections, demonstrating improvements of ~4–5-fold in spatial resolution relative to the original label-free microscopy inputs. By offering superior resolution and staining accuracy, our diffusion-model-based VS framework has the potential to improve current digital pathology workflows substantially.
Presenter
UCLA Samueli School of Engineering (United States)
Yijie Zhang received his Bachelor of Science degree in Optical Science and Engineering from Zhejiang University in Hangzhou, China, in 2018. He earned his Master of Science degree in Electrical and Computer Engineering from UCLA in 2020. That same year, he joined Professor Aydogan Ozcan’s Bio- and Nano-Photonics Group in the Department of Electrical and Computer Engineering at UCLA to pursue his Ph.D. His research primarily focuses on computational imaging for virtual tissue staining and biosensing.