Talk Announcement by Department of Mechanical Engineering, College of Engineering and Science, National United University            

Time/Date: 13:00-15:00 PM, 2025/04/18 (Friday)

Venue: Room A1-413, College of Engineering and Science Hall (I) (理工學院一館A1-413教室)

Title: SHG analysis and modeling of collagen fibril and fiber architecture via a rigorous wavelength dependent computational model of 3D quasi-phase matching and generative adversarial networks

Speaker: Prof. Paul J. Campagnola

Department of Biomedical Engineering, University of Wisconsin-Madison

Abstract: We have shown that diseases including cancers and fibroses have significant changes in the collagen fibril and fiber structure. We have used machine learning methods to classify normal and diseased tissues based on the respective fiber morphology. However, while we have developed these classifiers, they are still “black box’ and the important features remain unknown. To determine the most characteristic features of tissue classes, we use the generative adversarial network (GAN) framework StyleGAN which was developed for more input control over the image synthesis process. To quantify the respective important fiber morphologies, we trained the StyleGAN latent space and used PCA to determine the weights of fiber features. We found that curvature and density were the most important attributes in ovarian cancer. We have also previously demonstrated differences in the underlying fibril size and spacing in several diseased states. Here we used a combination of a heuristic phase matching model, SHG directional measurements and Monte Carlo simulations to determine the SHG emission direction. While providing insight into overall changes in the sub-resolution structure, the model is simplistic as only overall trends can be determined and little quantitative information is attainable. Here we develop a more complete computational model of 3D quasi-phasematching in collagen tissues. This in conjunction with wavelength- dependent position sensitive measurements of the spatial emission pattern, will afford the extraction of average fibril size and packing without performing TEM. Collectively these determinations will be broadly applicable to identifying characteristic diagnostic targets in a wide range of diseases.

Biography: Paul J. Campagnola obtained his PhD in Chemistry from Yale University in 1992 after which he was a postdoctoral associate at the University of Colorado from 1992-1995. He was on the faculty in the Department of Cell Biology, Center for Cell Analysis and Modeling at the University of Connecticut Health Center from 1995-2010, having adjunct appointments in the Physics Department and Biomedical Engineering Program. In 2010 became an Associate Professor in Departments of Biomedical Engineering and Medical Physics at the University of Wisconsin-Madison and was promoted to Professor in 2013. He is currently the Tong Biomedical Engineering Department Chair and UW Kellett Faculty Fellow. He is a Fellow of the Optical Society of America and American Institute for Medical and Bioengineering and currently a Fellow in the Big 10 Alliance Academic Leadership Program.

    His research is focused on studying structural and functional aspects of the extracellular matrix (ECM), where we have developed optical microscopy instrumental and analysis methods to study problems in basic science as well as those with translational potential. He has over 100 peer-reviewed journal articles, several review articles and book chapters, co-edited a book “Second Harmonic Generation microscopy” and given over 100 invited talks. He serves on the editorial board for the Journal of Biomedical Optics and serves on numerous NIH and NSF review panels.

機械工程學系特邀美國威斯康辛州大學麥迪遜校區生物醫學工程系Prof. Paul J. Campagnola蒞臨本校演講，歡迎本校各系所教師、學生前來參加。

時間：114年04月18日(五) 13：00 -15：00 PM

地點：(八甲校區)理工學院一館A1-413階梯教室

講題：SHG analysis and modeling of collagen fibril and fiber architecture via a rigorous wavelength dependent computational model of 3D quasi-phase matching and generative adversarial networks

主講者：Prof. Paul J. Campagnola 美國威斯康辛州大學麥迪遜校區生物醫學工程系(Department of Biomedical Engineering, University of Wisconsin-Madison)

摘要:

We have shown that diseases including cancers and fibroses have significant changes in the collagen fibril and fiber structure. We have used machine learning methods to classify normal and diseased tissues based on the respective fiber morphology. However, while we have developed these classifiers, they are still “black box’ and the important features remain unknown. To determine the most characteristic features of tissue classes, we use the generative adversarial network (GAN) framework StyleGAN which was developed for more input control over the image synthesis process. To quantify the respective important fiber morphologies, we trained the StyleGAN latent space and used PCA to determine the weights of fiber features. We found that curvature and density were the most important attributes in ovarian cancer. We have also previously demonstrated differences in the underlying fibril size and spacing in several diseased states. Here we used a combination of a heuristic phase matching model, SHG directional measurements and Monte Carlo simulations to determine the SHG emission direction. While providing insight into overall changes in the sub-resolution structure, the model is simplistic as only overall trends can be determined and little quantitative information is attainable. Here we develop a more complete computational model of 3D quasi-phasematching in collagen tissues. This in conjunction with wavelength- dependent position sensitive measurements of the spatial emission pattern, will afford the extraction of average fibril size and packing without performing TEM. Collectively these determinations will be broadly applicable to identifying characteristic diagnostic targets in a wide range of diseases.

主辦單位:  國立聯合大學機械工程學系

協辦單位：國立陽明交通大學