活動時間：2024-08-12 16:00
活動地點：2号學院樓454
主講人：梁友
主講人中文簡介：
梁友博士，現為加拿大都市大學（Toronto Metropolitan University ）副教授。
活動内容摘要：
Fluorescence microscopic imaging of tissue sections is a foundational tool in diagnostic pathology and biomedical research. The development of image processing methods and algorithms of hyperspectral fluorescence microscopy imaging (HFMI) has facilitated the detection of various fluorescent contrast sources within hyperspectral data cubes. However, effectively visualizing and analyzing these high-dimensional HFMI images remains a challenge. The demand for efficient image processing techniques for diverse types of FHSI data across biomedical applications is evident, especially concerning semantic segmentation, a critical process for creating labelled HFMI data. We propose a novel graph-based algorithm for the semantic segmentation of HFMI. First, superpixels are generated to remove the image noise and create small homogenous regions with similar spectral features. Second, the normalized graph cuts algorithm is used to perform an initial image segmentation. Moreover, linear unmixing is used to add extra abundance information to combine with the pixel spectral features and the normalized cuts algorithm is applied again to segment the image. This proposed segmentation algorithm holds significant promise for enhancing the comprehension and diagnosis of eye diseases.
主持人：童金英