MEMO
收藏资源简介:
MEMO是首个公开的多模态EMA和OCTA视网膜图像数据集,由马里兰大学生物工程系创建。该数据集包含30对EMA和OCTA图像,每对图像手动标注了6个对应点,用于研究多模态视网膜图像注册。MEMO特别之处在于其包含的两种模态间血管密度差异较大,为解决这一挑战,研究团队提出了基于分割的深度学习框架VDD-Reg和一个新的评估指标MSD。数据集不仅用于视网膜图像注册研究,还可能应用于自动红细胞追踪等多个研究领域。
MEMO is the first publicly available multimodal retinal image dataset encompassing EMA and OCTA modalities, developed by the Department of Bioengineering at the University of Maryland. This dataset includes 30 pairs of EMA and OCTA images, with each pair manually annotated with 6 corresponding landmarks for research on multimodal retinal image registration. A notable feature of MEMO is the significant difference in vascular density between its two imaging modalities. To address this challenge, the research team proposed a segmentation-based deep learning framework named VDD-Reg and a novel evaluation metric called MSD. In addition to retinal image registration research, this dataset also has potential applications in multiple research fields such as automated red blood cell tracking.




