Manifold Extraction in Fluorescent Stack via Deep Learning
Published in 15th International Conference on Signal-Image Technology & Internet-Based Systems, 2019
Jianfeng Cao, Hong Yan. 15th International Conference on Signal-Image Technology & Internet-Based Systems. SITIS 2019
Abstract
Fueled by the development of advanced imaging techniques, the biological research has recently experienced ever-growing improvements, especially on microscopy analysis. The utilization of microscopies, however, is hampered by either the quality or quantity of these images. At the same time, the equipment is inevitably constrained by the physical limitations. Here we present MF-Net, a framework to automate the extraction of 2.5D membrane manifold from 3D blurred stack image. MF-Net realizes the transformation from 3D to 2D index map, and further to 2.5D manifold efficiently. Accompanied with a scheme to synthesize data, MF-Net is trained without manual annotations. Out of the box, MF-Net gets promising results on both synthetic and real microscopy images. Source code is available at https://github.com/cao13jf/MF-Net