Three-dimensional GPU-accelerated active contours for automated localization of cells in large images

Abstract

Cell segmentation in microscopy is a challenging problem, since cells are often asymmetric and densely packed. Successful cell segmentation algorithms rely identifying seed points, and are highly sensitive to variablility in cell size. In this paper, we present an efficient and highly parallel formulation for symmetric three-dimensional contour evolution that extends previous work on fast two-dimensional snakes. We provide a formulation for optimization on 3D images, as well as a strategy for accelerating computation on consumer graphics hardware. The proposed software takes advantage of Monte-Carlo sampling schemes in order to speed up convergence and reduce thread divergence. Experimental results show that this method provides superior performance for large 2D and 3D cell localization tasks when compared to existing methods on large 3D brain images.

Whole-brain tissue mapping toolkit using large-scale highly multiplexed immunofluorescence imaging and deep neural networks

“…For practicality and efficiency, we focused on thin (10 μm) slices that are imaged two-dimensionally to avoid antibody/dye penetration limitations and complexities associated with 3D imaging and automated image analysis 24 . The proposed computational pipeline is specifically designed to generate 2D immunohistology results from individual whole-brain slices which is expandable for 3D segmentation 25 and reconstruction 26 to better understand the connectome of brain cytoarchitecture at cellular, niche, and organ levels 27 . Cellular measurements are exported to flow cytometry standard (FCS) and Image Cytometry Experiment (ICE) file formats for visualization and statistical profiling using common commercially available software tools (e.g., FCS Express, De Novo Software; FlowJo, BD Biosciences, Kaluza, Beckman Coulter, etc.).…”

Semisupervised Deep Learning for the Detection of Foreign Materials on Poultry Meat with Near-Infrared Hyperspectral Imaging

“…Another advantage of DL models is that they can handle large high-dimensional data such as hyperspectral images. Several studies have reported that the use of DL in HSI improved the classification accuracy, computational efficiency, and robustness to noise and solved other challenges [ 13 , 19 , 20 ]…”

“…Another advantage of DL models is that they can handle large high-dimensional data such as hyperspectral images. Several studies have reported that the use of DL in HSI improved the classification accuracy, computational efficiency, and robustness to noise and solved other challenges [13,19,20] Recently, a generative adversarial network (GAN) was used for anomaly detection [21,22] in an unsupervised way. Jiang et al [21] proposed a GAN-based anomaly detection model for hyperspectral images in remote sensing, where anomaly suppressed hyperspectral images were reconstructed and a detection map was predicted by using differences between the original and reconstructed images.…”