Figure 1.
A workflow diagram presents nuclei segmentation processing stages. The input image enters preprocessing, which includes hematoxylin extraction, contrast enhancement, and orthogonal color projection. The output is split into patches. Patches enter local processing with adaptive thresholding, instance refinement, and anomalous instance removal, producing a pseudolabel. The pseudolabel feeds N u S e g H o p, which outputs a heatmap. The heatmap enters global processing with heatmap filtering, local maxima detection, watershed plus thresholding, and self supervised classification. Intermediate outputs include pseudolabel, heatmap, candidate R O Is, and false positive instances. The final output is a predicted mask.An overview of LG-NuSegHop pipeline. Pre-processing applies image enhancement to prepare the image for the local processing modules. NuSegHop receives as input the pseudolabel and predicts a heatmap. In the last step, global processing modules increase the nuclei detection rate using information across the entire input image