A CNN-based approach for joint segmentation and quantification of nuclei and NORs in AgNOR-stained images

Maikel M. Rönnau, Tatiana W. Lepper, Luara N. Amaral, Pantelis V. Rados, Manuel M. Oliveira

Abstract: Highlights•A CNN architecture for automatic joint segmentation and counting of nuclei and AgNORs.•An AgNOR-stained image dataset from oral mucosa with 1,171 images from 48 patients.•Training and evaluation of 102 CNN models for AgNOR-stained image segmentation.•A semi-automatic image annotation strategy to reduce the workload from specialists to produce ground truth image annotations.•An algorithm to identify overlapping nuclei and exclude them from AgNOR counting.

External IDs:doi:10.1016/j.cmpb.2023.107788