CW-NET for Multi-type Cell Detection and Classification in Bone Marrow Examination and Mitotic Figure Examination.

Abstract

Bone marrow examination is one of the most important indicators in diagnosing hematologic disorders and is typically performed under the microscope via oil-immersion objective lens with a total 100X objective magnification. On the other hand, mitotic detection and identification is critical not only for accurate cancer diagnosis and grading but also for predicting therapy success and survival. Fully automated bone marrow examination and mitotic figure examination from whole-slide images is highly demanded but challenging and poorly explored. Firstly, the complexity and poor reproducibility of microscopic image examination are due to the cell type diversity, delicate intra-lineage discrepancy within the multi-type cell maturation process, cells overlapping, lipid interference and stain variation. Secondly, manual annotation on whole-slide images is tedious, laborious and subject to intra-observer variability, which causes the supervised information restricted to limited, easily identifiable and scattered cells annotated by humans. Thirdly, when the training data is sparsely labeled, many unlabeled objects of interest are wrongly defined as background, which severely confuses AI learners. This paper presents an efficient and fully automatic CW-Net approach to address the three issues mentioned above and demonstrates its superior performance on both bone marrow examination and mitotic figure examination. The experimental results demonstrate the robustness and generalizability of the proposed CW-Net on a large bone marrow WSI dataset with 16,456 annotated cells of 19 bone marrow cell types and a large-scale WSI dataset for mitotic figure assessment with 262,481 annotated cells of five cell types. An online web-based system of the proposed method has been created for demonstration. (see https://youtu.be/MRMR25Mls1A). Supplementary data are available at Bioinformatics online.