No abstract

Objectives Primary cold agglutinin disease is a type of autoimmune hemolytic anemia caused by circulating antibodies against I antigen, a carbohydrate expressed on most cells, including red blood cells. The underlying disease has been characterized in recent years as a distinct B-cell lymphoproliferative disease of the bone marrow, occurring mostly in the elderly. The disease has been now been included as a separate entity in the most recent classifications of mature B-cell neoplasms. Methods A review of the characteristics of cold agglutinin disease is provided, with an emphasis on the pathology features. Results A detailed description of the histopathology, immunophenotype, and genetics of cold agglutinin disease is provided and compared to other B-cell lymphoproliferative diseases in the bone marrow with similar features. Conclusions Recognition of the pathology features of cold agglutinin disease allows to distinguish it from other diseases, especially lymphoplasmacytic lymphoma and marginal zone lymphoma.

Background: The diagnosis of plasma cell neoplasms requires accurate, and ideally precise, percentages. This plasma cell percentage is often determined by visual estimation of CD138-stained bone marrow biopsies and clot sections. While not necessarily inaccurate, estimates are by definition imprecise. For this study, we hypothesized that deep learning can be used to improve precision. Methods: We trained a semantic segmentation-based convolutional neural network (CNN) using annotations of CD138+ and CD138- cells provided by one pathologist on small image patches of bone marrow and validated the CNN on an independent test set of images patches using annotations from two pathologists and a non-deep-learning commercial software. Once satisfied with performance, we scaled-up the CNN to evaluate whole slide images (WSIs), and deployed the system as a workflow friendly web application to measure plasma cell percentages using snapshots taken from microscope cameras. Results: On validation image patches, we found that the intraclass correlation coefficients for plasma cell percentages between the CNN and pathologist #1, a non-deep learning commercial software and pathologist #1, and pathologists #1 and #2 were 0.975, 0.892, and 0.994, respectively. The overall results show that CNN labels were almost as accurate pathologist labels at a cell-by-cell level. On WSIs from 10 clinical cases, the CNN continued to perform well, and identified two cases where the sign-out pathologist overestimated plasma cell percentages. Conclusions: The high labeling accuracy of the CNN supports its eventually application as a computational second-opinion tool for the measurement of plasma cell percentages in clinical practice.