Bone marrow biopsies play a central role in hematopathology for diagnosing a variety of diseases, staginglymphomas or performing follow-up progression. Tasks performed while examining biopsies include countingcells and estimating the ratio of various hematopoietic lineages. Inter- and intra-observer variability betweenhematopathologists in the outcome of these tasks has been shown to be significant, which could result in mul-tiple pathologists diagnosing some patients di erently. To that end, this paper presents a fully-convolutionalneural network (FCNN) architecture to segment six classes in bone marrow trephine biopsies, which could assisthematopathologists in identifying and delineating cells, thus reducing inter- and intra-observer variability. Ad-ditionally, to show an application of the neural network to a clinically relevant task, the output of the networkis used to train a classifier capable of distinguishing between normocellular and aplastic bone marrow. Resultsindicate the network is successfully capable of segmenting cells with an average detection rate of 83%. The clas-si er for distinguishing normocellular/aplastic bone marrow reaches an AUC of 0.990, showing that is capableof automatically identifying aplasia.
展开▼