Background: Bone marrow aspiration and biopsy remain the gold standard for the diagnosis of hematological diseases despite the development of flow cytometry (FCM) and molecular and gene analyses. However, the interpretation of the results is laborious and operator dependent. Furthermore, the obtained results exhibit inter- and intravariations among specialists. Therefore, it is important to develop a more objective and automated analysis system. Several deep learning models have been developed and applied in medical image analysis but not in the field of hematological histology, especially for bone marrow smear applications. Objective: The aim of this study was to develop a deep learning model (BMSNet) for assisting hematologists in the interpretation of bone marrow smears for faster diagnosis and disease monitoring. Methods: From January 1, 2016, to December 31, 2018, 122 bone marrow smears were photographed and divided into a development cohort (N=42), a validation cohort (N=70), and a competition cohort (N=10). The development cohort included 17,319 annotated cells from 291 high-resolution photos. In total, 20 photos were taken for each patient in the validation cohort and the competition cohort. This study included eight annotation categories: erythroid, blasts, myeloid, lymphoid, plasma cells, monocyte, megakaryocyte, and unable to identify. BMSNet is a convolutional neural network with the YOLO v3 architecture, which detects and classifies single cells in a single model. Six visiting staff members participated in a human-machine competition, and the results from the FCM were regarded as the ground truth. Results: In the development cohort, according to 6-fold cross-validation, the average precision of the bounding box prediction without consideration of the classification is 67.4%. After removing the bounding box prediction error, the precision and recall of BMSNet were similar to those of the hematologists in most categories. In detecting more than 5% of blasts in the validation cohort, the area under the curve (AUC) of BMSNet (0.948) was higher than the AUC of the hematologists (0.929) but lower than the AUC of the pathologists (0.985). In detecting more than 20% of blasts, the AUCs of the hematologists (0.981) and pathologists (0.980) were similar and were higher than the AUC of BMSNet (0.942). Further analysis showed that the performance difference could be attributed to the myelodysplastic syndrome cases. In the competition cohort, the mean value of the correlations between BMSNet and FCM was 0.960, and the mean values of the correlations between the visiting staff and FCM ranged between 0.952 and 0.990. Conclusions: Our deep learning model can assist hematologists in interpreting bone marrow smears by facilitating and accelerating the detection of hematopoietic cells. However, a detailed morphological interpretation still requires trained hematologists.

The National Comprehensive Cancer Network (NCCN) International Prognostic Index (IPI) is an enhanced prognostic tool that has identified some specific extranodal sites as a poor prognostic factor. We retrospectively analyzed 148 Taiwanese patients with newly diagnosed diffuse large B-cell lymphoma receiving rituximab, cyclophosphamide, doxorubicin, vincristine and prednisone (R-CHOP)-like regimens from January 2001 to December 2010 at the Tri-Service General Hospital. In univarate analysis, ≥ 2 extranodal involved sites had no significant prognostic relevance (p = 0.108), although extranodal involvement of the lung/pleura, liver, lower urinary tract or bone marrow was a statistically significant poor prognostic factor (p < 0.001). In multivariate analysis, specific extranodal sites had a stronger predictive value for poor prognosis (relative risk 3.654, 95% confidence interval 1.514-8.815, p = 0.004) compared with the number of extranodal sites involved. This finding suggests that specific extranodal involved sites have prognostic value in the R era.

The aim of this study was to compare the risk factors and clinical outcomes of bacteremia in allogeneic and autologous hematopoietic stem cell transplant (allo-HSCT and auto-HSCT) recipients with levofloxacin prophylaxis during the early period after transplantation. Characteristics of bacteremia within 45 days after transplantation between allo-HSCT and auto-HSCT recipients who received levofloxacin prophylaxis between January 2005 and December 2014 were retrospectively reviewed. Of 105 HSCT recipients included in this study, 55 (52.4%) received an allo-HSCT and 50 (47.6%) received an auto-HSCT. Twenty-five patients (23.8%) with HSCT developed 28 episodes of bacteremia. Of these 25 bacteremia patients, 15 received an allo-HSCT, while 10 received an auto-HSCT. The occurrence of Grade 3-4 graft-versus-host disease and longer engraftment duration were associated with bacteremia in allo- and auto-HSCT recipients (p = 0.001 and p = 0.002, respectively). Auto-HSCT recipients with bacteremia had a longer hospital stay after transplantation, while allo-HSCT recipients with bacteremia had an increased 45-day mortality rate as compared with those without bacteremia (p = 0.014 and p = 0.013, respectively). All 14 Gram-negative blood isolates in this study were resistant to fluoroquinolone. Levofloxacin prophylaxis in HSCT recipients is associated with the emergence of fluoroquinolone-resistant Gram-negative bacteria. The risk factors and clinical outcomes of bacteremia differ between allo- and auto-HSCT recipients, and these differences should be taken into account when designing strategies to prevent bacteremia.