Search

We propose a novel method that combines a convolutional neural network (CNN) with a long short-term memory (LSTM) mechanism for accurate prediction of intracranial hemorrhage on computed tomography (CT) scans. The CNN plays the role of a slice-wise feature extractor while the LSTM is responsible for linking the features across slices. The whole architecture is trained end-to-end with input being an RGB-like image formed by stacking 3 different viewing windows of a single slice. We validate the method on the recent RSNA Intracranial Hemorrhage Detection challenge and on the CQ500 dataset. For the RSNA challenge, our best single model achieves a weighted log loss of 0.0522 on the leaderboard, which is comparable to the top 3% performances, almost all of which make use of ensemble learning. Importantly, our method generalizes very well: the model trained on the RSNA dataset significantly outperforms the 2D model, which does not take into account the relationship between slices, on CQ500. Our codes and models is publicly avaiable at https://github.com/VinBDI-MedicalImagingTeam/midl2020-cnnlstm-ich.

Objectives: Widespread adoption of recent recommendations for minimal residual disease (MRD) detection in myeloma has partly been impeded by a paucity of studies detailing multiparameter flow cytometry (MPF) assay validation. In response, we have validated a novel and efficient single-tube 10-color assay for MRD detection that incorporates the recently recommended plasma cell markers. Methods: Aspirate samples from 53 patients with plasma cell disorder were analyzed using a novel single-tube 10-color method. The limit of detection, precision of measurement, and linearity of measurement of our new assay were determined using serial dilution experiments. The stability of the new antibody cocktail and the viability/specificity of stained samples were evaluated using serial time course measurements. Results: There was a high degree of quantitative agreement between our new 10-color method and an established eight-color method. Four positive samples detected by the 10-color assay were below or at the limit of detection of the eight-color assay, confirming its high sensitivity. In two cases, subsequent revision of the International Myeloma Working Group Uniform Response Criteria was necessary. Conclusion: Adoption of our validated 10-color assay would enable clinical laboratories to satisfy current MRD recommendations without significantly increasing the demands on current workflow practices.

Measurable residual disease is associated with inferior outcomes in patients with acute myeloid leukemia (AML). Measurable residual disease monitoring enhances risk stratification and may guide therapeutic intervention. The European LeukemiaNet working party recently came to a consensus recommendation incorporating leukemia associated immunophenotype-based different from normal approach by multi-color flow cytometry for measurable residual disease evaluation. However, the analytical approach is highly expertise-dependent and difficult to standardize. Here we demonstrate that loss of plasmacytoid dendritic cell differentiation after 7+3 induction in AML is highly specific for measurable residual disease positivity (specificity 97.4%) in a uniformly treated patient cohort. Moreover, loss of plasmacytoid dendritic cell differentiation as determined by a blast-to-plasmacytoid dendritic cell ratio >10 was strongly associated with inferior overall and relapse-free survival (RFS) [Hazard ratio 2.79, 95% confidence interval (95%CI): 0.98-7.97; P=0.077) and 3.83 (95%CI: 1.51-9.74; P=0.007), respectively), which is similar in magnitude to measurable residual disease positivity. Importantly, measurable residual disease positive patients who reconstituted plasmacytoid dendritic cell differentiation (blast/ plasmacytoid dendritic cell ratio