Intrathecal administration of methotrexate (IT-MTX) can lead to neurotoxicity. MTX-induced neurotoxicity occasionally manifests with a stroke-like presentation that is difficult to distinguish from genuine stroke. We retrospectively reviewed records of nine patients with leukemia or lymphoma and episodes of stroke-like presentation at our institute between 2010 and 2015 for whom magnetic resonance imaging (MRI) data were available. Coagulation test results were compared between the two diagnostic groups. Four patients were diagnosed with MTX-induced stroke-like neurotoxicity. The first neurological event occurred 10-13 days after the fourth or later IT-MTX treatment. All four patients had hemiparalysis, two exhibited disturbed consciousness and three presented with speech disorders. Fibrin/fibrinogen degradation products (FDP) and D-dimer values were within normal ranges. MRI revealed bilateral lesions with restricted diffusion in all four cases. Neurological symptoms fluctuated and resolved within 5 days, and IT-MTX was subsequently re-initiated in all four cases. One patient developed transient hemiparalysis after a subsequent IT-MTX treatment, but this did not recur thereafter. Bilateral lesions on MRI and normal coagulation are indicative of MTX-induced stroke-like neurotoxicity. Continuation of IT-MTX after these events is generally feasible, but adverse event risk should be carefully weighed against anti-tumor benefits.

No abstract

Deep learning algorithms have achieved great success in cancer image classification. However, it is imperative to understand the differences between the deep learning and human approaches. Using an explainable model, we aimed to compare the deep learning-focused regions of magnetic resonance (MR) images with cancerous locations identified by radiologists and pathologists. First, 307 prostate MR images were classified using a well-established deep neural network without locational information of cancers. Subsequently, we assessed whether the deep learning-focused regions overlapped the radiologist-identified targets. Furthermore, pathologists provided histopathological diagnoses on 896 pathological images, and we compared the deep learning-focused regions with the genuine cancer locations through 3D reconstruction of pathological images. The area under the curve (AUC) for MR images classification was sufficiently high (AUC = 0.90, 95% confidence interval 0.87–0.94). Deep learning-focused regions overlapped radiologist-identified targets by 70.5% and pathologist-identified cancer locations by 72.1%. Lymphocyte aggregation and dilated prostatic ducts were observed in non-cancerous regions focused by deep learning. Deep learning algorithms can achieve highly accurate image classification without necessarily identifying radiological targets or cancer locations. Deep learning may find clues that can help a clinical diagnosis even if the cancer is not visible.

A male patient diagnosed with severe congenital protein C (PC) deficiency during the neonatal period was treated with long‐term warfarin but frequently developed purpura fulminans and bleeding. At four years of age, edoxaban was initiated (direct oral anticoagulant [DOAC]). His d‐dimer and fibrin/fibrinogen degradation product levels were closely monitored. His PC activity increased from below the sensitivity range to 17%; this increase was thought to be due to a reduction in PC consumption during edoxaban therapy. After edoxaban introduction, he experienced just one episode of purpura fulminans over two years without any adverse events. Thus, DOAC may be a promising alternative for the management of congenital PC deficiency.

No abstract