The use of a deep neural network scheme is proposed to help clinicians solve a difficult diagnosis problem in neurology. The proposed multilayer architecture includes a feature engineering step (from time-frequency transformation), a double compressing stage trained by unsupervised learning, and a classification stage trained by supervised learning. After fine-tuning, the deep network is able to discriminate well the class of patients from controls with around 90% sensitivity and specificity. This deep model gives better classification performance than some other standard discriminative learning algorithms. As in clinical problems there is a need for explaining decisions, an effort has been carried out to qualitatively justify the classification results. The main novelty of this paper is indeed to give an entropic interpretation of how the deep scheme works and reach the final decision.

Background and Objectives To assess the current diagnostic yield of genetic testing for the progressive myoclonus epilepsies (PMEs) of an Italian series described in 2014 where Unverricht-Lundborg and Lafora diseases accounted for ∼50% of the cohort. Methods Of 47/165 unrelated patients with PME of indeterminate genetic origin, 38 underwent new molecular evaluations. Various next-generation sequencing (NGS) techniques were applied including gene panel analysis (n = 7) and/or whole-exome sequencing (WES) (WES singleton n = 29, WES trio n = 7, and WES sibling n = 4). In 1 family, homozygosity mapping was followed by targeted NGS. Clinically, the patients were grouped in 4 phenotypic categories: “Unverricht-Lundborg disease-like PME,” “late-onset PME,” “PME plus developmental delay,” and “PME plus dementia.” Results Sixteen of 38 (42%) unrelated patients reached a positive diagnosis, increasing the overall proportion of solved families in the total series from 72% to 82%. Likely pathogenic variants were identified in NEU1 (2 families), CERS1 (1 family), and in 13 nonfamilial patients in KCNC1 (3), DHDDS (3), SACS, CACNA2D2, STUB1, AFG3L2, CLN6, NAXE, and CHD2. Across the different phenotypic categories, the diagnostic rate was similar, and the same gene could be found in different phenotypic categories. Discussion The application of NGS technology to unsolved patients with PME has revealed a collection of very rare genetic causes. Pathogenic variants were detected in both established PME genes and in genes not previously associated with PME, but with progressive ataxia or with developmental encephalopathies. With a diagnostic yield >80%, PME is one of the best genetically defined epilepsy syndromes.

Status epilepticus (SE) in pregnancy represents a life-threatening medical emergency for both mother and fetus. Pregnancy-related pharmacokinetic modifications and the risks for fetus associated wi...