Japanese Braille Translation Using Deep Learning - Conversion from Phonetic Characters (Kana) to Homonymic Characters (Kanji).

A blind student writes and submits reports in Braille word processor, which is difficult for teachers to read. This study's purpose is to make a translator from Braille into mixed Kana-Kanji sentences for such teachers. Because Kanji has homonyms, it is not always possible to get correct results when converting. To overcome this difficulty, we used deep learning for translation. We built a training dataset composed from 15,000 pairs of Braille codes and mixed Kana-Kanji sentences, and a validation dataset. In training, we got an accuracy of 0.906 and a good Bleu score of 0.600. In validation, we found 5 mistaken words in selecting homonymous Kanji by examining translation mistakes from 100 pairs of the verification sentences. The choice of homonymous Kanji depends on the context. For decreasing such type of errors, it is necessary to introduce of translation of paragraphs by increasing the scale of the network model in deep learning, and to expand the network structure.

Pathological features of FTLD-FUS in a Japanese population: analyses of nine cases.

We investigated the pathological features of frontotemporal lobar degeneration (FTLD) with fused in sarcoma protein (FUS) accumulation (FTLD-FUS) in the Japanese population. Only one out of nine FTLD-FUS cases showed pathology that corresponds to atypical FTLD with ubiquitin-positive inclusions (aFTLD-U). Five were basophilic inclusion body disease (BIBD) and two were neuronal intermediate filament inclusion disease. The last case was unclassifiable and was associated with dystrophic neurites (DNs) as the predominant FUS pathology. The results of this study indicate an ethnic difference from western countries. In Japan, BIBD is the most common subtype of FTLD-FUS and aFTLD-U is rare, a finding which contrasts with aFTLD-U being the most common form in western countries. Immunohistochemical analyses of these FTLD-FUS cases reveal that FUS abnormally accumulated in neuronal cytoplasmic inclusions (NCIs) and DNs has an immunohistochemical profile distinct from that of normal, nuclear FUS. NCIs and DNs are more readily stained than the nuclei by antibodies to the middle portion of FUS. Antibodies to the carboxyl terminal portion, on the other hand, stain the nuclei more readily than NCIs and DNs. Such an immunohistochemical profile of NCIs and DNs was similar to that of cytoplasmic granular FUS staining which we previously reported to be associated with dendrites and synapses. Redistribution of FUS from the nucleus to the cytoplasm could be associated with the formation of abnormal FUS aggregates in FTLD-FUS.

Clinicopathological characteristics of FTLD-TDP showing corticospinal tract degeneration but lacking lower motor neuron loss.

The presence of frontotemporal lobar degeneration with TDP-43-positive inclusions (FTLD-TDP) showing corticospinal tract (CST) degeneration but lacking lower motor neuron (LMN) loss has been reported, and the term primary lateral sclerosis (PLS) is used to distinguish motor neuron disease (MND) of these cases from amyotrophic lateral sclerosis (ALS). To date, however, details of clinicopathological findings of FTLD-MND-PLS type (FTLD-MND-P) have not been reported. We evaluated medical records and histopathological findings of ten cases of FTLD-MND-P, in comparison with those of six FTLD-MND-ALS type (FTLD-MND-A) cases. The mean age at onset and disease duration of FTLD-MND-P cases were 54 and 12 years, respectively. The first symptoms were frontotemporal dementia showing behavioral abnormality and/or personality change in five cases, semantic dementia in three cases, progressive non-fluent aphasia in one case, and auditory hallucination in one case. Upper motor neuron signs were clinically identified in six of the ten cases. There were no LMN signs throughout the clinical course in any case. Histopathologically, there was no obvious LMN loss or Bunina bodies in the hypoglossal nucleus or spinal cord in any case, whereas the CST was involved in all cases. The cerebral cortex of the six cases showed type 1 of TDP-43 histology defined by Cairns et al., whereas three cases showed type 3 histology, and one case showed type 2 histology. In all cases, TDP-43 positive neuronal cytoplasmic inclusions were absent or rare in the LMNs, while TDP-43 positive round structures were frequently identified in the neuropil of the spinal cord anterior horn in some cases. This study clarified that FTLD-MND-P cases have characteristic clinicopathological features distinct from those of FTLD-MND-A.

Relationship of phosphorylated alpha-synuclein and tau accumulation to Abeta deposition in the cerebral cortex of dementia with Lewy bodies.

Alpha-synuclein accumulated in the brain of dementia with Lewy bodies (DLB) is phosphorylated at serine129 (palpha-synuclein). We investigated the accumulation of palpha-synuclein in the brains of patients with DLB and Alzheimer's disease (AD). We employed 18 DLB patients with neocortical Lewy body type pathology (nLBTP) with or without AD. We also employed the same number of AD cases without significant nLBTP. We refer to the former group as the nLBTP group and to the latter as the AD type pathology (ADTP) group. In the nLBTP group, palpha-synuclein positive neurite pathology such as threads and dots occurs in all layers of the temporal neocortex. It was comparable in degree with tau pathology in AD. Fifteen cases in the nLBTP group were associated with Abeta deposition that meets the CERAD plaque score "C" and one case with a score "B". In these plaque-associated cases, the severity of palpha-synuclein pathology was related to the degree of Abeta deposition. In the cases with relatively moderate Abeta deposition, tau pathology was disproportionately mild in the nLBTP group, while the total of tau and palpha-synuclein pathology was proportionate to Abeta deposition in both the nLBTP and ADTP groups. Our results support the ideas that there is an overlap in the pathology between AD and DLB and that Abeta promotes accumulation of both alpha-synuclein and tau. The procession from Abeta to neurite pathology in the cerebral cortex of AD and DLB may be unifiable.

Relative paucity of tau accumulation in the small areas with abundant Abeta42-positive capillary amyloid angiopathy within a given cortical region in the brain of patients with Alzheimer pathology.

Cerebral amyloid angiopathy (CAA) is a manifestation of amyloid beta-protein (Abeta) accumulation in the elderly as well as in patients with Alzheimer's disease (AD). Two types of CAA have been noted, based on the type of vasculature in which Abeta is deposited: cerebral capillary amyloid angiopathy (capCAA) and non-capCAA. Non-capCAA is a common form of CAA that consists of Abeta deposited in arteries and arterioles. Recent information on Abeta metabolism in the brain suggests that non-capCAA is associated with Abeta secretion into the cerebrospinal fluid via the perivascular space, whereas capCAA is associated with Abeta removal to blood plasma via the capillary endothelium. Abeta40, a major and relatively soluble Abeta isoform, is deposited predominantly in non-capCAA, and Abeta42, which is insoluble and associated more closely than Abeta40 with AD, is deposited predominantly in capCAA. Studying small areas of microscopic size within a given cortical region, we found an inverse association of capCAA and senile plaques. We also found a relative paucity of tau pathology in the small areas with abundant capCAA compared with the small areas with abundant senile plaques within a cortical region with the same cytoarchitecture. We suppose that both capCAA and senile plaques indicate high Abeta42 in the neuropil but that only Abeta42 in the form of insoluble deposits in senile plaques promotes tau abnormality.