Neuroimaging studies have shown that schizophrenia is associated with disruption of resting-state local functional connectivity. However, these findings vary considerably, which hampers our understanding of the underlying pathophysiological mechanisms of schizophrenia. Here, we performed an updated and extended meta-analysis to identify the most consistent changes of local functional connectivity measured by regional homogeneity (ReHo) in schizophrenia. Specifically, a systematic search of ReHo studies in patients with schizophrenia in PubMed, Embase, and Web of Science identified 18 studies (20 datasets), including 652 patients and 596 healthy controls. In addition, we included three whole-brain statistical maps of ReHo differences calculated based on independent datasets (163 patients and 194 controls). A voxel-wise meta-analysis was then conducted to investigate ReHo alterations and their relationship with clinical characteristics using the newly developed seed-based d mapping with permutation of subject images (SDM-PSI) meta-analytic approach. Compared with healthy controls, patients with schizophrenia showed significantly higher ReHo in the bilateral medial superior frontal gyrus, while lower ReHo in the bilateral postcentral gyrus, right precentral gyrus, and right middle occipital gyrus. The following sensitivity analyses including jackknife analysis, subgroup analysis, heterogeneity test, and publication bias test demonstrated that our results were robust and highly reliable. Meta-regression analysis revealed that illness duration was negatively correlated with ReHo abnormalities in the right precentral/postcentral gyrus. This comprehensive meta-analysis not only identified consistent and reliably aberrant local functional connectivity in schizophrenia but also helped to further deepen our understanding of its pathophysiology.

No abstract

In this paper we introduce our Chinese-English simultaneous translation system participating in AutoSimTrans 2021. In simultaneous translation, translation quality and latency are both important. In order to reduce the translation latency, we cut the streaming-input source sentence into segments and translate the segments before the full sentence is received. In order to obtain high-quality translations, we pretrain a translation model with adequate corpus and fine-tune the model with domain adaptation and sentence length adaptation. The experimental results on the development dataset show that our system performs better than the baseline system.