3D convolutional neural networks have achieved promising results for video tasks in computer vision, including video saliency prediction that is explored in this paper. However, 3D convolution encodes visual representation merely on fixed local spacetime according to its kernel size, while human attention is always attracted by relational visual features at different time of a video. To overcome this limitation, we propose a novel Spatio-Temporal Self-Attention 3D Network (STSANet) for video saliency prediction, in which multiple Spatio-Temporal Self-Attention (STSA) modules are employed at different levels of 3D convolutional backbone to directly capture long-range relations between spatio-temporal features of different time steps. Besides, we propose an Attentional Multi-Scale Fusion (AMSF) module to integrate multi-level features with the perception of context in semantic and spatio-temporal subspaces. Extensive experiments demonstrate the contributions of key components of our method, and the results on DHF1K, Hollywood-2, UCF, and DIEM benchmark datasets clearly prove the superiority of the proposed model compared with all state-of-the-art models.

Background: This study aim to derive and validate a simple and well-performing risk calculator (RC) for predicting psychosis in individual patients at clinical high risk (CHR).Methods: From the ongoing ShangHai-At-Risk-for-Psychosis (SHARP) program, 417 CHR cases were identified based on the Structured Interview for Prodromal Symptoms (SIPS), of whom 349 had at least 1-year follow-up assessment. Of these 349 cases, 83 converted to psychosis. Logistic regression was used to build a multivariate model to predict conversion. The area under the receiver operating characteristic (ROC) curve (AUC) was used to test the effectiveness of the SIPS-RC. Second, an independent sample of 100 CHR subjects was recruited based on an identical baseline and follow-up procedures to validate the performance of the SIPS-RC.Results: Four predictors (each based on a subset of SIPS-based items) were used to construct the SIPS-RC: (1) functional decline; (2) positive symptoms (unusual thoughts, suspiciousness); (3) negative symptoms (social anhedonia, expression of emotion, ideational richness); and (4) general symptoms (dysphoric mood). The SIPS-RC showed moderate discrimination of subsequent transition to psychosis with an AUC of 0.744 (p < 0.001). A risk estimate of 25% or higher had around 75% accuracy for predicting psychosis. The personalized risk generated by the SIPS-RC provided a solid estimate of conversion outcomes in the independent validation sample, with an AUC of 0.804 [95% confidence interval (CI) 0.662–0.951].Conclusion: The SIPS-RC, which is simple and easy to use, can perform in the same manner as the NAPLS-2 RC in the Chinese clinical population. Such a tool may be used by clinicians to counsel appropriately their patients about clinical monitor v. potential treatment options.

Background.: Chinese psychiatrists have gradually started to focus on those who are deemed to be at ‘clinical high-risk (CHR)’ for psychosis; however, it is still unknown how often those individuals identified as CHR from a different country background than previously studied would transition to psychosis. The objectives of this study are to examine baseline characteristics and the timing of symptom onset, help-seeking, or transition to psychosis over a 2-year period in China.Method.: The presence of CHR was determined with the Structured Interview for Prodromal Syndromes (SIPS) at the participants' first visit to the mental health services. A total of 86 (of 117) CHR participants completed the clinical follow-up of at least 2 years (73.5%). Conversion was determined using the criteria of presence of psychotic symptoms (in SIPS). Analyses examined baseline demographic and clinical predictors of psychosis and trajectory of symptoms over time. Survival analysis (Kaplan–Meier) methods along with Log-rank tests were performed to illustrate the relationship of baseline data to either conversion or non-conversion over time. Cox regression was performed to identify baseline predictors of conversion by the 2-year follow-up.Results.: In total 25 (29.1%) of 86 completers transitioned to a psychotic disorder over the course of follow-up. Among the CHR sample, the mean time between attenuated symptom onset and professional help-seeking was about 4 months on average, and converters developed fully psychotic symptoms about 12 months after symptom onset. Compared with those CHR participants whose risk syndromes remitted over the course of the study, converters had significantly longer delays (p = 0.029) for their first visit to a professional in search of help. At baseline assessment, the conversion subgroup was younger, had poorer functioning, higher total SIPS positive symptom scores, longer duration of untreated prodromal symptoms, and were more often given psychosis-related diagnoses and subsequently prescribed antipsychotics in the clinic.Conclusions.: Chinese CHR identified primarily by a novel clinical screening approach had a 2-year transition rate comparable with those of specialised help-seeking samples world-wide. Early clinical intervention with this functionally deteriorating clinical population who are suffering from attenuated psychotic symptoms, is a next step in applying the CHR construct in China.