No abstract

Preterm infants face lifelong disabilities, including learning disorders, as well as visual, auditory and behavioral problems. Recent studies have demonstrated that leptin, an adipocytokine encoded by a gene associated with obesity and expressed in adipose tissue, affects neurocognitive and motor function; however, the mechanisms of brain damage in preterm infants are unclear. In the present study, the neuroprotective effects of leptin in a rat model of preterm hypoxic‑ischemic brain damage were investigated. Rats (2‑days‑old) were subjected to brain damage (ligation of the common carotid artery followed by exposure to 6% oxygen for 2 h) and treated with vehicle (control) or leptin. Spatial memory was analyzed in the present study using the Morris water maze test 19 days following ligation. Over the 24‑day post‑surgical observation period, capture‑resistance test, forelimb suspension and open field tests were conducted to evaluate motor function and anxiety‑associated behavior. Treatment with leptin did not affect survival rate or body weight. Treatment with leptin increased the number of platform crossings in rats with premature brain damage in the Morris water maze test, which was used to assess spatial memory. Multivariate analysis revealed that leptin reduced the latency to finding the platform location, independent of gender and weight. In the capture‑resistance, forelimb suspension and open field tests, there were no differences between animals administered leptin and the sham group. Collectively, the results of the present study suggested that leptin may alleviate spatial memory impairment resulting from premature brain damage, independent of gender or weight. These results may improve understanding of the neuroprotective effects exhibited by leptin in infants with preterm brain damage.

More and more babies are born preterm, and most of them could be saved, but many of the survivors face lifelong disabilities, including learning disorder and visual and hearing problems. Unclear mechanisms of premature brain damage lead to the underlying safety and reliable treatment issues. Recently, scientists found that biological effects of leptin on non-hypothalamus area were such as learning and memory, cognitive function and neuroprotective effect. Therefore, we aimed to determine whether the neuroprotective effect appears on developing brain of very early preterm. Combining capture-resistance experiment, suspension experiment, open field test and Morris water maze test, we observed the effects of neurocognitive and motor function items by lepin in premature brain damage rats from 2 days postnatal to 21 days juvenile rat. Univariate and multivariate analysis (using multiple linear regression model) showed leptin alleviate the spatial memory impairment of premature brain damage, independent from gender and weight. These findings have important implications for our standing of leptin neuroprotective effects on preterm with brain damage.

In this paper, we investigate a challenging but interesting task in the research of speech emotion recognition (SER), i.e., cross-corpus SER. Unlike the conventional SER, the training (source) and testing (target) samples in cross-corpus SER come from different speech corpora, which results in a feature distribution mismatch between them. Hence, the performance of most existing SER methods may sharply decrease. To cope with this problem, we propose a simple yet effective deep transfer learning method called progressive distribution adapted neural networks (PDAN). PDAN employs convolutional neural networks (CNN) as the backbone and the speech spectrum as the inputs to achieve an end-to-end learning framework. More importantly, its basic idea for solving cross-corpus SER is very straightforward, i.e., enhancing the backbone's corpus invariant feature learning ability by incorporating a progressive distribution adapted regularization term into the original loss function to guide the network training. To evaluate the proposed PDAN, extensive cross-corpus SER experiments on speech emotion corpora including EmoDB, eNTERFACE, and CASIA are conducted. Experimental results showed that the proposed PDAN outperforms most well-performing deep and subspace transfer learning methods in dealing with the cross-corpus SER tasks.