Abstract-Deep neural networks (DNNs) are now a central component of nearly all state-of-the-art speech recognition systems. Building neural network acoustic models requires several design decisions including network architecture, size, and training loss function. This paper offers an empirical investigation on which aspects of DNN acoustic model design are most important for speech recognition system performance. We report DNN classifier performance and final speech recognizer word error rates, and compare DNNs using several metrics to quantify factors influencing differences in task performance. Our first set of experiments use the standard Switchboard benchmark corpus, which contains approximately 300 hours of conversational telephone speech. We compare standard DNNs to convolutional networks, and present the first experiments using locally-connected, untied neural networks for acoustic modeling. We additionally build systems on a corpus of 2,100 hours of training data by combining the Switchboard and Fisher corpora. This larger corpus allows us to more thoroughly examine performance of large DNN models -with up to ten times more parameters than those typically used in speech recognition systems. Our results suggest that a relatively simple DNN architecture and optimization technique produces strong results. These findings, along with previous work, help establish a set of best practices for building DNN hybrid speech recognition systems with maximum likelihood training. Our experiments in DNN optimization additionally serve as a case study for training DNNs with discriminative loss functions for speech tasks, as well as DNN classifiers more generally.

Mitochondrial dysfunction and oxidative stress damage are hallmarks of osteoarthritis (OA). Mesenchymal stem cell (MSC)-derived exosomes are important in intercellular mitochondria communication. However, the use of MSC exosomes for regulating mitochondrial function in OA has not been reported. This study aimed to explore the therapeutic effect of MSC exosomes in a three dimensional (3D) printed scaffold for early OA therapeutics. Methods : We first examined the mitochondria-related proteins in normal and OA human cartilage samples and investigated whether MSC exosomes could enhance mitochondrial biogenesis in vitro . We subsequently designed a bio-scaffold for MSC exosomes delivery and fabricated a 3D printed cartilage extracellular matrix (ECM)/gelatin methacrylate (GelMA)/exosome scaffold with radially oriented channels using desktop-stereolithography technology. Finally, the osteochondral defect repair capacity of the 3D printed scaffold was assessed using a rabbit model. Results : The ECM/GelMA/exosome scaffold effectively restored chondrocyte mitochondrial dysfunction, enhanced chondrocyte migration, and polarized the synovial macrophage response toward an M2 phenotype. The 3D printed scaffold significantly facilitated the cartilage regeneration in the animal model. Conclusion : This study demonstrated that the 3D printed, radially oriented ECM/GelMA/exosome scaffold could be a promising strategy for early OA treatment.

ObjectivesCircular RNAs (circRNA) expression aberration has been identified in various human diseases. In this study, we investigated whether circRNAs could act as competing endogenous RNAs to regulate the pathological process of osteoarthritis (OA).MethodsCircRNA deep sequencing was performed to the expression of circRNAs between OA and control cartilage tissues. The regulatory and functional role of CircSERPINE2 upregulation was examined in OA and was validated in vitro and in vivo, downstream target of CircSERPINE2 was explored. RNA pull down, a luciferase reporter assay, biotin-coupled microRNA capture and fluorescence in situ hybridisation were used to evaluate the interaction between CircSERPINE2 and miR-1271-5 p, as well as the target mRNA, E26 transformation-specific-related gene (ERG). The role and mechanism of CircSERPINE2 in OA was also explored in rabbit models.ResultsThe decreased expression of CircSERPINE2 in the OA cartilage tissues was directly associated with excessive apoptosis and imbalance between anabolic and catabolic factors of extracellular matrix (ECM). Mechanistically, CircSERPINE2 acted as a sponge of miR-1271-5 p and functioned in human chondrocytes (HCs) through targeting miR-1271-5 p and ERG. Intra-articular injection of adeno-associated virus-CircSERPINE2-wt alleviated OA in the rabbit model.ConclusionsOur results reveal an important role for a novel circRNA-CircSERPINE2 in OA progression. CircSERPINE2 overexpression could alleviate HCs apoptosis and promote anabolism of ECM through miR-1271-ERG pathway. It provides a potentially effective therapeutic strategy for OA progression.