We present MorphNet, an approach to automate the design of neural network structures. MorphNet iteratively shrinks and expands a network, shrinking via a resourceweighted sparsifying regularizer on activations and expanding via a uniform multiplicative factor on all layers. In contrast to previous approaches, our method is scalable to large networks, adaptable to specific resource constraints (e.g. the number of floating-point operations per inference), and capable of increasing the network's performance. When applied to standard network architectures on a wide variety of datasets, our approach discovers novel structures in each domain, obtaining higher performance while respecting the resource constraint.

Strategies for selectively imaging and delivering drugs to tumours typically leverage differentially upregulated surface molecules on cancer cells. Here, we show that intravenously injected carbon quantum dots, functionalized with multiple paired α-carboxyl and amino groups that bind to the large neutral amino acid transporter 1 (which is expressed in most tumours), selectively accumulate in human tumour xenografts in mice and in an orthotopic mouse model of human glioma. The functionalized quantum dots, which structurally mimic large amino acids and can be loaded with aromatic drugs through π-π stacking interactions, enabled-in the absence of detectable toxicity-near-infrared fluorescence and photoacoustic imaging of the tumours and a reduction in tumour burden after the targeted delivery of chemotherapeutics to the tumours. The versatility of functionalization and high tumour selectivity of the quantum dots make them broadly suitable for tumour-specific imaging and drug delivery.

The synthesis, optical properties, and biomedical applications of CDs are summarized.

Both phosphatidylinositol 3-kinase (PI3K)/AKT and mitogen activated protein kinase (MAPK) signaling cascades play an important role in cell proliferation, survival, angiogenesis, and metastasis of tumor cells. In the present report, we investigated the effects of licochalcone A (LA), a flavonoid extracted from licorice root, on the PI3K/AKT/mTOR and MAPK activation pathways in human gastric cancer BGC-823 cells. LA increased reactive oxygen species (ROS) levels, which is associated with the induction of apoptosis as characterized by positive Annexin V binding and activation of caspase-3, and cleavage of poly-ADP-ribose polymerase (PARP). Inhibition of ROS generation by N-acetylcysteine (NAC) significantly prevented LA-induced apoptosis. Interestingly, we also observed that LA caused the activation of ERK, JNK, and p38 MAPK in BGC-823 cells. The antitumour activity of LA-treated BGC-823 cells was significantly distinct in KM mice in vivo. All the findings from our study suggest that LA can interfere with MAPK signaling cascades, initiate ROS generation, induce oxidative stress and consequently cause BGC cell apoptosis.

The cardiac ischemia-reperfusion (I/R) injury greatly influences the therapeutic effect and remains an urgent challenge in clinical therapy. Polypharmacology opens a new therapeutic opportunity to design drugs with a specific target for improving the efficacy. In this study, we first forecasted that Rosmarinic acid (RosA) could be used for the treatment of cardiovascular disease using text mining, chemometric and chemogenomic methods. Consistent with the effect of the positive drug (pioglitazone, PIO), we subsequently validated that RosA pretreatment could restore the decreased cardiac hemodynamic parameters (LVDP, ± dp/dtmin, ± dp/dtmax and CF), decreased the infarct size and the cardiomyocyte apoptosis in a rat model of cardiac I/R injury. Furthermore, RosA pre-treatment inhibited the levels of inflammatory cytokines (IL-6, TNF-α and CRP), up-regulated PPARγ expression and down-regulated NF-κB expression in myocardial tissue isolated from the rat model of I/R-induced myocardial injury. In addition, the effects of RosA were reversed by co-treatment with PPAR-γ inhibitor GW9662 and T0070907, respectively. These data suggest that RosA attenuates cardiac injury through activating PPARγ and down-regulating NF-κB-mediated signaling pathway, which inhibiting inflammation and cardiomyocyte apoptosis in a rat model of cardiac I/R injury.

Detecting anomaly of chest X-ray images by advanced technologies, such as deep learning, is an urgent need to improve the work efficiency and diagnosis accuracy. Fine-tuning existing deep learning networks for medical image processing suffers from over-fitting and low transfer efficiency. To overcome such limitations, we design a hierarchical convolutional neural network (CNN) structure for ChestX-ray14 and propose a new network CXNet-m1, which is much shorter, thinner but more powerful than fine-tuning. We also raise a novel loss function sin-loss, which can learn discriminative information from misclassified and indistinguishable images. Besides, we optimize the convolutional kernels of CXNet-m1 to achieve better classification accuracy. The experimental results show that our light model CXNet-m1 with sin-loss function achieves better accuracy rate, recall rate, F1-score, and AUC value. It illustrates that designing a proper CNN is better than fine-tuning deep networks, and the increase of training data is vital to enhance the performance of CNN. INDEX TERMS Chest X-Rays image, anomaly detection, deep neural network, self-adapting loss function. HAO WU received the B.E. and Ph.D. degrees

In recent years, carbon dots (CDs), including carbon nanodots, carbonized polymer dots, carbon quantum dots, and graphene quantum dots have attracted a mounting interest as readily accessible, nontoxic, and relatively inexpensive carbon-based nanomaterials. Yet, despite intense research for a number of years, a unifying picture is still lacking to clarify the exact definition, clear chemical structure, and unique optical properties of this family of nanomaterials. In this review, we systematically summarize the recent development of CDs from molecular design to related properties of excited states as well as their applications in optoelectronic devices and biology. We point out the current challenges, including exploring precise synthesis, clarifying the structure-property relationship, and regulating singlet and triplet states of fluorescence, phosphorescence, and delayed fluorescence. Moreover, the structural optimization of optoelectronic devices, tumor targeting mechanism, selective imaging, and drug delivery of CDs are also highlighted. We hope that the information provided in this review will inspire more exciting research on CDs from a brand-new perspective and promote practical application of CDs in multiple directions of current and future research.

In this paper, we present an RIP module with the features of supporting multiple inductive sensors, no variable frequency LC oscillator, low power consumption, and automatic gain adjustment for each channel. Based on the method of inductance measurement without using a variable frequency LC oscillator, we further integrate pulse amplitude modulation and time division multiplexing scheme into a module to support multiple RIP sensors. All inductive sensors are excited by a high-frequency electric current periodically and momentarily, and the inductance of each sensor is measured during the time when the electric current is fed to it. To improve the amplitude response of the RIP sensors, we optimize the sensing unit with a matching capacitor parallel with each RIP sensor forming a frequency selection filter. Performance tests on the linearity of the output with cross-sectional area and the accuracy of respiratory volume estimation demonstrate good linearity and accurate lung volume estimation. Power consumption of this new RIP module with two sensors is very low. The performance of respiration measurement during movement is also evaluated. This RIP module is especially desirable for wearable systems with multiple RIP sensors for long-term respiration monitoring.