A Spatio-Temporal Ensemble Deep Learning Architecture for Real-Time Defect Detection during Laser Welding on Low Power Embedded Computing Boards.

In modern production environments, advanced and intelligent process monitoring strategies are required to enable an unambiguous diagnosis of the process situation and thus of the final component quality. In addition, the ability to recognize the current state of product quality in real-time is an important prerequisite for autonomous and self-improving manufacturing systems. To address these needs, this study investigates a novel ensemble deep learning architecture based on convolutional neural networks (CNN), gated recurrent units (GRU) combined with high-performance classification algorithms such as k-nearest neighbors (kNN) and support vector machines (SVM). The architecture uses spatio-temporal features extracted from infrared image sequences to locate critical welding defects including lack of fusion (false friends), sagging, lack of penetration, and geometric deviations of the weld seam. In order to evaluate the proposed architecture, this study investigates a comprehensive scheme based on classical machine learning methods using manual feature extraction and state-of-the-art deep learning algorithms. Optimal hyperparameters for each algorithm are determined by an extensive grid search. Additional work is conducted to investigate the significance of various geometrical, statistical and spatio-temporal features extracted from the keyhole and weld pool regions. The proposed method is finally validated on previously unknown welding trials, achieving the highest detection rates and the most robust weld defect recognition among all classification methods investigated in this work. Ultimately, the ensemble deep neural network is implemented and optimized to operate on low-power embedded computing devices with low latency (1.1 ms), demonstrating sufficient performance for real-time applications.

Combined anticancer effects of sphingosine kinase inhibitors and sorafenib.

The pro-apoptotic lipid sphingosine is phosphorylated by sphingosine kinases 1 and 2 (SK1 and SK2) to generate the mitogenic lipid sphingosine-1-phosphate (S1P). We previously reported that inhibition of SK activity delays tumor growth in a mouse mammary adenocarcinoma model. Because SK inhibitors and the multikinase inhibitor sorafenib both suppress the MAP kinase pathway, we hypothesized that their combination may provide enhanced inhibition of tumor growth. Therefore, we evaluated the effects of two SK inhibitors, ABC294640 (a SK2-specific inhibitor) and ABC294735 (a dual SK1/SK2 inhibitor), alone and in combination with sorafenib on human pancreatic adenocarcinoma (Bxpc-3) and kidney carcinoma (A-498) cells in vitro and in vivo. Exposure of either Bxpc-3 or A-498 cells to combinations of ABC294640 and sorafenib or ABC294735 and sorafenib resulted in synergistic cytotoxicity, associated with activation of caspases 3/7 and DNA fragmentation. Additionally, strong decreases in ERK phosphorylation were observed in Bxpc-3 and A-498 cells exposed to either the sorafenib/ABC294640 or the sorafenib/ABC294735 combination. Oral administration of either ABC294640 or ABC294735 to mice led to a delay in tumor growth in both xenograft models without overt toxicity to the animals. Tumor growth delay was potentiated by co-administration of sorafenib. These studies show that combination of an SK inhibitor with sorafenib causes synergistic inhibition of cell growth in vitro, and potentiates antitumor activity in vivo. Thus, a foundation is established for clinical trials evaluating the efficacy of combining these signaling inhibitors.

A novel sphingosine kinase inhibitor induces autophagy in tumor cells.

The sphingolipids ceramide, sphingosine, and sphingosine 1-phosphate (S1P) regulate cell signaling, proliferation, apoptosis, and autophagy. Sphingosine kinase-1 and -2 (SK1 and SK2) phosphorylate sphingosine to form S1P, shifting the balanced activity of these lipids toward cell proliferation. We have previously reported that pharmacological inhibition of SK activity delays tumor growth in vivo. The present studies demonstrate that the SK2-selective inhibitor 3-(4-chlorophenyl)-adamantane-1-carboxylic acid (pyridin-4-ylmethyl)amide (ABC294640) induces nonapoptotic cell death that is preceded by microtubule-associated protein light chain 3 cleavage, morphological changes in lysosomes, formation of autophagosomes, and increases in acidic vesicles in A-498 kidney carcinoma cells. ABC294640 caused similar autophagic responses in PC-3 prostate and MDA-MB-231 breast adenocarcinoma cells. Simultaneous exposure of A-498 cells to ABC294640 and 3-methyladenine, an inhibitor of autophagy, switched the mechanism of toxicity to apoptosis, but decreased the potency of the SK2 inhibitor, indicating that autophagy is a major mechanism for tumor cell killing by this compound. Induction of the unfolded protein response by the proteasome inhibitor N-(benzyloxycarbonyl)leucinylleucinylleucinal Z-Leu-Leu-Leu-al (MG-132) or the heat shock protein 90 inhibitor geldanamycin synergistically increased the cytotoxicity of ABC294640 in vitro. In severe combined immunodeficient mice bearing A-498 xenografts, daily administration of ABC294640 delayed tumor growth and elevated autophagy markers, but did not increase terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling-positive cells in the tumors. These data suggest that ABC294640 promotes tumor cell autophagy, which ultimately results in nonapoptotic cell death and a delay of tumor growth in vivo. Consequently, ABC294640 may effectively complement anticancer drugs that induce tumor cell apoptosis.

Complement receptor 2-mediated targeting of complement inhibitors to sites of complement activation.

In a strategy to specifically target complement inhibitors to sites of complement activation and disease, recombinant fusion proteins consisting of a complement inhibitor linked to a C3 binding region of complement receptor (CR) 2 were prepared and characterized. Natural ligands for CR2 are C3 breakdown products deposited at sites of complement activation. Fusion proteins were prepared consisting of a human CR2 fragment linked to either the N terminus or C terminus of soluble forms of the membrane complement inhibitors decay accelerating factor (DAF) or CD59. The targeted complement inhibitors bound to C3-opsonized cells, and all were significantly more effective (up to 20-fold) than corresponding untargeted inhibitors at protecting target cells from complement. CR2 fusion proteins also inhibited CR3-dependent adhesion of U937 cells to C3 opsonized erythrocytes, indicating a second potential anti-inflammatory mechanism of CR2 fusion proteins, since CR3 is involved in endothelial adhesion and diapedesis of leukocytes at inflammatory sites. Finally, the in vivo validity of the targeting strategy was confirmed by the demonstration that CR2-DAF, but not soluble DAF, targets to the kidney in mouse models of lupus nephritis that are associated with renal complement deposition.

Synthesis and evaluation of novel inhibitors of Pim-1 and Pim-2 protein kinases.

The Pim protein kinases are frequently overexpressed in prostate cancer and certain forms of leukemia and lymphoma. 5-(3-Trifluoromethylbenzylidene)thiazolidine-2,4-dione (4a) was identified by screening to be a Pim-1 inhibitor and was found to attenuate the autophosphorylation of tagged Pim-1 in intact cells. Although 4a is a competitive inhibitor with respect to ATP, a screen of approximately 50 diverse protein kinases demonstrated that it has high selectivity for Pim kinases. Computational docking of 4a to Pim-1 provided a model for lead optimization, and a series of substituted thiazolidine-2,4-dione congeners was synthesized. The most potent new compounds exhibited IC(50)s of 13 nM for Pim-1 and 2.3 microM for Pim-2. Additional compounds in the series demonstrated selectivities of more than 2500-fold and 400-fold for Pim-1 or Pim-2, respectively, while other congeners were essentially equally potent toward the two isozymes. Overall, these compounds are new Pim kinase inhibitors that may provide leads to novel anticancer agents.

Megalin functions as an endocytic sonic hedgehog receptor.

Embryos deficient in the morphogen Sonic hedgehog (Shh) or the endocytic receptor megalin exhibit common neurodevelopmental abnormalities. Therefore, we have investigated the possibility that a functional relationship exists between the two proteins. During embryonic development, megalin was found to be expressed along the apical surfaces of neuroepithelial cells and was coexpressed with Shh in the ventral floor plate of the neural tube. Using enzyme-linked immunosorbent assay, homologous ligand displacement, and surface plasmon resonance techniques, it was found that the amino-terminal fragment of Shh (N-Shh) bound to megalin with high affinity. Megalin-expressing cells internalized N-Shh through a mechanism that was inhibited by antagonists of megalin, viz. anti-receptor-associated protein and anti-megalin antibodies. Heparin also inhibited N-Shh endocytosis, implicating proteoglycans in the internalization process, as has been described for other megalin ligands. Use of chloroquine to inhibit lysosomal proteinase activity showed that N-Shh endocytosed via megalin was not efficiently targeted to the lysosomes for degradation. The ability of megalin-internalized N-Shh to bypass lysosomes may relate to the finding that the interaction between N-Shh and megalin was resistant to dissociation with low pH. Together, these findings show that megalin is an efficient endocytic receptor for N-Shh. Furthermore, they implicate megalin as a new regulatory component of the Shh signaling pathway.