Analysis of Training Deep Learning Models for PCB Defect Detection.

Recently, many companies have introduced automated defect detection methods for defect-free PCB manufacturing. In particular, deep learning-based image understanding methods are very widely used. In this study, we present an analysis of training deep learning models to perform PCB defect detection stably. To this end, we first summarize the characteristics of industrial images, such as PCB images. Then, the factors that can cause changes (contamination and quality degradation) to the image data in the industrial field are analyzed. Subsequently, we organize defect detection methods that can be applied according to the situation and purpose of PCB defect detection. In addition, we review the characteristics of each method in detail. Our experimental results demonstrated the impact of various degradation factors, such as defect detection methods, data quality, and image contamination. Based on our overview of PCB defect detection and experiment results, we present knowledge and guidelines for correct PCB defect detection.

Machine-Learning Model for the Prediction of Hypoxaemia during Endoscopic Retrograde Cholangiopancreatography under Monitored Anaesthesia Care.

PURPOSE: Hypoxaemia is a significant adverse event during endoscopic retrograde cholangiopancreatography (ERCP) under monitored anaesthesia care (MAC); however, no model has been developed to predict hypoxaemia. We aimed to develop and compare logistic regression (LR) and machine learning (ML) models to predict hypoxaemia during ERCP under MAC. MATERIALS AND METHODS: We collected patient data from our institutional ERCP database. The study population was randomly divided into training and test sets (7:3). Models were fit to training data and evaluated on unseen test data. The training set was further split into k-fold (k=5) for tuning hyperparameters, such as feature selection and early stopping. Models were trained over k loops; the i-th fold was set aside as a validation set in the i-th loop. Model performance was measured using area under the curve (AUC). RESULTS: We identified 6114 cases of ERCP under MAC, with a total hypoxaemia rate of 5.9%. The LR model was established by combining eight variables and had a test AUC of 0.693. The ML and LR models were evaluated on 30 independent data splits. The average test AUC for LR was 0.7230, which improved to 0.7336 by adding eight more variables with an l1 regularisation-based selection technique and ensembling the LRs and gradient boosting algorithm (GBM). The high-risk group was discriminated using the GBM ensemble model, with a sensitivity and specificity of 63.6% and 72.2%, respectively. CONCLUSION: We established GBM ensemble model and LR model for risk prediction, which demonstrated good potential for preventing hypoxaemia during ERCP under MAC.

TFF3 interacts with LINGO2 to regulate EGFR activation for protection against colitis and gastrointestinal helminths.

Intestinal epithelial cells (IEC) have important functions in nutrient absorption, barrier integrity, regeneration, pathogen-sensing, and mucus secretion. Goblet cells are a specialized cell type of IEC that secrete Trefoil factor 3 (TFF3) to regulate mucus viscosity and wound healing, but whether TFF3-responsiveness requires a receptor is unclear. Here, we show that leucine rich repeat receptor and nogo-interacting protein 2 (LINGO2) is essential for TFF3-mediated functions. LINGO2 immunoprecipitates with TFF3, co-localizes with TFF3 on the cell membrane of IEC, and allows TFF3 to block apoptosis. We further show that TFF3-LINGO2 interactions disrupt EGFR-LINGO2 complexes resulting in enhanced EGFR signaling. Excessive basal EGFR activation in Lingo2 deficient mice increases disease severity during colitis and augments immunity against helminth infection. Conversely, TFF3 deficiency reduces helminth immunity. Thus, TFF3-LINGO2 interactions de-repress inhibitory LINGO2-EGFR complexes, allowing TFF3 to drive wound healing and immunity.

Validation of the Korean Version of the Manchester-Oxford Foot Questionnaire in Patients With Hallux Valgus.

The Manchester-Oxford Foot Questionnaire (MOXFQ) is a practical, reliable, and valid questionnaire for hallux valgus surgery and has been translated into several languages. However, the MOXFQ has not been translated into Korean. In the present study, we aimed to translate and evaluate the validity and reliability of the Korean version of the MOXFQ for patients affected by hallux valgus. In accordance with the guidelines of cross-cultural adaptation, we translated the English version of MOXFQ into Korean and then backward translated it into English. We sent out letters that included the Korean version of the MOXFQ, a visual analog scale measure of pain, and a validated Korean version of the short-form 36-item Health Survey to 135 patients with hallux valgus. A retest was administered after 2 weeks. Of the 135 patients, 104 responded to the first questionnaire, and 82 of the first-time responders returned their second questionnaires. We evaluated the test-retest reliability, internal consistency, concurrent validity, and construct validity of the Korean version of the MOXFQ. The intraclass correlation coefficient for test-retest reliability was 0.82 for the total MOXFQ and ranged from 0.81 to 0.82 for the 3 subscales. Cronbach's alpha for the total MOXFQ was 0.85 and ranged from 0.8 to 0.92 for the 3 subscales. Concurrent and construct validity was supported by significant correlation with the visual analog scale and short-form 36-item Health Survey subscale scores. The Korean version of the MOXFQ was tested, and it was found to be a valid and reliable instrument for patients with hallux valgus.

Reliability and Validity of the Korean Version of the Foot Function Index.

Although translated versions of the Foot Function Index (FFI) in several languages are available, the absence of a Korean version has precluded comparing the data from Korea with the data from other countries using the FFI. We, therefore, evaluated the reliability and validity of the adapted Korean version of the FFI. We translated the English version of the FFI into Korean and back into English. We mailed the Korean version of the visual analog scale, FFI, and the previously validated Medical Outcomes Study Short-Form 36-item questionnaire (SF-36) to 121 patients with treated foot complaints. To evaluate the test-retest reliability and internal consistency, we used the intraclass correlation coefficient and Cronbach's α, respectively. We also evaluated the concurrent and construct validity of Korean version of the FFI by comparing the visual analog scale and SF-36. Cronbach's α was 0.91 and 0.95 for the pain and disability subscales, respectively. The reproducibility was good, and a strong correlation between the FFI and the SF-36 and visual analog scale with related content was observed, indicating good construct validity.The Korean version of the FFI is a reliable and valid questionnaire for the self-assessment of pain and disability in Korean patients with foot complaints.

Impact of Statin Use on Development of New-Onset Diabetes Mellitus in Asian Population.

There have been several reports showing that the statin use is associated with new-onset diabetes mellitus (DM). The aim of the present study was to evaluate the impact of chronic statin use on development of new-onset DM in a series of Asian population. The patients were retrospectively enrolled using the electronic database of Korea University Guro Hospital from January 2004 to February 2010. A total of 10,994 patients without a history of diabetes were analyzed. Baseline lipid profiles, fasting glucose, Hemoglobin (Hb) A1c, and glucose tolerance tests were measured in all patients before statin treatment. Included patients had HbA1c ≤ 5.7% and fasting glucose level ≤ 100 (mg/dl). The patients were divided into 2 groups according to the use of statins (the statin group, n = 2,324 patients and the nonstatin group, n = 8,670 patients). To adjust baseline potential confounders, a propensity score-matched analysis was performed using logistic regression model. After propensity score matching, 2 propensity-matched groups (1,699 pairs, n = 3,398, C statistic = 0.859) were generated and analyzed. After propensity score matching, baseline characteristics of both groups were balanced except that the statin group was older and had higher rate of coronary artery disease compared with the nonstatin group. During a 3-year follow-up, the statin group had higher incidence of new-onset DM compared with the nonstatin group (hazard ratio 1.99, 95% CI 1.36 to 2.92, p <0.001), but the statin group showed lower incidence of major adverse cerebral-cardiovascular events compared with the nonstatin group (hazard ratio 0.40, 95% CI 0.19 to 0.85, p <0.001). In the present study, although the use of statins was associated with higher rate of new-onset DM, it markedly improved 3-year cardiovascular outcomes in Asian population.

DNA-PK triggers histone ubiquitination and signaling in response to DNA double-strand breaks produced during the repair of transcription-blocking topoisomerase I lesions.

Although defective repair of DNA double-strand breaks (DSBs) leads to neurodegenerative diseases, the processes underlying their production and signaling in non-replicating cells are largely unknown. Stabilized topoisomerase I cleavage complexes (Top1cc) by natural compounds or common DNA alterations are transcription-blocking lesions whose repair depends primarily on Top1 proteolysis and excision by tyrosyl-DNA phosphodiesterase-1 (TDP1). We previously reported that stabilized Top1cc produce transcription-dependent DSBs that activate ATM in neurons. Here, we use camptothecin (CPT)-treated serum-starved quiescent cells to induce transcription-blocking Top1cc and show that those DSBs are generated during Top1cc repair from Top1 peptide-linked DNA single-strand breaks generated after Top1 proteolysis and before excision by TDP1. Following DSB induction, ATM activates DNA-PK whose inhibition suppresses H2AX and H2A ubiquitination and the later assembly of activated ATM into nuclear foci. Inhibition of DNA-PK also reduces Top1 ubiquitination and proteolysis as well as resumption of RNA synthesis suggesting that DSB signaling further enhances Top1cc repair. Finally, we show that co-transcriptional DSBs kill quiescent cells. Together, these new findings reveal that DSB production and signaling by transcription-blocking Top1 lesions impact on non-replicating cell fate and provide insights on the molecular pathogenesis of neurodegenerative diseases such as SCAN1 and AT syndromes, which are caused by TDP1 and ATM deficiency, respectively.

Effect of a Dose-Escalation Regimen for Improving Adherence to Roflumilast in Patients with Chronic Obstructive Pulmonary Disease.

BACKGROUND: The adverse effects of the phosphodiesterase-4 inhibitor roflumilast, appear to be more frequent in clinical practice than what was observed in chronic obstructive pulmonary disease (COPD) clinical trials. Thus, we designed this study to determine whether adverse effects could be reduced by starting roflumilast at half the dose, and then increasing a few weeks later to 500 µg daily. METHODS: We retrospectively investigated 85 patients with COPD who had taken either 500 µg roflumilast, or a starting dose of 250 µg and then increased to 500 µg. We analyzed all adverse events and assessed differences between patients who continued taking the drug after dose escalation and those who had stopped. RESULTS: Adverse events were reported by 22 of the 85 patients (25.9%). The most common adverse event was diarrhea (10.6%). Of the 52 patients who had increased from a starting dose of 250 µg roflumilast to 500 µg, 43 (82.7%) successfully maintained the 500 µg roflumilast dose. No difference in factors likely to affect the risk of adverse effects, was detected between the dose-escalated and the discontinued groups. Of the 26 patients who started with the 500 µg roflumilast regimen, seven (26.9%) discontinued because of adverse effects. There was no statistically significant difference in discontinuation rate between the dose-escalated and the control groups (p=0.22). CONCLUSION: Escalating the roflumilast dose may reduce treatment-related adverse effects and improve tolerance to the full dose. This study suggests that the dose-escalated regimen reduced the rate of discontinuation. However, longer-term and larger-scale studies are needed to support the full benefit of a dose escalation strategy.

Identification and characterization of Paragonimus westermani leucine aminopeptidase.

Paragonimus westermani is a tissue-invading trematode parasite that causes inflammatory lung disease as well as systemic infections including cerebral invasion in carnivorous mammals. While aminopeptidases play important roles in trematodes in the catabolism of host hemoglobin, an essential source of nutrient for the parasite, little is known about aminopeptidase in Paragonimus. Presently, we isolated a cDNA encoding a 58 kDa P. westermani leucine aminopeptidase (PwLAP). Deduced amino acid sequence of PwLAP exhibited significant sequence homology with LAP from Schistosoma spp. and Fasciola hepatica. Biochemical analysis of the recombinant PwLAP protein demonstrated preferential substrate specificity for Leu-NHMec and inhibition by EDTA, 1,10-phenanthroline, and bestatin, which are conserved characteristics of the M17 family of leucine aminopeptidase. PwLAP exhibited relatively higher enzyme activity in the presence of Mn2+ compared to Schistosoma mansoni LAP. Based on the biochemical properties and immunohistochemical analysis, PwLAP is concluded to represent a leucine aminopeptidase. The enzyme is most likely responsible for the catabolism of host hemoglobin, and, hence, represents a potential target of Paragonimus chemotherapy.