An Intelligent Diagnostic Model for Melasma Based on Deep Learning and Multimode Image Input.

INTRODUCTION: The diagnosis of melasma is often based on the naked-eye judgment of physicians. However, this is a challenge for inexperienced physicians and non-professionals, and incorrect treatment might have serious consequences. Therefore, it is important to develop an accurate method for melasma diagnosis. The objective of this study is to develop and validate an intelligent diagnostic system based on deep learning for melasma images. METHODS: A total of 8010 images in the VISIA system, comprising 4005 images of patients with melasma and 4005 images of patients without melasma, were collected for training and testing. Inspired by four high-performance structures (i.e., DenseNet, ResNet, Swin Transformer, and MobileNet), the performances of deep learning models in melasma and non-melasma binary classifiers were evaluated. Furthermore, considering that there were five modes of images for each shot in VISIA, we fused these modes via multichannel image input in different combinations to explore whether multimode images could improve network performance. RESULTS: The proposed network based on DenseNet121 achieved the best performance with an accuracy of 93.68% and an area under the curve (AUC) of 97.86% on the test set for the melasma classifier. The results of the Gradient-weighted Class Activation Mapping showed that it was interpretable. In further experiments, for the five modes of the VISIA system, we found the best performing mode to be "BROWN SPOTS." Additionally, the combination of "NORMAL," "BROWN SPOTS," and "UV SPOTS" modes significantly improved the network performance, achieving the highest accuracy of 97.4% and AUC of 99.28%. CONCLUSIONS: In summary, deep learning is feasible for diagnosing melasma. The proposed network not only has excellent performance with clinical images of melasma, but can also acquire high accuracy by using multiple modes of images in VISIA.

A miniaturized ECR plasma flood gun for wafer charge neutralization.

In modern ion implanters, a plasma flood gun (PFG) is used to neutralize wafer charge during the doping process, preventing the breakdown of floating wafers caused by the space charge accumulation. Typically, there are two kinds of PFGs, namely, dc arc discharge with filament and RF discharge. As a PFG, the filament one has limited lifetime and cannot avoid metallic contamination because of the thermal emitting filament. RF discharge PFG has been developed to solve these problems, including prolonging the source lifetime and avoiding metal pollution. Recently, a 2.45 GHz electron cyclotron resonance (ECR) ion source is also regarded as a potential choice for PFG. However, the dimension of the 2.45 GHz ECR source system including the size of the source itself and its meter's length RF subsidiary limits its application within an ion implanter. At Peking University, a miniaturized 2.45 GHz permanent magnet electron cyclotron resonance plasma flood gun with a coaxial RF transmission line has been built and tested. The dimensions of the ECR source body are Φ60 mm × Φ88 mm with a Φ30 mm × Φ40 mm plasma chamber. Its RF transmission line consists of a 200 W microwave generator, a 30 cm coaxial line, a 7 cm coaxial-to-waveguide transducer, and a microwave window that also serves as a vacuum seal. In continuous wave experiments, the electron extraction currents can be as high as 8.8 mA at an input RF power of 22 W with argon gas. The gas flow is less than 1.0 SCCM for this test.

Possibility of generating H+, or H2 +, or H3 + dominated ion beams with a 2.45 GHz permanent magnet ECR ion source.

At Peking University (PKU), experimental research as well as theoretical study on how to produce high intense H+, H2 +, or H3 + dominated ion beams with a compact permanent magnet 2.45 GHz electron cyclotron resonance (PMECR) ion source have been continuously carried out in the past few decades. Based on the comprehension of hydrogen plasma processes inside a 2.45 GHz PMECR discharge chamber, a three-phase diagram of ion fraction dominant regions that illustrates the relationship between the H+, H2 +, and H3 + ion species and working parameters was presented. Meanwhile, a numerical model based on the particle population balance equations was developed for quantitative comprehension of electron cyclotron heated hydrogen plasma. Calculated results of H+, H2 +, and H3 + fractions against gas pressure, microwave density, and wall material obtained with this numerical model agree well with the measured ones. Recently, a miniaturized ECR ion source has been developed, and a 52 mA hydrogen beam was extracted. Under the guidance of the model, H+, H2 +, and H3 + beams with a fraction of 88%, 80%, and 82%, respectively, were obtained with this miniaturized ECR ion source under suitable working parameters. A PMECR ion source for a proton therapy facility has been built at PKU recently. A 34 mA beam H+ fraction of 91% was obtained at the first attempt.

Correlations of MCP-1 -2518A>G polymorphism and serum levels with cerebral infarction risk: a meta-analysis.

This meta-analysis was performed to evaluate the relationships between the monocyte chemoattractant protein-1 (MCP-1) -2518A>G (rs1024611 A>G) polymorphism and its serum levels, and the risk of cerebral infarction. The PubMed, CISCOM, CINAHL, Web of Science, Google Scholar, EBSCO, Cochrane Library, and CBM databases were searched for relevant articles published before October 1st, 2013 without language restrictions. Meta-analysis was conducted using the STATA 12.0 software. Crude odds ratios (ORs) or standardized mean difference (SMD) with their 95% confidence intervals (95% CIs) were calculated. Twelve case-control studies that met all the inclusion criteria were included in this meta-analysis. A total of 1272 patients with cerebral infarction and 1210 healthy control subjects were involved in this meta-analysis. Our meta-analysis results reveal that the MCP-1 -2518A>G polymorphism might increase the risk of cerebral infarction (A allele vs. G allele: OR=1.37, 95% CI: 1.18-1.60, p<0.001; GA+AA vs. GG: OR=1.33, 95% CI: 1.09-1.62, p=0.005; respectively). Furthermore, cerebral infarction patients had higher levels of serum MCP-1 than did healthy control subjects (SMD=2.96, 95% CI: 2.00-3.92, p<0.001). Statistical analysis revealed no evidence of publication bias in this meta-analysis (all p>0.05). Our findings indicate that the MCP-1 -2518A>G polymorphism and serum MCP-1 levels may contribute to the development of cerebral infarction. Thus, the MCP-1 -2518A>G polymorphism and serum MCP-1 levels could be potential biomarkers for the early detection of cerebral infarction.

Genetic polymorphisms in the ESR1 gene and cerebral infarction risk: a meta-analysis.

A number of studies have documented that estrogen receptor α (ESR1) may play an important role in the development and progression of cerebral infarction, but many existing studies have yielded inconclusive results. This meta-analysis was performed to evaluate the relationships between ESR1 genetic polymorphisms and cerebral infarction risk. The PubMed, CISCOM, CINAHL, Web of Science, Google Scholar, EBSCO, Cochrane Library, and CBM databases were searched for relevant articles published before October 1, 2013, without any language restrictions. Meta-analysis was conducted using the STATA 12.0 software. Seven case-control studies were included with a total of 1471 patients with cerebral infarction and 4688 healthy control subjects. Two common single-nucleotide polymorphisms (SNPs) in the ESR1 gene (rs2234693 T>C and rs9340799 A>G) were assessed. Our meta-analysis results revealed that ESR1 genetic polymorphisms might increase the risk of cerebral infarction. Subgroup analysis by SNP type indicated that both rs2234693 and rs9340799 polymorphisms in the ESR1 gene were strongly associated with an increased risk of cerebral infarction. Further subgroup analysis by ethnicity showed significant associations between ESR1 genetic polymorphisms and increased risk of cerebral infarction among both Asians and Caucasians. In the stratified subgroup analysis by gender, the results suggested that ESR1 genetic polymorphisms were associated with an increased risk of cerebral infarction in the female population. However, there were no statistically significant associations between ESR1 genetic polymorphisms and cerebral infarction risk in the male population. Meta-regression analyses also confirmed that gender might be a main source of heterogeneity. Our findings indicate that ESR1 genetic polymorphisms may contribute to the development of cerebral infarction, especially in the female population.