Efficient Mixed-Type Wafer Defect Pattern Recognition Based on Light-Weight Neural Network.

Wafer defect pattern recognition can help engineers improve the production process of semiconductor chips. In real industrial scenarios, the recognition of mixed-type wafer defects is difficult and the production scale of semiconductor wafers is large, which requires high accuracy and speed in wafer defect pattern recognition. This study proposes a light-weight neural network model to efficiently recognize mixed-type wafer defects. The proposed model is constructed via inverted residual convolution blocks with attention mechanisms and large kernel convolution downsampling layers. The inference speed of the inverted residual convolution block is fast, and the attention mechanism can enhance feature extraction capabilities. Large kernel convolutions help the network retain more important feature information during downsampling operations. The experimental results on the real Mixed-type WM38 dataset show that the proposed model achieves a recognition accuracy of 98.69% with only 1.01 M parameters. Compared with some popular high-performance models and light-weight models, our model has advantages in both recognition accuracy and inference speed. Finally, we deploy the model as a TensorRT engine, which significantly improves the inference speed of the model, enabling it to process more than 1300 wafer maps per second.

Molecular epidemiology and genomic dynamics of Pseudomonas aeruginosa isolates causing relapse infections.

Pseudomonas aeruginosa (P. aeruginosa) is one of the leading causes of chronic infections, including reinfection, relapse, and persistent infection, especially in cystic fibrosis patients. Relapse P. aeruginosa infections are more harmful because of repeated hospitalization and undertreatment of antimicrobials. However, relapse P. aeruginosa infection in China remains largely unknown. Herein, we performed a 3-year retrospective study from 2019 to 2022 in a tertiary hospital, which included 442 P. aeruginosa isolates from 196 patients. Relapse infection was identified by screening clinical records and whole-genome sequencing (WGS). We found that 31.6% (62/196) of patients had relapsed infections. The relapse incidence of carbapenem-resistant P. aeruginosa infection (51.4%) is significantly higher than that of carbapenem-susceptible P. aeruginosa infection (20.2%, P < 0.0001). These isolates were assigned to 50 distinct sequence types and sporadically distributed in phylogeny, indicating that relapsed infections were not caused by certain lineages. Fast adaptation and evolution of P. aeruginosa isolates were reflected by dynamic changes of antimicrobial resistance, gene loss and acquisition, and single-nucleotide polymorphisms during relapse episodes. Remarkably, a convergent non-synonymous mutation that occurs in a pyochelin-associated virulence gene fptA (T1056C, M252T) could be a considerable target for the diagnosis and treatment of relapse P. aeruginosa infection. These findings suggest that integrated utilization of WGS and medical records provides opportunities for improved diagnostics of relapsed infections. Continued surveillance of the genomic dynamics of relapse P. aeruginosa infection will generate further knowledge for optimizing treatment and prevention in the future.IMPORTANCEPseudomonas aeruginosa is a predominant pathogen that causes various chronic infections. Relapse infections promote the adaptation and evolution of antimicrobial resistance and virulence of P. aeruginosa, which obscure evolutionary trends and complicate infection management. We observed a high incidence of relapse P. aeruginosa infection in this study. Whole-genome sequencing (WGS) revealed that relapse infections were not caused by certain lineages of P. aeruginosa isolates. Genomic dynamics of relapse P. aeruginosa among early and later stages reflected a plasticity scattered through the entire genome and fast adaptation and genomic evolution in different ways. Remarkably, a convergent evolution was found in a significant virulence gene fptA, which could be a considerable target for diagnosis and treatment. Taken together, our findings highlight the importance of longitudinal surveillance of relapse P. aeruginosa infection in China since cystic fibrosis is rare in Chinese. Integrated utilization of WGS and medical records provides opportunities for improved diagnostics of relapse infections.

Francisella salimarina sp. nov., isolated from coastal seawater.

Four strains (SYSU SYW-1T, SYW-2, SYW-3 and XLW-1) were isolated from seawater near the shore in Guangdong Province, China. Cells were Gram-stain-negative, aerobic, non-motile and non-spore-forming. Growth was observed at a temperature range of 16-40 °C (optimum, 32 °C), a pH range of 4-8 (optimum, pH 7) and in the presence of up to 10 % (w/v) NaCl. The major polar lipids were diphosphatidylglycerol, phosphatidylcholine, phosphatidylethanolamine and an unidentified phospholipid. The respiratory quinone was ubiquinone 8 (UQ-8), and the predominant fatty acids were C18 : 0 3-OH, C10 : 0, C14 : 0 and C18 : 1ω9c. Comparison of 16S rRNA gene and genome sequences confirmed that these strains represented a novel member of the genus Francisella, with less than 98.8 % 16S rRNA gene sequence similarity and less than 95 % genomic average nucleotide identity to recognized Francisella species. The phylogenetic tree based on 16S rRNA gene sequences and the protein-concatamer tree based on a concatenation of 28 protein marker sequences both indicated that the strains clustered with 'Francisella salina' TX07-7308 and 'Francisella marina' E95-16, but formed a distinct lineage group among the other members of the genus Francisella. The DNA G+C contents of the four strains were determined to be 32.9, 32.7, 32.9 and 32.9 %, respectively (genome). On the basis of phenotypic and genotypic features, the strains are considered to represent a novel species of the genus Francisella, for which the name Francisella salimarina sp. nov. is proposed. The type strain is SYSU SYW-1T (=CGMCC 1.17031T=NBRC 113781T).