Objective:To develop CNN-Dinov2, a deep learning-based automatic classification model for Gram-stained images, enabling rapid diagnosis of three prevalent Gram-negative bacilli in bloodstream infections: Escherichia coli ( E.coli), Klebsiella pneumoniae ( K.pneumoniae), and Pseudomonas aeruginosa ( P.aeruginosa). Methods:This evaluation study analyzed 1 425 Gram-stained microscopic images from patients with bloodstream infections at Houjie Hospital, in Dongguan City, collected between January 2023 and January 2024. The images, all positive for blood culture and identified as target strains, were categorized into Escherichia coli (419 images), Klebsiella pneumoniae (411 images), Pseudomonas aeruginosa (413 images), and other Gram-negative bacilli (182 images). They were randomly split into a training set (1 141 images), a validation set (141 images), and a test set (143 images) in an 8∶1∶1 ratio. A hybrid CNN-Dinov2 model was developed by integrating ResNet′s local feature extraction with Dinov2′s global pre-trained features, followed by a fully connected layer. The model was optimized by inputting the preprocessed images and adjusting parameters through loss calculation and backpropagation. AlexNet, Dinov2, and ResNet18 served as control models. The models′ classification performance was assessed using accuracy, precision, weighted F1 score, and recall rate, derived from the confusion matrix. The PR curve and AP value further evaluated each model′s classification capability across the four image categories. Results:The CNN-Dinov2 model achieved a training accuracy of 99.74%, a validation accuracy of 98.12%, and a validation loss of 0.070 6, demonstrating robust generalization without overfitting. Validation metrics revealed superior performance with an accuracy of 98.60%, precision of 98.65%, a weighted F1 score of 98.60%, and a recall rate of 98.60%, outperforming other models. The confusion matrix confirmed its strong classification capability, with the highest sum of diagonal values for identifying four types of bacteria. The macro average precision (AP) values under the precision-recall (PR) curves were all 1, indicating excellent discrimination across all categories. Overall, the CNN-Dinov2 model exhibited the best performance among the four models evaluated.Conclusion:This study successfully developed CNN-Dinov2, an automated classification model for Gram staining images. It offers valuable support for the rapid diagnosis of bloodstream infections caused by Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa, demonstrating practical utility.

Objective:To develop CNN-Dinov2, a deep learning-based automatic classification model for Gram-stained images, enabling rapid diagnosis of three prevalent Gram-negative bacilli in bloodstream infections: Escherichia coli ( E.coli), Klebsiella pneumoniae ( K.pneumoniae), and Pseudomonas aeruginosa ( P.aeruginosa). Methods:This evaluation study analyzed 1 425 Gram-stained microscopic images from patients with bloodstream infections at Houjie Hospital, in Dongguan City, collected between January 2023 and January 2024. The images, all positive for blood culture and identified as target strains, were categorized into Escherichia coli (419 images), Klebsiella pneumoniae (411 images), Pseudomonas aeruginosa (413 images), and other Gram-negative bacilli (182 images). They were randomly split into a training set (1 141 images), a validation set (141 images), and a test set (143 images) in an 8∶1∶1 ratio. A hybrid CNN-Dinov2 model was developed by integrating ResNet′s local feature extraction with Dinov2′s global pre-trained features, followed by a fully connected layer. The model was optimized by inputting the preprocessed images and adjusting parameters through loss calculation and backpropagation. AlexNet, Dinov2, and ResNet18 served as control models. The models′ classification performance was assessed using accuracy, precision, weighted F1 score, and recall rate, derived from the confusion matrix. The PR curve and AP value further evaluated each model′s classification capability across the four image categories. Results:The CNN-Dinov2 model achieved a training accuracy of 99.74%, a validation accuracy of 98.12%, and a validation loss of 0.070 6, demonstrating robust generalization without overfitting. Validation metrics revealed superior performance with an accuracy of 98.60%, precision of 98.65%, a weighted F1 score of 98.60%, and a recall rate of 98.60%, outperforming other models. The confusion matrix confirmed its strong classification capability, with the highest sum of diagonal values for identifying four types of bacteria. The macro average precision (AP) values under the precision-recall (PR) curves were all 1, indicating excellent discrimination across all categories. Overall, the CNN-Dinov2 model exhibited the best performance among the four models evaluated.Conclusion:This study successfully developed CNN-Dinov2, an automated classification model for Gram staining images. It offers valuable support for the rapid diagnosis of bloodstream infections caused by Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa, demonstrating practical utility.

To investigate the genomic features and perform cluster analysis of Carbapenem-resistant Klebsiella pneumoniae (CRKP) to provide an experimental basis for guiding the prevention and treatment of CRKP infections.A retrospective case-cohort study was conducted on 19 non-redundant CRKP strains isolated from the Tenth Affiliated Hospital of Southern Medical University between January and June 2023. Whole genome sequencing (WGS) and multilocus sequence typing (MLST) were performed to compare genomic features and analyze the resistance genes and homology of the strains.The results showed that the 19 CRKP strains were isolated from 8 different clinical departments, mainly from respiratory specimens. The whole genome sequencing revealed that the genomic lengths of CRKP ranged from 4.90 to 5.85 Mbp, with contigs N50 values>20 kb for each genome. The median overall GC content was 57.0% (50.4%-57.1%). Comparative genomic analysis identified three regions with high genomic variability. WGS detected 32 resistance genes across 11 categories. All 19 strains carried carbapenem resistance genes ( blaKPC-2 and blaOXA-48), blaTEM-1B extended-spectrum β-lactamase resistance genes, qnrS1 quinolone resistance gene, and fosA fosfomycin resistance gene, with each strain carrying only one carbapenemase gene. The detection rate of blaKPC-2 was 94.7% (18/19). MLST identified three sequence types: ST11, ST437 and ST147, with ST11 being predominant (89.5%, 17/19). Clustering analysis based on acquired resistance genes revealed three clonal transmission patterns among strains 72 and 90, and strains 88, 84, 66 and 79.In conclusion, CRKP strains carry multiple resistance genes, and clustering analysis indicating that nosocomial clonal transmission is closely related to acquired resistance genes. The ST11- blaKPC-2 type strain is the predominant clone. Strengthened surveillance and effective control strategies are necessary to reduce nosocomial transmission of CRKP.

To investigate the genomic features and perform cluster analysis of Carbapenem-resistant Klebsiella pneumoniae (CRKP) to provide an experimental basis for guiding the prevention and treatment of CRKP infections.A retrospective case-cohort study was conducted on 19 non-redundant CRKP strains isolated from the Tenth Affiliated Hospital of Southern Medical University between January and June 2023. Whole genome sequencing (WGS) and multilocus sequence typing (MLST) were performed to compare genomic features and analyze the resistance genes and homology of the strains.The results showed that the 19 CRKP strains were isolated from 8 different clinical departments, mainly from respiratory specimens. The whole genome sequencing revealed that the genomic lengths of CRKP ranged from 4.90 to 5.85 Mbp, with contigs N50 values>20 kb for each genome. The median overall GC content was 57.0% (50.4%-57.1%). Comparative genomic analysis identified three regions with high genomic variability. WGS detected 32 resistance genes across 11 categories. All 19 strains carried carbapenem resistance genes ( blaKPC-2 and blaOXA-48), blaTEM-1B extended-spectrum β-lactamase resistance genes, qnrS1 quinolone resistance gene, and fosA fosfomycin resistance gene, with each strain carrying only one carbapenemase gene. The detection rate of blaKPC-2 was 94.7% (18/19). MLST identified three sequence types: ST11, ST437 and ST147, with ST11 being predominant (89.5%, 17/19). Clustering analysis based on acquired resistance genes revealed three clonal transmission patterns among strains 72 and 90, and strains 88, 84, 66 and 79.In conclusion, CRKP strains carry multiple resistance genes, and clustering analysis indicating that nosocomial clonal transmission is closely related to acquired resistance genes. The ST11- blaKPC-2 type strain is the predominant clone. Strengthened surveillance and effective control strategies are necessary to reduce nosocomial transmission of CRKP.

Objective:Based on the principle that the aggregation-induced emission (AIE) fluorescent probe 6PD-DPAN could bind and aggregate with bacteria, and the fluorescence intensity could reflect the quantity of bacteria, a new method for rapid, convenient, and accurate bacterial drug sensitivity testing was established, which provided a basis for rapid and accurate clinical drug use.Methods:This was a methodological evaluation study. A total of 107 clinical isolates were collected from Houjie Hospital of Dongguan City from January to December 2022, among which 46 isolates were used for the establishment of the new method, and 61 isolates were used for methodological validation. The minimum inhibitory concentration (MIC) determined by broth microdilution method was used as the gold standard, and three antibacterial drugs, gentamicin, levofloxacin, and cefotaxime, were used as experimental drugs. The AIE plate was incubated for 4 hours, and the fluorescence intensity was measured every half an hour to draw a fluorescence change curve. The MIC results were compared with the CLSI breakpoints to determine the bacteria as sensitive, intermediate, or resistant. To simplify the detection process, the ratio of fluorescence intensity at 4 hours(R) was calculated, and the ROC curve was used to analyze the efficacy of R in determining bacterial growth and establish its cutoff value. The new method was used to determine the MIC of 61 clinical isolates, with broth microdilution method as the gold standard. The basic consistency, categorical consistency, very major errors, and major errors of the new method were analyzed, and the consistency between the two methods was determined by the Kappa test.Results:ROC curve analysis of the R after 4 hours of culture: The cut-off value was 3.0, with both sensitivity and specificity for determining bacterial growth being 100%. The median (interquartile) R for bacterial growth inhibition was 11.1 (8.6, 14.4); the median R-value for bacterial growth was 1.1 (1.0, 1.2). Compared to the gold standard, the newly established method showed 100% (61/61) essential agreement in detecting MICs of 61 clinical isolates, with a categorical agreement of 96.7% (59/61). There were no very major or major errors, and the Kappa value was 0.94, indicating good consistency between the newly established method and the microbroth dilution method.Conclusions:This study successfully established a new method for bacterial drug sensitivity testing based on AIE technology, which could obtain satisfactory results within 5 hours, providing a basis for early precision drug treatment in clinical practice.

Objective: To establish a method for simultaneous and rapid detecting of the polymorphisms in Cytochrome P450 2C9 (CYP2C9), CYP2C19, CYP4F2, Vitamin K epoxide reductase (VKORC1) and ATP-binding cassette subfamily B member1 (ABCB1) gene, which were associated with warfarin and clopidogrel, based on liquid phase chip technology. Methods: Method establishment. The eight gene sequences near targeted sites related to warfarin and clopidogrel were found in Genbank, and the specific primers and probes were designed. Through multiple PCR amplification, followed by allele specific primer extension (ASPE), and MagPlex-Tag microspheres hybridization, the suspension array Luminex 200 system step-by-step, the genotypes were determined by fluorescence signal. The reaction system was optimized and its methodological evaluation was performed. 260 patients with antithrombotic therapy from Dongguan houjie hospital were recruited in this study form June 2017 to December 2018. The eight genotypes of the 260 patients were detected by the established method, and the results were compared with the sequencing results. Results: The results of 260 samples showed that allelic median fluorescence intensity (MFI) ratios of homozygotes (mutant/wild-type) were all greater than 0.9 or less than 0.1, and all the allelic MFI ratios of heterozygotes were between 0.3 and 0.6. The within run and between run coefficients of variance for allelic MFI ratios were lower than 6.4% and 10.9%, respectively. The minimum DNA template requirements was 0.75ng. The genotypes of 260 patients determined by the established method were completely concordant with the sequencing results. Conclusion A method was established successfully for rapid detecting the genotypes which associated with warfarin and clopidogrel based on liquid phase chip technology.