Fig. 3 Single-cell classification of parental and antibiotic-resistant cells using deep neural networks. (IMAGE)

The University of Osaka

Fig. 3 Single-cell classification of parental and antibiotic-resistant cells using deep neural networks.

Caption

A: The classification workflow. The insets show enlarged views of the contours of the cell region extracted from microscopy images of the parental and resistant (ENX) strains. The corresponding cell contours aligned horizontally and evenly interpolated to 128 points are shown as these formed the input data. A cartoon of the deep learning architecture shows a circular convolution (circ-conv) integrated into the ResNet architecture. For the circ-conv layer, r represents the radius or neighborhood size, resulting in a kernel size of 2r+1, followed by the number of channels and optionally a stride. The circ-conv layers are followed by batch normalization and rectified linear unit functions. The curved arrows depict a shortcut connection, and the dashed lines indicate the increase in the number of channels. The term fc stands for a fully-connected layer. Threefold cross-validation was conducted for evaluation of the classifier models.

B: The receiver operating characteristic (ROC) curve for the contour classification. Each curve shows an average of the ROC curves obtained from the threefold cross-validation. The vertical and horizontal axes represent the values of sensitivity and 1 − specificity, respectively.

C: Performance of the network in classifying resistant strains. The classification results (mean values) from the test sets in the threefold cross-validation are presented as bar graphs with standard deviations. The resistant strains are listed in descending order of the area under the curve (AUC). Sens. = sensitivity (correctly classified resistant cells); Spec. = specificity (correctly classified parental cells). Accu. = accuracy (collectly classified parental and resistant cells); AMK = Amikacin; AZM = Azithromycin; CFIX = Cefixime; CP = Chloramphenicol; CPFX = Ciprofloxacin; CPZ = Cefoperazone; DOXY = Doxycycline; ENX = Enoxacin; NM = Neomycin; TP = Trimethoprim; AR = aspect ratio; Circ = circularity; MaxFeret = maximum Feret’s diameter; MinFeret = minimum Feret’s diameter; Perim = perimeter; Round = roundness; Solid = solidity.

Credit

2024 Nishino et al., Bioinformatic analysis reveals the association between bacterial morphology and antibiotic resistance using light microscopy with deep learning., Frontiers in Microbiology

Usage Restrictions

Credit must be given to the creator.

License

CC BY

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.