<p>The objective of this study was to evaluate the performances of the automated digital imaging of Gram-stained slides against manual microscopy. Four hundred forty-three identified Gram-stained slides were included in this study. When both methods agreed, we considered the results as correct, and no further examination was carried out. Whenever the methods gave discrepant results, we reviewed the digital images and the glass slides by manual microscopy to avoid incorrectly read smears. The final result was a consensus of multiple independent reader interpretations. Among the 443 slides analyzed in this study, 101 (22.8%) showed discrepant results between the compared methods. The rates of discrepant results according to the specimen types were 5.7% (9/157) for positive blood cultures, 42% (60/142) for respiratory tract specimens, and 22% (32/144) for sterile site specimens. After a subsequent review of the discrepant slides, the final rate of discrepancies dropped to 7.0% (31/443). The overall agreement between the compared methods and the culture results reached 78% (345/443) and 79% (349/443) for manual microscopy and automated digital imaging, respectively. According to culture results, the specificity for automated digital imaging and manual microscopy were 90.8% and 87.7% respectively. In contrast, sensitivity was 84.1% for the two compared methods. The discrepant results were mostly encountered with microorganism morphologies of rare occurrence. The results reported in this study emphasize that on-screen reading is challenging, since the recognition of morphologies on-screen can appear different as compared to routine manual microscopy. Monitoring of Gram stain errors, which is facilitated by automated digital imaging, remains crucial for the quality control of reported Gram stain results.</p>

Digital object identifier (DOI): 10.1007/s10096-021-04233-2

Microscopic interpretation of stained smears is one of the most operator-dependent and time-intensive activities in the clinical microbiology laboratory. Here, we investigated application of an automated image acquisition and convolutional neural network (CNN)-based approach for automated Gram stain classification. Using an automated microscopy platform, uncoverslipped slides were scanned with a 40× dry objective, generating images of sufficient resolution for interpretation. We collected 25,488 images from positive blood culture Gram stains prepared during routine clinical workup. These images were used to generate 100,213 crops containing Gram-positive cocci in clusters, Gram-positive cocci in chains/pairs, Gram-negative rods, or background (no cells). These categories were targeted for proof-of-concept development as they are associated with the majority of bloodstream infections. Our CNN model achieved a classification accuracy of 94.9% on a test set of image crops. Receiver operating characteristic (ROC) curve analysis indicated a robust ability to differentiate between categories with an area under the curve of &gt;0.98 for each. After training and validation, we applied the classification algorithm to new images collected from 189 whole slides without human intervention. Sensitivity and specificity were 98.4% and 75.0% for Gram-positive cocci in chains and pairs, 93.2% and 97.2% for Gram-positive cocci in clusters, and 96.3% and 98.1% for Gram-negative rods. Taken together, our data support a proof of concept for a fully automated classification methodology for blood-culture Gram stains. Importantly, the algorithm was highly adept at identifying image crops with organisms and could be used to present prescreened, classified crops to technologists to accelerate smear review. This concept could potentially be extended to all Gram stain interpretive activities in the clinical laboratory.

Digital object identifier (DOI): 10.1128/JCM.01521-17