INTRODUCTION: Methicillin-Resistant Staphylococcus aureus (MRSA) presenting reduced susceptibility to vancomycin has been associated to therapeutic failure. Some methods used by clinical laboratories may not be sufficiently accurate to detect this phenotype, compromising results and the outcome of the patient. OBJECTIVES: To evaluate the performance of methods in the detection of vancomycin MIC values among clinical isolates of MRSA. MATERIAL AND METHODS: The Vancomycin Minimal Inhibitory Concentration was determined for 75 MRSA isolates from inpatients of Mãe de Deus Hospital, Porto Alegre, Brazil. The broth microdilution (BM) was used as the gold-standard technique, as well as the following methods: E-test® strips (BioMérieux), M.I.C.E® strips (Oxoid), PROBAC® commercial panel and the automated system MicroScan® (Siemens). Besides, the agar screening test was carried out with 3 µg/mL of vancomycin. RESULTS: All isolates presented MIC ≤ 2 µg/mL for BM. E-test® had higher concordance (40%) in terms of global agreement with the gold standard, and there was not statistical difference among E-test® and broth microdilution results. PROBAC® panels presented MICs, in general, lower than the gold-standard panels (58.66% major errors), while M.I.C.E.® MICs were higher (67.99% minor errors). CONCLUSIONS: For the population of MRSA in question, E-test® presented the best performance, although with a heterogeneous accuracy, depending on MIC values.

The objective of this work was to evaluate the rate of meiosis resumption and nuclear maturation of rat (Rattus norvegicus) oocytes selected for in vitro maturation (IVM) after staining of cumulus-oocyte complexes (COCs) with blue cresyl brilliant (BCB) using different protocols: exposure for 30, 60 or 90 min at 26 μM BCB (Experiment 1), and exposure for 60 min at 13, 20 or 26 μM BCB (Experiment 2). In Experiment 1, the selection of oocytes exposed to BCB for 60 min was found to be the most suitable, as meiosis resumption rates in the BCB(+) group (n = 35/61; 57.37%) were the closest to the observed in the control (not exposed) group (n = 70/90; 77.77%) and statistically higher than the values observed for the BCB(-) group (n = 3/41; 7.32%). Additionally, the more effective evaluation of diagnostic tests (sensitivity and negative predictive value 100%) was observed in COCs exposed for 60 min. In Experiment 2, the 13 μM BCB(+) group presented rates of meiosis resumption (n = 57/72; 72.22%) similar to the control group (n = 87/105; 82.86%) and higher than other concentration groups. However, this results of the analysis between BCB(-) oocytes was also higher in the 13 μM BCB group (n = 28/91; 30.78%) when compared with BCB(-) COCs exposed to 20 μM (n = 3/62; 4.84%) or 26 μM (n = 3/61; 4.92%) BCB. The nuclear maturation rate in the 13 μM BCB group was similar between BCB(+) or BCB(-) oocytes. The 20 μM BCB group had a lower rate of nuclear maturation of BCB(-) oocytes than other groups. Thus, our best results in the selection of Rattus norvegicus oocytes by staining with BCB were obtained using the concentration of 13 μM and 20 μM, and an incubation period of 60 min.

No abstract

No abstract

optical signature of each type of microscope, and the presence of foreign objects, such as a holding micropipette present in the image. If a certain optical signature or foreign artifact appeared more in the positive training class compared to the negative training class, the CNN models were found to learn these biases and give a higher score to those images regardless of the embryo morphology. With these insights, a new dataset was prepared that balanced the ratios of positive-to-negative samples for each type of microscope and for each group containing foreign objects. This provided a non-inflated total AUC of 0.72 and significantly raised the lowest per-site AUC from 0.51 to 0.61.CONCLUSIONS: There has been significant recent interest in using deep learning for analyzing images of embryos at the blastocyst stage. The blackbox nature of deep learning models such as CNNs makes it difficult to recognize when potential sources of bias have been introduced during the training process. We performed a series of experiments that identified and reduced two sources of bias, and improved per-clinic performance of the CNN. Future work will continue to search for other sources of bias and address them accordingly.IMPACT STATEMENT: Naive approaches to preparing training data for deep learning models for embryo ranking can create bias in the models. Our work illustrates the need for careful preparation of training data and monitoring of different metrics to identify and reduce potential sources of bias.