Deep Convolutional Neural Networks Implementation for the Analysis of Urine Culture.

BACKGROUND: Urine culture images collected using bacteriology automation are currently interpreted by technologists during routine standard-of-care workflows. Machine learning may be able to improve the harmonization of and assist with these interpretations. METHODS: A deep learning model, BacterioSight, was developed, trained, and tested on standard BD-Kiestra images of routine blood agar urine cultures from 2 different medical centers. RESULTS: BacterioSight displayed performance on par with standard-of-care-trained technologist interpretations. BacterioSight accuracy ranged from 97% when compared to standard-of-care (single technologist) and reached 100% when compared to a consensus reached by a group of technologists (gold standard in this study). Variability in image interpretation by trained technologists was identified and annotation "fuzziness" was quantified and found to correlate with reduced confidence in BacterioSight interpretation. Intra-testing (training and testing performed within the same institution) performed well giving Area Under the Curve (AUC) ≥0.98 for negative and positive plates, whereas, cross-testing on images (trained on one institution's images and tested on images from another institution) showed decreased performance with AUC ≥0.90 for negative and positive plates. CONCLUSIONS: Our study provides a roadmap on how BacterioSight or similar deep learning prototypes may be implemented to screen for microbial growth, flag difficult cases for multi-personnel review, or auto-verify a subset of cultures with high confidence. In addition, our results highlight image interpretation variability by trained technologist within an institution and globally across institutions. We propose a model in which deep learning can enhance patient care by identifying inherent sample annotation variability and improving personnel training.

Specificity of SARS-CoV-2 Real-Time PCR Improved by Deep Learning Analysis.

Real-time PCR (RT-PCR) is widely used to diagnose human pathogens. RT-PCR data are traditionally analyzed by estimating the threshold cycle (CT ) at which the fluorescence signal produced by emission of a probe crosses a baseline level. Current models used to estimate the CT value are based on approximations that do not adequately account for the stochastic variations of the fluorescence signal that is detected during RT-PCR. Less common deviations become more apparent as the sample size increases, as is the case in the current SARS-CoV-2 pandemic. In this work, we employ a method independent of CT value to interpret RT-PCR data. In this novel approach, we built and trained a deep learning model, qPCRdeepNet, to analyze the fluorescent readings obtained during RT-PCR. We describe how this model can be deployed as a quality assurance tool to monitor result interpretation in real time. The model's performance with the TaqPath COVID19 Combo Kit assay, widely used for SARS-CoV-2 detection, is described. This model can be applied broadly for the primary interpretation of RT-PCR assays and potentially replace the CT interpretive paradigm.

Characterization of minority HIV-1 drug resistant variants in the United Kingdom following the verification of a deep sequencing-based HIV-1 genotyping and tropism assay.

BACKGROUND: The widespread global access to antiretroviral drugs has led to considerable reductions in morbidity and mortality but, unfortunately, the risk of virologic failure increases with the emergence, and potential transmission, of drug resistant viruses. Detecting and quantifying HIV-1 drug resistance has therefore become the standard of care when designing new antiretroviral regimens. The sensitivity of Sanger sequencing-based HIV-1 genotypic assays is limited by its inability to identify minority members of the quasispecies, i.e., it only detects variants present above ~ 20% of the viral population, thus, failing to detect minority variants below this threshold. It is clear that deep sequencing-based HIV-1 genotyping assays are an important step change towards accurately monitoring HIV-infected individuals. METHODS: We implemented and verified a clinically validated HIV-1 genotyping assay based on deep sequencing (DEEPGEN™) in two clinical laboratories in the United Kingdom: St. George's University Hospitals Healthcare NHS Foundation Trust (London) and at NHS Lothian (Edinburgh), to characterize minority HIV-1 variants in 109 plasma samples from ART-naïve or -experienced individuals. RESULTS: Although subtype B HIV-1 strains were highly prevalent (44%, 48/109), most individuals were infected with non-B subtype viruses (i.e., A1, A2, C, D, F1, G, CRF02_AG, and CRF01_AE). DEEPGEN™ was able to accurately detect drug resistance-associated mutations not identified using standard Sanger sequencing-based tests, which correlated significantly with patient's antiretroviral treatment histories. A higher proportion of minority PI-, NRTI-, and NNRTI-resistance mutations was detected in NHS Lothian patients compared to individuals from St. George's, mainly M46I/L and I50 V (associated with PIs), D67 N, K65R, L74I, M184 V/I, and K219Q (NRTIs), and L100I (NNRTIs). Interestingly, we observed an inverse correlation between intra-patient HIV-1 diversity and CD4+ T cell counts in the NHS Lothian patients. CONCLUSIONS: This is the first study evaluating the transition, training, and implementation of DEEPGEN™ between three clinical laboratories in two different countries. More importantly, we were able to characterize the HIV-1 drug resistance profile (including minority variants), coreceptor tropism, subtyping, and intra-patient viral diversity in patients from the United Kingdom, providing a rigorous foundation for basing clinical decisions on highly sensitive and cost-effective deep sequencing-based HIV-1 genotyping assays in the country.

Impaired human immunodeficiency virus type 1 replicative fitness in atypical viremic non-progressor individuals.

BACKGROUND: Progression rates from initial HIV-1 infection to advanced AIDS vary significantly among infected individuals. A distinct subgroup of HIV-1-infected individuals-termed viremic non-progressors (VNP) or controllers-do not seem to progress to AIDS, maintaining high CD4+ T cell counts despite high levels of viremia for many years. Several studies have evaluated multiple host factors, including immune activation, trying to elucidate the atypical HIV-1 disease progression in these patients; however, limited work has been done to characterize viral factors in viremic controllers. METHODS: We analyzed HIV-1 isolates from three VNP individuals and compared the replicative fitness, near full-length HIV-1 genomes and intra-patient HIV-1 genetic diversity with viruses from three typical (TP) and one rapid (RP) progressor individuals. RESULTS: Viremic non-progressors and typical patients were infected for >10 years (range 10-17 years), with a mean CD4+ T-cell count of 472 cells/mm3 (442-529) and 400 cells/mm3 (126-789), respectively. VNP individuals had a less marked decline in CD4+ cells (mean -0.56, range -0.4 to -0.7 CD4+/month) than TP patients (mean -10.3, -8.2 to -13.1 CD4+/month). Interestingly, VNP individuals carried viruses with impaired replicative fitness, compared to HIV-1 isolates from the TP and RP patients (p < 0.05, 95% CI). Although analyses of the near full-length HIV-1 genomes showed no clear patterns of single-nucleotide polymorphisms (SNP) that could explain the decrease in replicative fitness, both the number of SNPs and HIV-1 population diversity correlated inversely with the replication capacity of the viruses (r = -0.956 and r = -0.878, p < 0.01, respectively). CONCLUSION: It is likely that complex multifactorial parameters govern HIV-1 disease progression in each individual, starting with the infecting virus (phenotype, load, and quasispecies diversity) and the intrinsic ability of the host to respond to the infection. Here we analyzed a subset of viremic controller patients and demonstrated that similar to the phenomenon observed in patients with a discordant response to antiretroviral therapy (i.e., high CD4+ cell counts with detectable plasma HIV-1 RNA load), reduced viral replicative fitness seems to be linked to slow disease progression in these antiretroviral-naïve individuals.