Search

Abstract Objectives Visual inspection with acetic acid (VIA) is a low‐cost approach for cervical cancer screening used in most low‐ and middle‐income countries (LMICs) but, similar to other visual tests, is subjective and requires sustained training and quality assurance. We developed, trained, and validated an artificial‐intelligence‐based “Automated Visual Evaluation” (AVE) tool that can be adapted to run on smartphones to assess smartphone‐captured images of the cervix and identify precancerous lesions, helping augment VIA performance. Design Prospective study. Setting Eight public health facilities in Zambia. Participants A total of 8204 women aged 25–55. Interventions Cervical images captured on commonly used low‐cost smartphone models were matched with key clinical information including human immunodeficiency virus (HIV) and human papillomavirus (HPV) status, plus histopathology analysis (where applicable), to develop and train an AVE algorithm and evaluate its performance for use as a primary screen and triage test for women who are HPV positive. Main Outcome Measures Area under the receiver operating curve (AUC); sensitivity; specificity. Results As a general population screening tool for cervical precancerous lesions, AVE identified cases of cervical precancerous and cancerous (CIN2+) lesions with high performance (AUC = 0.91, 95% confidence interval [CI] = 0.89–0.93), which translates to a sensitivity of 85% (95% CI = 81%–90%) and specificity of 86% (95% CI = 84%–88%) based on maximizing the Youden's index. This represents a considerable improvement over naked eye VIA, which as per a meta‐analysis by the World Health Organization (WHO) has a sensitivity of 66% and specificity of 87%. For women living with HIV, the AUC of AVE was 0.91 (95% CI = 0.88–0.93), and among those testing positive for high‐risk HPV types, the AUC was 0.87 (95% CI = 0.83–0.91). Conclusions These results demonstrate the feasibility of utilizing AVE on images captured using a commonly available smartphone by nurses in a screening program, and support our ongoing efforts for moving to more broadly evaluate AVE for its clinical sensitivity, specificity, feasibility, and acceptability across a wider range of settings. Limitations of this study include potential inflation of performance estimates due to verification bias (as biopsies were only obtained from participants with visible aceto‐white cervical lesions) and due to this being an internal validation (the test data, while independent from that used to develop the algorithm was drawn from the same study).

Abstract The endoscopic features associated with eosinophilic esophagitis (EoE) may be missed during routine endoscopy. We aimed to develop and evaluate an Artificial Intelligence (AI) algorithm for detecting and quantifying the endoscopic features of EoE in white light images, supplemented by the EoE Endoscopic Reference Score (EREFS). An AI algorithm (AI-EoE) was constructed and trained to differentiate between EoE and normal esophagus using endoscopic white light images extracted from the database of the University Hospital Augsburg. In addition to binary classification, a second algorithm was trained with specific auxiliary branches for each EREFS feature (AI-EoE-EREFS). The AI algorithms were evaluated on an external data set from the University of North Carolina, Chapel Hill (UNC), and compared with the performance of human endoscopists with varying levels of experience. The overall sensitivity, specificity, and accuracy of AI-EoE were 0.93 for all measures, while the AUC was 0.986. With additional auxiliary branches for the EREFS categories, the AI algorithm (AI-EoE-EREFS) performance improved to 0.96, 0.94, 0.95, and 0.992 for sensitivity, specificity, accuracy, and AUC, respectively. AI-EoE and AI-EoE-EREFS performed significantly better than endoscopy beginners and senior fellows on the same set of images. An AI algorithm can be trained to detect and quantify endoscopic features of EoE with excellent performance scores. The addition of the EREFS criteria improved the performance of the AI algorithm, which performed significantly better than endoscopists with a lower or medium experience level.

Abstract WGS is used to define if isolates are “in” or “out” of an outbreak and/or microbial root cause investigation. No threshold of genetic differences is fixed and the conclusions on similarity between isolates are mainly based on the knowledge generated from previous outbreak investigations and reported mutation rates. Mutation rates in Salmonella when exposed to food processing conditions are lacking. Thus, in this study, the ability of heat and dry stress to cause genetic changes in two Salmonella serotypes frequently isolated from low moisture foods was investigated. S. enterica serovars S. Agona ATCC 51,957 and S. Mbandaka NCTC 7892 (ATCC 51,958) were repeatedly exposed to heat (90 °C for 5 min) in a low water activity and high fat matrix. No increased fitness of the strains was observed after 10 repeated heat treatments. However, genetic changes were introduced and the number of genetic differences increased with every heat treatment cycle. The genetic changes appeared randomly in the genome and were responsible for a population of diverse isolates with 0 to 28 allelic differences (0 to 38 SNPs) between them. This knowledge is key to interpret WGS results for source tracking investigations as part of a root cause analysis in a contamination event as isolates are exposed to stress conditions.

Salmonella is one of the most common causes of food-borne diseases worldwide. While Salmonella molecular subtyping by Whole Genome Sequencing (WGS) is increasingly used for outbreak and source tracking investigations, serotyping remains as a first-line characterization of Salmonella isolates. The traditional phenotypic method for serotyping is logistically challenging, as it requires the use of more than 150 specific antisera and well trained personnel to interpret the results. Consequently, it is not a routine method for the majority of laboratories. Several rapid molecular methods targeting O and H loci or surrogate genomic markers have been developed as alternative solutions. With the expansion of WGS, in silico Salmonella serotype prediction using WGS data is available. Here, we compared a microarray method using molecular markers, the Check and Trace Salmonella assay (CTS) and a WGS-based serotype prediction tool that targets molecular determinants of serotype (SeqSero) to the traditional phenotypic method using 100 strains representing 45 common and uncommon serotypes. Compared to the traditional method, the CTS assay correctly serotyped 97% of the strains, four strains gave a double serotype prediction. Among the inconclusive data, one strain was not predicted and two strains were incorrectly identified. SeqSero was evaluated with two versions (SeqSero 1 and the alpha test version of SeqSero 2). The correct antigenic formula was predicted by SeqSero 1 for 96 and 95% of strains using raw reads and assembly, respectively. However, 34 and 33% of these predictions included multiple serotypes by raw reads and assembly. With raw reads, one strain was not identified and three strains were discordant with phenotypic serotyping result. With assembly, three strains were not predicted and two strains were incorrectly predicted. While still under development, SeqSero 2 maintained the accuracy of antigenic formula prediction at 98% and reduced multiple serotype prediction rate to 13%. One strain had no prediction and one strain was incorrectly predicted. Our study indicates that the CTS assay is a good alternative for routine laboratories as it is an easy to use method with a short turn-around-time. SeqSero is a reliable replacement for phenotypic serotyping if WGS is routinely implemented.

Whole genome sequencing (WGS) has arisen as a powerful tool to perform pathogen source tracking in the food industry thanks to several developments in recent years. However, the cost associated to this technology and the degree of expertise required to accurately process and understand the data has limited its adoption at a wider scale. Additionally, the time needed to obtain actionable information is often seen as an impairment for the application and use of the information generated via WGS. Ongoing work towards standardization of wet lab including sequencing protocols, following guidelines from the regulatory authorities and international standardization efforts make the technology more and more accessible. However, data analysis and results interpretation guidelines are still subject to initiatives coming from distinct groups and institutions. There are multiple bioinformatics software and pipelines developed to handle such information. Nevertheless, little consensus exists on a standard way to process the data and interpret the results. Here, we want to present the constraints we face in an industrial setting and the steps we consider necessary to obtain high quality data, reproducible results and a robust interpretation of the obtained information. All of this, in a time frame allowing for data-driven actions supporting factories and their needs.

Whole genome sequencing (WGS), using high throughput sequencing technology, reveals the complete sequence of the bacterial genome in a few days. WGS is increasingly being used for source tracking, pathogen surveillance and outbreak investigation due to its high discriminatory power. In the food industry, WGS used for source tracking is beneficial to support contamination investigations. Despite its increased use, no standards or guidelines are available today for the use of WGS in outbreak and/or trace-back investigations. Here we present a validation of our complete (end-to-end) WGS workflow for Listeria monocytogenes and Salmonella enterica including: subculture of isolates, DNA extraction, sequencing and bioinformatics analysis. This end-to-end WGS workflow was evaluated according to the following performance criteria: stability, repeatability, reproducibility, discriminatory power, and epidemiological concordance. The current study showed that few single nucleotide polymorphism (SNPs) were observed for L. monocytogenes and S. enterica when comparing genome sequences from five independent colonies from the first subculture and five independent colonies after the tenth subculture. Consequently, the stability of the WGS workflow for L. monocytogenes and S. enterica was demonstrated despite the few genomic variations that can occur during subculturing steps. Repeatability and reproducibility were also demonstrated. The WGS workflow was shown to have a high discriminatory power and has the ability to show genetic relatedness. Additionally, the WGS workflow was able to reproduce published outbreak investigation results, illustrating its capability of showing epidemiological concordance. The current study proposes a validation approach comprising all steps of a WGS workflow and demonstrates that the workflow can be applied to L. monocytogenes or S. enterica.