[The model transferability of AI in digital pathology : Potential and reality]. / Die Modelltransferierbarkeit von KI in der digitalen Pathologie : Potenzial und Realität.

OBJECTIVE: Artificial intelligence (AI) holds the potential to make significant advancements in pathology. However, its actual implementation and certification for practical use are currently limited, often due to challenges related to model transferability. In this context, we investigate the factors influencing transferability and present methods aimed at enhancing the utilization of AI algorithms in pathology. MATERIALS AND METHODS: Various convolutional neural networks (CNNs) and vision transformers (ViTs) were trained using datasets from two institutions, along with the publicly available TCGA-MIBC dataset. These networks conducted predictions in urothelial tissue and intrahepatic cholangiocarcinoma (iCCA). The objective was to illustrate the impact of stain normalization, the influence of various artifacts during both training and testing, as well as the effects of the NoisyEnsemble method. RESULTS: We were able to demonstrate that stain normalization of slides from different institutions has a significant positive effect on the inter-institutional transferability of CNNs and ViTs (respectively +13% and +10%). In addition, ViTs usually achieve a higher accuracy in the external test (here +1.5%). Similarly, we showcased how artifacts in test data can negatively affect CNN predictions and how incorporating these artifacts during training leads to improvements. Lastly, NoisyEnsembles of CNNs (better than ViTs) were shown to enhance transferability across different tissues and research questions (+7% Bladder, +15% iCCA). DISCUSSION: It is crucial to be aware of the transferability challenge: achieving good performance during development does not necessarily translate to good performance in real-world applications. The inclusion of existing methods to enhance transferability, such as stain normalization and NoisyEnsemble, and their ongoing refinement, is of importance.

Food and Environmental Virology: Use of Passive Sampling to Characterize the Presence of SARS-CoV-2 and Other Viruses in Wastewater.

Fecal shedding of SARS-CoV-2 leads to a renaissance of wastewater-based epidemiology (WBE) as additional tool to follow epidemiological trends in the catchment of treatment plants. As alternative to the most commonly used composite samples in surveillance programs, passive sampling is increasingly studied. However, the many sorbent materials in different reports hamper the comparison of results and a standardization of the approach is necessary. Here, we compared different cost-effective sorption materials (cheesecloths, gauze swabs, electronegative filters, glass wool, and tampons) in torpedo-style housings with composite samples. Despite a remarkable variability of the concentration of SARS-CoV-2-specific gene copies, analysis of parallel-deposited passive samplers in the sewer demonstrated highest rate of positive samples and highest number of copies by using cheesecloths. Using this sorption material, monitoring of wastewater of three small catchments in the City of Dresden resulted in a rate of positive samples of 50% in comparison with composite samples (98%). During the investigation period, incidence of reported cases of SARS-CoV-2 in the catchments ranged between 16 and 170 per 100,000 persons and showed no correlation with the measured concentrations of E gene in wastewater. In contrast, constantly higher numbers of gene copies in passive vs. composite samples were found for human adenovirus and crAssphage indicating strong differences of efficacy of methods concerning the species investigated. Influenza virus A and B were sporadically detected allowing no comparison of results. The study contributes to the further understanding of possibilities and limits of passive sampling approaches in WBE.

Drug consumption in German cities and municipalities during the COVID-19 lockdown: a wastewater analysis.

Analysis of illicit drugs, medicines, and pathogens in wastewater is a powerful tool for epidemiological studies to monitor public health trends. The aims of this study were to (i) assess spatial and temporal trends of population-normalized mass loads of illicit drugs and nicotine in raw wastewater in the time of regulations against SARS-CoV-2 infections (2020-21) and (ii) find substances that are feasible markers for characterizing the occurrence of selected drugs in wastewater. Raw sewage 24-h composite samples were collected in catchment areas of 15 wastewater treatment plants (WWTPs) in urban, small-town, and rural areas in Germany during different lockdown phases from April 2020 to December 2021. Parent substances (amphetamine, methamphetamine, MDMA, carbamazepine, gabapentin, and metoprolol) and the metabolites of cocaine (benzoylecgonine) and nicotine (cotinine) were measured. The daily discharge of WWTP influents were used to calculate the daily load (mg/day) normalized by population equivalents (PE) in drained catchment areas (in mg/1,000 persons/day). A weekend trend for illicit drugs was visible with higher amounts on Saturdays and Sundays in larger WWTPs. An influence of the regulations to reduce SARS-CoV-2 infections such as contact bans and border closures on drug consumption has been proven in some cases and refuted in several. In addition, metoprolol and cotinine were found to be suitable as marker substances for the characterization of wastewater. A change in drug use was visible at the beginning of the SARS-CoV-2 crisis. Thereafter from mid-2020, no obvious effect was detected with regard to the regulations against SARS-CoV-2 infections on concentration of drugs in wastewater. Wastewater-based epidemiology is suitable for showing changes in drug consumption during the COVID-19 lockdown.

Regional and temporal differences in the relation between SARS-CoV-2 biomarkers in wastewater and estimated infection prevalence - Insights from long-term surveillance.

Wastewater-based epidemiology provides a conceptual framework for the evaluation of the prevalence of public health related biomarkers. In the context of the Coronavirus disease-2019, wastewater monitoring emerged as a complementary tool for epidemic management. In this study, we evaluated data from six wastewater treatment plants in the region of Saxony, Germany. The study period lasted from February to December 2021 and covered the third and fourth regional epidemic waves. We collected 1065 daily composite samples and analyzed SARS-CoV-2 RNA concentrations using reverse transcription-quantitative polymerase chain reaction (RT-qPCR). Regression models quantify the relation between RNA concentrations and disease prevalence. We demonstrated that the relation is site and time specific. Median loads per diagnosed case differed by a factor of 3-4 among sites during both waves and were on average 45 % higher during the third wave. In most cases, log-log-transformed data achieved better regression performance than non-transformed data and local calibration outperformed global models for all sites. The inclusion of lag/lead time, discharge and detection probability improved model performance in all cases significantly, but the importance of these components was also site and time specific. In all cases, models with lag/lead time and log-log-transformed data obtained satisfactory goodness-of-fit with adjusted coefficients of determination higher than 0.5. Back-estimation of testing efficiency from wastewater data confirmed state-wide prevalence estimation from individual testing statistics, but revealed pronounced differences throughout the epidemic waves and among the different sites.

Simultaneous Detection of SARS-CoV-2 and Influenza Virus in Wastewater of Two Cities in Southeastern Germany, January to May 2022.

Dependent on the excretion pattern, wastewater monitoring of viruses can be a valuable approach to characterizing their circulation in the human population. Using polyethylene glycol precipitation and reverse transcription-quantitative PCR, the occurrence of RNA of SARS-CoV-2 and influenza viruses A/B in the raw wastewater of two treatment plants in Germany between January and May 2022 was investigated. Due to the relatively high incidence in both exposal areas (plant 1 and plant 2), SARS-CoV-2-specific RNA was determined in all 273 composite samples analyzed (concentration of E gene: 1.3 × 104 to 3.2 × 106 gc/L). Despite a nation-wide low number of confirmed infections, influenza virus A was demonstrated in 5.2% (concentration: 9.8 × 102 to 8.4 × 104 gc/L; plant 1) and in 41.6% (3.6 × 103 to 3.0 × 105 gc/L; plant 2) of samples. Influenza virus B was detected in 36.0% (7.2 × 102 to 8.5 × 106 gc/L; plant 1) and 57.7% (9.6 × 103 to 2.1 × 107 gc/L; plant 2) of wastewater samples. The results of the study demonstrate the frequent detection of two primary respiratory viruses in wastewater and offer the possibility to track the epidemiology of influenza by wastewater-based monitoring.

How to learn with intentional mistakes: NoisyEnsembles to overcome poor tissue quality for deep learning in computational pathology.

There is a lot of recent interest in the field of computational pathology, as many algorithms are introduced to detect, for example, cancer lesions or molecular features. However, there is a large gap between artificial intelligence (AI) technology and practice, since only a small fraction of the applications is used in routine diagnostics. The main problems are the transferability of convolutional neural network (CNN) models to data from other sources and the identification of uncertain predictions. The role of tissue quality itself is also largely unknown. Here, we demonstrated that samples of the TCGA ovarian cancer (TCGA-OV) dataset from different tissue sources have different quality characteristics and that CNN performance is linked to this property. CNNs performed best on high-quality data. Quality control tools were partially able to identify low-quality tiles, but their use did not increase the performance of the trained CNNs. Furthermore, we trained NoisyEnsembles by introducing label noise during training. These NoisyEnsembles could improve CNN performance for low-quality, unknown datasets. Moreover, the performance increases as the ensemble become more consistent, suggesting that incorrect predictions could be discarded efficiently to avoid wrong diagnostic decisions.

Deep learning based on hematoxylin-eosin staining outperforms immunohistochemistry in predicting molecular subtypes of gastric adenocarcinoma.

In gastric cancer (GC), there are four molecular subclasses that indicate whether patients respond to chemotherapy or immunotherapy, according to the TCGA. In clinical practice, however, not every patient undergoes molecular testing. Many laboratories have used well-implemented in situ techniques (IHC and EBER-ISH) to determine the subclasses in their cohorts. Although multiple stains are used, we show that a staining approach is unable to correctly discriminate all subclasses. As an alternative, we trained an ensemble convolutional neuronal network using bagging that can predict the molecular subclass directly from hematoxylin-eosin histology. We also identified patients with predicted intra-tumoral heterogeneity or with features from multiple subclasses, which challenges the postulated TCGA-based decision tree for GC subtyping. In the future, deep learning may enable targeted testing for molecular subtypes and targeted therapy for a broader group of GC patients. © 2022 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.