Deep learning based tomosynthesis denoising: a bias investigation across different breast types.

Purpose: High noise levels due to low X-ray dose are a challenge in digital breast tomosynthesis (DBT) reconstruction. Deep learning algorithms show promise in reducing this noise. However, these algorithms can be complex and biased toward certain patient groups if the training data are not representative. It is important to thoroughly evaluate deep learning-based denoising algorithms before they are applied in the medical field to ensure their effectiveness and fairness. In this work, we present a deep learning-based denoising algorithm and examine potential biases with respect to breast density, thickness, and noise level. Approach: We use physics-driven data augmentation to generate low-dose images from full field digital mammography and train an encoder-decoder network. The rectified linear unit (ReLU)-loss, specifically designed for mammographic denoising, is utilized as the objective function. To evaluate our algorithm for potential biases, we tested it on both clinical and simulated data generated with the virtual imaging clinical trial for regulatory evaluation pipeline. Simulated data allowed us to generate X-ray dose distributions not present in clinical data, enabling us to separate the influence of breast types and X-ray dose on the denoising performance. Results: Our results show that the denoising performance is proportional to the noise level. We found a bias toward certain breast groups on simulated data; however, on clinical data, our algorithm denoises different breast types equally well with respect to structural similarity index. Conclusions: We propose a robust deep learning-based denoising algorithm that reduces DBT projection noise levels and subject it to an extensive test that provides information about its strengths and weaknesses.

Conventional versus video-assisted laryngoscopy for perioperative endotracheal intubation (COVALENT) - a randomized, controlled multicenter trial.

BACKGROUND: Data on the routine use of video-assisted laryngoscopy in peri-operative intubations are rather inconsistent and ambiguous, in part due to small populations and non-uniform outcome measures in past trials. Failed or prolonged intubation procedures are a reason for relevant morbidity and mortality. This study aims to determine whether video-assisted laryngoscopy (with both Macintosh-shaped and hyperangulated blades) is at least equal to the standard method of direct laryngoscopy with respect to the first-pass success rate. Furthermore, validated tools from the field of human factors will be applied to examine within-team communication and task load during this critical medical procedure. METHODS: In this randomized, controlled, three-armed parallel group design, multi-centre trial, a total of more than 2500 adult patients scheduled for perioperative endotracheal intubation will be randomized. In equally large arms, video-assisted laryngoscopy with a Macintosh-shaped or a hyperangulated blade will be compared to the standard of care (direct laryngoscopy with Macintosh blade). In a pre-defined hierarchical analysis, we will test the primary outcome for non-inferiority first. If this goal should be met, the design and projected statistical power also allow for subsequent testing for superiority of one of the interventions. Various secondary outcomes will account for patient safety considerations as well as human factors interactions within the provider team and will allow for further exploratory data analysis and hypothesis generation. DISCUSSION: This randomized controlled trial will provide a solid base of data in a field where reliable evidence is of major clinical importance. With thousands of endotracheal intubations performed every day in operating rooms around the world, every bit of performance improvement translates into increased patient safety and comfort and may eventually prevent significant burden of disease. Therefore, we feel confident that a large trial has the potential to considerably benefit patients and anaesthetists alike. TRIAL REGISTRATION: ClincalTrials.gov NCT05228288. PROTOCOL VERSION: 1.1, November 15, 2021.

Contaminant concentration versus flow velocity: drivers of biodegradation and microbial growth in groundwater model systems.

Aromatic hydrocarbons belong to the most abundant contaminants in groundwater systems. They can serve as carbon and energy source for a multitude of indigenous microorganisms. Predictions of contaminant biodegradation and microbial growth in contaminated aquifers are often vague because the parameters of microbial activity in the mathematical models used for predictions are typically derived from batch experiments, which don't represent conditions in the field. In order to improve our understanding of key drivers of natural attenuation and the accuracy of predictive models, we conducted comparative experiments in batch and sediment flow-through systems with varying concentrations of contaminant in the inflow and flow velocities applying the aerobic Pseudomonas putida strain F1 and the denitrifying Aromatoleum aromaticum strain EbN1. We followed toluene degradation and bacterial growth by measuring toluene and oxygen concentrations and by direct cell counts. In the sediment columns, the total amount of toluene degraded by P. putida F1 increased with increasing source concentration and flow velocity, while toluene removal efficiency gradually decreased. Results point at mass transfer limitation being an important process controlling toluene biodegradation that cannot be assessed with batch experiments. We also observed a decrease in the maximum specific growth rate with increasing source concentration and flow velocity. At low toluene concentrations, the efficiencies in carbon assimilation within the flow-through systems exceeded those in the batch systems. In all column experiments the number of attached cells plateaued after an initial growth phase indicating a specific "carrying capacity" depending on contaminant concentration and flow velocity. Moreover, in all cases, cells attached to the sediment dominated over those in suspension, and toluene degradation was performed practically by attached cells only. The observed effects of varying contaminant inflow concentration and flow velocity on biodegradation could be captured by a reactive-transport model. By monitoring both attached and suspended cells we could quantify the release of new-grown cells from the sediments to the mobile aqueous phase. Studying flow velocity and contaminant concentrations as key drivers of contaminant transformation in sediment flow-through microcosms improves our system understanding and eventually the prediction of microbial biodegradation at contaminated sites.

Compound-Specific Stable Isotope Fractionation of Pesticides and Pharmaceuticals in a Mesoscale Aquifer Model.

Compound-specific isotope analysis (CSIA) receives increasing interest for its ability to detect natural degradation of pesticides and pharmaceuticals. Despite recent laboratory studies, CSIA investigations of such micropollutants in the environment are still rare. To explore the certainty of information obtainable by CSIA in a near-environmental setting, a pulse of the pesticide bentazone, the pesticide metabolite 2,6-dichlorobenzamide (BAM), and the pharmaceuticals diclofenac and ibuprofen was released into a mesoscale aquifer with quasi-two-dimensional flow. Concentration breakthrough curves (BTC) of BAM and ibuprofen demonstrated neither degradation nor sorption. Bentazone was transformed but did not sorb significantly, whereas diclofenac showed both degradation and sorption. Carbon and nitrogen CSIA could be accomplished in similar concentrations as for "traditional" priority pollutants (low µg/L range), however, at the cost of uncertainties (0.4-0.5‰ (carbon), 1‰ (nitrogen)). Nonetheless, invariant carbon and nitrogen isotope values confirmed that BAM was neither degraded nor sorbed, while significant enrichment of (13)C and in particular (15)N corroborated transformation of diclofenac and bentazone. Retardation of diclofenac was reflected in additional (15)N sorption isotope effects, whereas isotope fractionation of transverse dispersion could not be identified. These results provide a benchmark on the performance of CSIA to monitor the reactivity of micropollutants in aquifers and may guide future efforts to accomplish CSIA at even lower concentrations (ng/L range).

Dynamics of suspended and attached aerobic toluene degraders in small-scale flow-through sediment systems under growth and starvation conditions.

The microbially mediated reactions, that are responsible for field-scale natural attenuation of organic pollutants, are governed by the concurrent presence of a degrading microbial community, suitable energy and carbon sources, electron acceptors, as well as nutrients. The temporal lack of one of these essential components for microbial activity, arising from transient environmental conditions, might potentially impair in situ biodegradation. This study presents results of small scale flow-through experiments aimed at ascertaining the effects of substrate-starvation periods on the aerobic degradation of toluene by Pseudomonas putida F1. During the course of the experiments, concentrations of attached and mobile bacteria, as well as toluene and oxygen were monitored. Results from a fitted reactive-transport model, along with the observed profiles, show the ability of attached cells to survive substrate-starvation periods of up to four months and suggest a highly dynamic exchange between attached and mobile cells under growth conditions and negligible cell detachment under substrate-starvation conditions. Upon reinstatement of toluene, it was readily degraded without a significant lag period, even after a starvation period of 130 days. Our experimental and modeling results strongly suggest that aerobic biodegradation of BTEX-hydrocarbons at contaminated field sites is not hampered by intermittent starvation periods of up to four months.

Fringe-controlled biodegradation under dynamic conditions: quasi 2-D flow-through experiments and reactive-transport modeling.

Biodegradation in contaminated aquifers has been shown to be most pronounced at the fringe of contaminant plumes, where mixing of contaminated water and ambient groundwater, containing dissolved electron acceptors, stimulates microbial activity. While physical mixing of contaminant and electron acceptor by transverse dispersion has been shown to be the major bottleneck for biodegradation in steady-state plumes, so far little is known on the effect of flow and transport dynamics (caused, e.g., by a seasonally fluctuating groundwater table) on biodegradation in these systems. Towards this end we performed experiments in quasi-two-dimensional flow-through microcosms on aerobic toluene degradation by Pseudomonas putida F1. Plume dynamics were simulated by vertical alteration of the toluene plume position and experimental results were analyzed by reactive-transport modeling. We found that, even after disappearance of the toluene plume for two weeks, the majority of microorganisms stayed attached to the sediment and regained their full biodegradation potential within two days after reappearance of the toluene plume. Our results underline that besides microbial growth, also maintenance and dormancy are important processes that affect biodegradation performance under transient environmental conditions and therefore deserve increased consideration in future reactive-transport modeling.

Direct experimental evidence of non-first order degradation kinetics and sorption-induced isotopic fractionation in a mesoscale aquifer: 13C/12C analysis of a transient toluene pulse.

The injection of a mixed toluene and D2O (conservative tracer) pulse into a pristine mesoscale aquifer enabled a first direct experimental comparison of contaminant-specific isotopic fractionation from sorption versus biodegradation and transverse dispersion on a relevant scale. Water samples were taken from two vertically resolved sampling ports at 4.2 m distance. Analysis of deuterium and toluene concentrations allowed quantifying the extent of sorption (R = 1.25) and biodegradation (37% and 44% of initial toluene at the two sampling ports). Sorption and biodegradation were found to directly affect toluene (13)C/(12)C breakthrough curves. In particular, isotope trends demonstrated that biodegradation underwent Michaelis-Menten kinetics rather than first-order kinetics. Carbon isotope enrichment factors obtained from an optimized reactive transport model (Eckert et al., this issue) including a possible isotope fractionation of transverse dispersion were ε(equ)(sorption) = -0.31 ‰, ε(kin)(transverse-dispersion) = -0.82 ‰, and ε(kin)(biodegradation) = -2.15 ‰. Extrapolation of our results to the scenario of a continuous injection predicted that (i) the bias in isotope fractionation from sorption, but not transverse dispersion, may be avoided when the plume reaches steady-state; and (ii) the relevance from both processes is expected to decrease at longer flow distances when isotope fractionation of degradation increasingly dominates.

Model complexity needed for quantitative analysis of high resolution isotope and concentration data from a toluene-pulse experiment.

Separating microbial- and physical-induced effects on the isotope signals of contaminants has been identified as a challenge in interpreting compound-specific isotope data. In contrast to simple analytical tools, such as the Rayleigh equation, reactive-transport models can account for complex interactions of different fractionating processes. The question arises how complex such models must be to reproduce the data while the model parameters remain identifiable. In this study, we reanalyze the high-resolution data set of toluene concentration and toluene-specific δ(13)C from the toluene-pulse experiment performed by Qiu et al. (this issue). We apply five reactive-transport models, differing in their degree of complexity. We uniquely quantify degradation and sorption properties of the system for each model, estimate the contributions of biodegradation-induced, sorption-induced, and transverse-dispersion-induced isotope fractionation to the overall isotope signal, and investigate the error introduced in the interpretation of the data when individual processes are neglected. Our results show that highly resolved data of both concentration and isotope ratios are needed for unique process identification facilitating reliable model calibration. Combined analysis of these highly resolved data demands reactive transport models accounting for nonlinear degradation kinetics and isotope fractionation by both reactive and physical processes such as sorption and transverse dispersion.

Numerical simulation of isotope fractionation in steady-state bioreactive transport controlled by transverse mixing.

Compound-specific stable isotope analysis (CSIA) has increasingly been used as a tool to assess intrinsic biodegradation at contaminated field sites. Typically, the Rayleigh equation is used to estimate the extent of biodegradation from measured isotope ratios of the contaminant. However, if the rate-limiting step in overall degradation is not the microbial reaction itself, the Rayleigh equation may no more be applicable. In this study we simulate biodegradation of continuously emitted petroleum hydrocarbons in groundwater systems. These contaminants are effectively degraded at the plume fringe where transverse dispersion makes them mix with dissolved electron acceptors present in the ambient groundwater. We simulate reactive transport to study the coupled effects of transverse mixing and biodegradation on the spatial patterns of carbon isotope signatures and their interpretation based on depth-integrated sampling which represents the most common setup in the assessment of contaminated sites. We present scenarios mimicking a hydraulically uniform laboratory experiment and a field-scale application considering heterogeneous conductivity fields. We compare cases in which isotopologue-specific transverse dispersion is considered to those where this additional fractionation process is neglected. We show that these effects cause significant shifts in the isotopic signals and may lead to overestimation of biodegradation. Moreover, our results provide evidence that the rate-limiting effect of transverse mixing on the overall degradation spatially varies along the length of a steady-state contaminant plume. The control of biodegradation by transverse dispersion and the fractionating effect of dispersion modulate the fractionation caused by the microbial reaction alone. As a consequence, significantly nonlinear isotopic patterns are observed in a Rayleigh plot. Simulations in heterogeneous flow domains show that these effects persist at larger field scales and are sensitive to the degree of mixing enhancement, determined by the heterogeneity of the hydraulic conductivity fields, and to the groundwater flow velocity.