Evaluation of data uncertainty for deep-learning-based CT noise reduction using ensemble patient data and a virtual imaging trial framework

Deep learning-based image reconstruction and noise reduction (DLIR) methods have been increasingly deployed in clinical CT. Accurate assessment of their data uncertainty properties is essential to understand the stability of DLIR in response to noise. In this work, we aim to evaluate the data uncertainty of a DLIR method using real patient data and a virtual imaging trial framework and compare it with filtered-backprojection (FBP) and iterative reconstruction (IR). The ensemble of noise realizations was generated by using a realistic projection domain noise insertion technique. The impact of varying dose levels and denoising strengths were investigated for a ResNet-based deep convolutional neural network (DCNN) model trained using patient images. On the uncertainty maps, DCNN shows more detailed structures than IR although its bias map has less structural dependency, which implies that DCNN is more sensitive to small changes in the input. Both visual examples and histogram analysis demonstrated that hotspots of uncertainty in DCNN may be associated with a higher chance of distortion from the truth than IR, but it may also correspond to a better detection performance for some of the small structures.

A Multivariate Analysis-Driven Workflow to Tackle Uncertainties in Miniaturized NIR Data

This study focuses on exploring and understanding measurement errors in analytical procedures involving miniaturized near-infrared instruments. Despite recent spreading in different application fields, there remains a lack of emphasis on the accuracy and reliability of these devices, which is a critical concern for accurate scientific outcomes. The study investigates multivariate measurement errors, revealing their complex nature and the influence that preprocessing techniques can have. The research introduces a possible workflow for practical error analysis in experiments involving diverse samples and instruments. Notably, it investigates how sample characteristics impact errors in the case of solid pills and tablets, typical pharmaceutical samples. ASCA was used for understanding critical instrumental factors and the potential and limitations of the method in the current application were discussed. The joint interpretation of multivariate error matrices and their resume through image histograms and K index are discussed in order to evaluate the impact of common preprocessing methods and to assess their influence on signals.

Evaluation of Uncertainties in the Anthropogenic SO2 Emissions in the USA from the OMI Point Source Catalog

Satellite remote sensing is a promising method of monitoring emissions that may be missing in inventories, but the accuracy of these estimates is often not clear. We demonstrate here a comprehensive evaluation of errors in anthropogenic sulfur dioxide (SO₂) emission estimates from NASA's OMI point source catalog for the contiguous US by comparing emissions from the catalog with high-quality emission inventory data over different dimensions including size of individual sources, aggregate vs individual source errors, and potential bias in individual source estimates over time. For sources that are included in the catalog, we find that errors in aggregate (sum of error for all included sources) are relatively low. Errors for individual sources in any given year can be substantial, however, with over- or underestimates in terms of total error ranging from -80 to 110 kt (roughly 10-90th percentile). We find that these errors are not necessarily random over time and that there can be consistently positive or negative biases for individual sources. We did not find any overall statistical relationship between the degree of isolation of a source and bias, either at a 40 or 70 km scales. For a sub-set of sources where inventory emissions over a radius of 70 km around an OMI detection are larger than twice the emissions within 40 km, the OMI value is consistently overestimated. We find, as expected, that emission sources not included in the catalog are the largest aggregate source of difference between the satellite estimates and inventories, especially in more recent years where source emission magnitudes have been decreasing and note that trends in satellite detections do not necessarily track trends in total emissions. We find that the OMI-based SO₂ emissions are accurate in aggregate, when summed over a number of sources, but must be interpreted more cautiously at the individual source level. Similar analyses would be valuable for other satellite emission estimates; however, in many cases, the appropriate high-quality reference data may need to be generated.

Exploiting Data Uncertainty for Improving the Performance of a Quantitative Analysis Model for Laser-Induced Breakdown Spectroscopy

The accuracy and precision of laser-induced breakdown spectroscopy (LIBS) quantitative analysis are significantly limited by the spectral noise. Normalization and ensemble averaging of multiple spectra were often used to preprocess spectra. However, these methods cannot completely remove the spectral noise. Data uncertainty due to the irremovable spectral noise will affect LIBS quantitative analysis. Therefore, this paper proposes a method using data uncertainty to improve the performance of LIBS quantitative analysis. The proposed method uses several spectra to characterize each sample to preserve some data uncertainty in the calibration data matrix. Thus, the data uncertainty is used to optimize the calibration model for improving the toleration to the spectral signal variation. As a result, the optimized calibration model had better accuracy and robustness than the calibration model trained by conventional method. The best root mean square error of prediction (RMSEP) of the ash content of coal was 1.152% for the optimized calibration model, while that for the conventional calibration model was 1.718%. The optimized calibration model also showed a lower relative standard deviation (RSD) value of repeated predictions. Moreover, the calibration model for predicting the ash content in biomass was also optimized by the proposed method. The optimized calibration model outperformed the conventional calibration model again, which demonstrated the extensive applicability of the proposed method.

A diversity index for model space selection in the estimation of benchmark and infectious doses via model averaging

In chemical and microbial risk assessments, risk assessors fit dose-response models to high-dose data and extrapolate downward to risk levels in the range of 1-10%. Although multiple dose-response models may be able to fit the data adequately in the experimental range, the estimated effective dose (ED) corresponding to an extremely small risk can be substantially different from model to model. In this respect, model averaging (MA) provides more robustness than a single dose-response model in the point and interval estimation of an ED. In MA, accounting for both data uncertainty and model uncertainty is crucial, but addressing model uncertainty is not achieved simply by increasing the number of models in a model space. A plausible set of models for MA can be characterized by goodness of fit and diversity surrounding the truth. We propose a diversity index (DI) to balance between these two characteristics in model space selection. It addresses a collective property of a model space rather than individual performance of each model. Tuning parameters in the DI control the size of the model space for MA.

A Bayesian Maximum Entropy approach to address the change of support problem in the spatial analysis of childhood asthma prevalence across North Carolina

The spatial analysis of data observed at different spatial observation scales leads to the change of support problem (COSP). A solution to the COSP widely used in linear spatial statistics consists in explicitly modeling the spatial autocorrelation of the variable observed at different spatial scales. We present a novel approach that takes advantage of the non-linear Bayesian Maximum Entropy (BME) extension of linear spatial statistics to address the COSP directly without relying on the classical linear approach. Our procedure consists in modeling data observed over large areas as soft data for the process at the local scale. We demonstrate the application of our approach to obtain spatially detailed maps of childhood asthma prevalence across North Carolina (NC). Because of the high prevalence of childhood asthma in NC, the small number problem is not an issue, so we can focus our attention solely to the COSP of integrating prevalence data observed at the county-level together with data observed at a targeted local scale equivalent to the scale of school-districts. Our spatially detailed maps can be used for different applications ranging from exploratory and hypothesis generating analyses to targeting intervention and exposure mitigation efforts.