Common methodological pitfalls in ICI pneumonitis risk prediction studies.

Background: Pneumonitis is one of the most common adverse events induced by the use of immune checkpoint inhibitors (ICI), accounting for a 20% of all ICI-associated deaths. Despite numerous efforts to identify risk factors and develop predictive models, there is no clinically deployed risk prediction model for patient risk stratification or for guiding subsequent monitoring. We believe this is due to systemic suboptimal approaches in study designs and methodologies in the literature. The nature and prevalence of different methodological approaches has not been thoroughly examined in prior systematic reviews. Methods: The PubMed, medRxiv and bioRxiv databases were used to identify studies that aimed at risk factor discovery and/or risk prediction model development for ICI-induced pneumonitis (ICI pneumonitis). Studies were then analysed to identify common methodological pitfalls and their contribution to the risk of bias, assessed using the QUIPS and PROBAST tools. Results: There were 51 manuscripts eligible for the review, with Japan-based studies over-represented, being nearly half (24/51) of all papers considered. Only 2/51 studies had a low risk of bias overall. Common bias-inducing practices included unclear diagnostic method or potential misdiagnosis, lack of multiple testing correction, the use of univariate analysis for selecting features for multivariable analysis, discretization of continuous variables, and inappropriate handling of missing values. Results from the risk model development studies were also likely to have been overoptimistic due to lack of holdout sets. Conclusions: Studies with low risk of bias in their methodology are lacking in the existing literature. High-quality risk factor identification and risk model development studies are urgently required by the community to give the best chance of them progressing into a clinically deployable risk prediction model. Recommendations and alternative approaches for reducing the risk of bias were also discussed to guide future studies.

The impact of imputation quality on machine learning classifiers for datasets with missing values.

BACKGROUND: Classifying samples in incomplete datasets is a common aim for machine learning practitioners, but is non-trivial. Missing data is found in most real-world datasets and these missing values are typically imputed using established methods, followed by classification of the now complete samples. The focus of the machine learning researcher is to optimise the classifier's performance. METHODS: We utilise three simulated and three real-world clinical datasets with different feature types and missingness patterns. Initially, we evaluate how the downstream classifier performance depends on the choice of classifier and imputation methods. We employ ANOVA to quantitatively evaluate how the choice of missingness rate, imputation method, and classifier method influences the performance. Additionally, we compare commonly used methods for assessing imputation quality and introduce a class of discrepancy scores based on the sliced Wasserstein distance. We also assess the stability of the imputations and the interpretability of model built on the imputed data. RESULTS: The performance of the classifier is most affected by the percentage of missingness in the test data, with a considerable performance decline observed as the test missingness rate increases. We also show that the commonly used measures for assessing imputation quality tend to lead to imputed data which poorly matches the underlying data distribution, whereas our new class of discrepancy scores performs much better on this measure. Furthermore, we show that the interpretability of classifier models trained using poorly imputed data is compromised. CONCLUSIONS: It is imperative to consider the quality of the imputation when performing downstream classification as the effects on the classifier can be considerable.

Many artificial intelligence (AI) methods aim to classify samples of data into groups, e.g., patients with disease vs. those without. This often requires datasets to be complete, i.e., that all data has been collected for all samples. However, in clinical practice this is often not the case and some data can be missing. One solution is to 'complete' the dataset using a technique called imputation to replace those missing values. However, assessing how well the imputation method performs is challenging. In this work, we demonstrate why people should care about imputation, develop a new method for assessing imputation quality, and demonstrate that if we build AI models on poorly imputed data, the model can give different results to those we would hope for. Our findings may improve the utility and quality of AI models in the clinic.

A pipeline to further enhance quality, integrity and reusability of the NCCID clinical data.

The National COVID-19 Chest Imaging Database (NCCID) is a centralized UK database of thoracic imaging and corresponding clinical data. It is made available by the National Health Service Artificial Intelligence (NHS AI) Lab to support the development of machine learning tools focused on Coronavirus Disease 2019 (COVID-19). A bespoke cleaning pipeline for NCCID, developed by the NHSx, was introduced in 2021. We present an extension to the original cleaning pipeline for the clinical data of the database. It has been adjusted to correct additional systematic inconsistencies in the raw data such as patient sex, oxygen levels and date values. The most important changes will be discussed in this paper, whilst the code and further explanations are made publicly available on GitLab. The suggested cleaning will allow global users to work with more consistent data for the development of machine learning tools without being an expert. In addition, it highlights some of the challenges when working with clinical multi-center data and includes recommendations for similar future initiatives.

Aß-induced vulnerability propagates via the brain's default mode network.

The link between brain amyloid-ß (Aß), metabolism, and dementia symptoms remains a pressing question in Alzheimer's disease. Here, using positron emission tomography ([18F]florbetapir tracer for Aß and [18F]FDG tracer for glucose metabolism) with a novel analytical framework, we found that Aß aggregation within the brain's default mode network leads to regional hypometabolism in distant but functionally connected brain regions. Moreover, we found that an interaction between this hypometabolism with overlapping Aß aggregation is associated with subsequent cognitive decline. These results were also observed in transgenic Aß rats that do not form neurofibrillary tangles, which support these findings as an independent mechanism of cognitive deterioration. These results suggest a model in which distant Aß induces regional metabolic vulnerability, whereas the interaction between local Aß with a vulnerable environment drives the clinical progression of dementia.

A Functional Approach to Deconvolve Dynamic Neuroimaging Data.

Positron emission tomography (PET) is an imaging technique which can be used to investigate chemical changes in human biological processes such as cancer development or neurochemical reactions. Most dynamic PET scans are currently analyzed based on the assumption that linear first-order kinetics can be used to adequately describe the system under observation. However, there has recently been strong evidence that this is not the case. To provide an analysis of PET data which is free from this compartmental assumption, we propose a nonparametric deconvolution and analysis model for dynamic PET data based on functional principal component analysis. This yields flexibility in the possible deconvolved functions while still performing well when a linear compartmental model setup is the true data generating mechanism. As the deconvolution needs to be performed on only a relative small number of basis functions rather than voxel by voxel in the entire three-dimensional volume, the methodology is both robust to typical brain imaging noise levels while also being computationally efficient. The new methodology is investigated through simulations in both one-dimensional functions and 2D images and also applied to a neuroimaging study whose goal is the quantification of opioid receptor concentration in the brain.

Estimating the population local wavelet spectrum with application to non-stationary functional magnetic resonance imaging time series.

Functional magnetic resonance imaging (fMRI) is a dynamic four-dimensional imaging modality. However, in almost all fMRI analyses, the time series elements of this data are assumed to be second-order stationary. In this paper, we examine, using time series spectral methods, whether such stationary assumptions can be made and whether estimates of non-stationarity can be used to gain understanding into fMRI experiments. A non-stationary version of replicated stationary time series analysis is proposed that takes into account the replicated time series that are available from nearby voxels in a region of interest (ROI). These are used to investigate non-stationarities in both the ROI itself and the variations within the ROI. The proposed techniques are applied to simulated data and to an anxiety-inducing fMRI experiment.

Adaptive Multivariate Global Testing.

We present a methodology for dealing with recent challenges in testing global hypotheses using multivariate observations. The proposed tests target situations, often arising in emerging applications of neuroimaging, where the sample size n is relatively small compared with the observations' dimension K. We employ adaptive designs allowing for sequential modifications of the test statistics adapting to accumulated data. The adaptations are optimal in the sense of maximizing the predictive power of the test at each interim analysis while still controlling the Type I error. Optimality is obtained by a general result applicable to typical adaptive design settings. Further, we prove that the potentially high-dimensional design space of the tests can be reduced to a low-dimensional projection space enabling us to perform simpler power analysis studies, including comparisons to alternative tests. We illustrate the substantial improvement in efficiency that the proposed tests can make over standard tests, especially in the case of n smaller or slightly larger than K. The methods are also studied empirically using both simulated data and data from an EEG study, where the use of prior knowledge substantially increases the power of the test. Supplementary materials for this article are available online.

Mutation rate plasticity in rifampicin resistance depends on Escherichia coli cell-cell interactions.

Variation of mutation rate at a particular site in a particular genotype, in other words mutation rate plasticity (MRP), can be caused by stress or ageing. However, mutation rate control by other factors is less well characterized. Here we show that in wild-type Escherichia coli (K-12 and B strains), the mutation rate to rifampicin resistance is plastic and inversely related to population density: lowering density can increase mutation rates at least threefold. This MRP is genetically switchable, dependent on the quorum-sensing gene luxS--specifically its role in the activated methyl cycle--and is socially mediated via cell-cell interactions. Although we identify an inverse association of mutation rate with fitness under some circumstances, we find no functional link with stress-induced mutagenesis. Our experimental manipulation of mutation rates via the social environment raises the possibility that such manipulation occurs in nature and could be exploited medically.

Reprint of "Pharmacokinetic modelling of the anti-malarial drug artesunate and its active metabolite dihydroartemisinin".

A four compartment mechanistic mathematical model is developed for the pharmacokinetics of the commonly used anti-malarial drug artesunate and its principle metabolite dihydroartemisinin following oral administration of artesunate. The model is structurally unidentifiable unless additional constraints are imposed. Combinations of mechanistically derived constraints are considered to assess their effects on structural identifiability and on model fits. Certain combinations of the constraints give rise to locally or globally identifiable model structures. Initial validation of the model under various combinations of the constraints leading to identifiable model structures was performed against a dataset of artesunate and dihydroartemisinin concentration-time profiles of 19 malaria patients. When all the discussed constraints were imposed on the model, the resulting globally identifiable model structure was found to fit reasonably well to those patients with normal drug absorption profiles. However, there is wide variability in the fitted parameters and further investigation is warranted.

Where antibiotic resistance mutations meet quorum-sensing.

We do not need to rehearse the grim story of the global rise of antibiotic resistant microbes. But what if it were possible to control the rate with which antibiotic resistance evolves by de novo mutation? It seems that some bacteria may already do exactly that: they modify the rate at which they mutate to antibiotic resistance dependent on their biological environment. In our recent study [Krasovec, et al. Nat. Commun. (2014), 5, 3742] we find that this modification depends on the density of the bacterial population and cell-cell interactions (rather than, for instance, the level of stress). Specifically, the wild-type strains of Escherichia coli we used will, in minimal glucose media, modify their rate of mutation to rifampicin resistance according to the density of wild-type cells. Intriguingly, the higher the density, the lower the mutation rate (Figure 1). Why this novel density-dependent 'mutation rate plasticity' (DD-MRP) occurs is a question at several levels. Answers are currently fragmentary, but involve the quorum-sensing gene luxS and its role in the activated methyl cycle.

Pharmacokinetic modelling of the anti-malarial drug artesunate and its active metabolite dihydroartemisinin.

A four compartment mechanistic mathematical model is developed for the pharmacokinetics of the commonly used anti-malarial drug artesunate and its principle metabolite dihydroartemisinin following oral administration of artesunate. The model is structurally unidentifiable unless additional constraints are imposed. Combinations of mechanistically derived constraints are considered to assess their effects on structural identifiability and on model fits. Certain combinations of the constraints give rise to locally or globally identifiable model structures. Initial validation of the model under various combinations of the constraints leading to identifiable model structures was performed against a dataset of artesunate and dihydroartemisinin concentration-time profiles of 19 malaria patients. When all the discussed constraints were imposed on the model, the resulting globally identifiable model structure was found to fit reasonably well to those patients with normal drug absorption profiles. However, there is wide variability in the fitted parameters and further investigation is warranted.

Characterizing fundamental frequency in Mandarin: a functional principal component approach utilizing mixed effect models.

A model for fundamental frequency (F0, or commonly pitch) employing a functional principal component (FPC) analysis framework is presented. The model is applied to Mandarin Chinese; this Sino-Tibetan language is rich in pitch-related information as the relative pitch curve is specified for most syllables in the lexicon. The approach yields a quantification of the influence carried by each identified component in relation to original tonal content, without formulating any assumptions on the shape of the tonal components. The original five speaker corpus is preprocessed using a locally weighted least squares smoother to produce F0 curves. These smoothed curves are then utilized as input for the computation of FPC scores and their corresponding eigenfunctions. These scores are analyzed in a series of penalized mixed effect models, through which meaningful categorical prototypes are built. The prototypes appear to confirm known tonal characteristics of the language, as well as suggest the presence of a sinusoid tonal component that is previously undocumented.

A self-organizing state-space-model approach for parameter estimation in hodgkin-huxley-type models of single neurons.

Traditional approaches to the problem of parameter estimation in biophysical models of neurons and neural networks usually adopt a global search algorithm (for example, an evolutionary algorithm), often in combination with a local search method (such as gradient descent) in order to minimize the value of a cost function, which measures the discrepancy between various features of the available experimental data and model output. In this study, we approach the problem of parameter estimation in conductance-based models of single neurons from a different perspective. By adopting a hidden-dynamical-systems formalism, we expressed parameter estimation as an inference problem in these systems, which can then be tackled using a range of well-established statistical inference methods. The particular method we used was Kitagawa's self-organizing state-space model, which was applied on a number of Hodgkin-Huxley-type models using simulated or actual electrophysiological data. We showed that the algorithm can be used to estimate a large number of parameters, including maximal conductances, reversal potentials, kinetics of ionic currents, measurement and intrinsic noise, based on low-dimensional experimental data and sufficiently informative priors in the form of pre-defined constraints imposed on model parameters. The algorithm remained operational even when very noisy experimental data were used. Importantly, by combining the self-organizing state-space model with an adaptive sampling algorithm akin to the Covariance Matrix Adaptation Evolution Strategy, we achieved a significant reduction in the variance of parameter estimates. The algorithm did not require the explicit formulation of a cost function and it was straightforward to apply on compartmental models and multiple data sets. Overall, the proposed methodology is particularly suitable for resolving high-dimensional inference problems based on noisy electrophysiological data and, therefore, a potentially useful tool in the construction of biophysical neuron models.

A hybrid procedure for detecting global treatment effects in multivariate clinical trials: theory and applications to fMRI studies.

In multivariate clinical trials, a key research endpoint is ascertaining whether a candidate treatment is more efficacious than an established alternative. This global endpoint is clearly of high practical value for studies, such as those arising from neuroimaging, where the outcome dimensions are not only numerous but they are also highly correlated and the available sample sizes are typically small. In this paper, we develop a two-stage procedure testing the null hypothesis of global equivalence between treatments effects and demonstrate its application to analysing phase II neuroimaging trials. Prior information such as suitable statistics of historical data or suitably elicited expert clinical opinions are combined with data collected from the first stage of the trial to learn a set of optimal weights. We apply these weights to the outcome dimensions of the second-stage responses to form the linear combination z and t tests statistics while controlling the test's false positive rate. We show that the proposed tests hold desirable asymptotic properties and characterise their power functions under wide conditions. In particular, by comparing the power of the proposed tests with that of Hotelling's T(2), we demonstrate their advantages when sample sizes are close to the dimension of the multivariate outcome. We apply our methods to fMRI studies, where we find that, for sufficiently precise first stage estimates of the treatment effect, standard single-stage testing procedures are outperformed.

Linguistic and human effects on F(0) in a tonal dialect of Qiang.

While both human and linguistic factors affect fundamental frequency (F(0)) in spoken language, capturing the influence of multiple effects and their interactions presents special challenges, especially when there are strict time constraints on the data-gathering process. A lack of speaker literacy can further impede the collection of identical utterances across multiple speakers. This study employs linear mixed effects analysis to elucidate how various effects and their interactions contribute to the production of F(0) in Luobuzhai, a tonal dialect of the Qiang language. In addition to the effects of speaker sex and tone, F(0) in this language is affected by previous and following tones, sentence type, vowel, position in the phrase, and by numerous combinations of these effects. Under less than ideal data collecting conditions, a single experiment was able to yield an extensive model of F(0) output in an endangered language of the Himalayas.

Smoothing dynamic positron emission tomography time courses using functional principal components.

A functional smoothing approach to the analysis of PET time course data is presented. By borrowing information across space and accounting for this pooling through the use of a nonparametric covariate adjustment, it is possible to smooth the PET time course data thus reducing the noise. A new model for functional data analysis, the Multiplicative Nonparametric Random Effects Model, is introduced to more accurately account for the variation in the data. A locally adaptive bandwidth choice helps to determine the correct amount of smoothing at each time point. This preprocessing step to smooth the data then allows subsequent analysis by methods such as Spectral Analysis to be substantially improved in terms of their mean squared error.

Beyond p-values: averaged and reproducible evidence in fMRI experiments.

In functional magnetic resonance imaging studies, there might exist activation regions routinely involved in experimental sessions, but modest in response magnitude. These regions may not be easily detectable by the conventional p-value approach using a rigid threshold. With particular reference to the reproducibility analysis method proposed in Liou and colleagues, this study presents some within- and between-subject brain-activation patterns that are replicable between experimental modalities, and robust to the method used for generating the patterns. There is a neurophysiological basis behind these reproducible patterns, and the conventional p-value approach using averaged data across subjects might not suggest the complete patterns. For example, recent studies based on the group-averaged data showed a task-induced deactivation in the precuneus and posterior cingulate, but our reproducibility analysis suggests both increased and decreased responses in the two regions. The increased responses localize in these regions with differentially distributed patterns for individual subjects and for different experimental tasks. In this study, we discuss the neurophysiological basis of the reproducible patterns and propose some applications of our research findings to scientific and clinical studies.

MR image segmentation using a power transformation approach.

This study proposes a segmentation method for brain MR images using a distribution transformation approach. The method extends traditional Gaussian mixtures expectation-maximization segmentation to a power transformed version of mixed intensity distributions, which includes Gaussian mixtures as a special case. As MR intensities tend to exhibit non-Gaussianity due to partial volume effects, the proposed method is designed to fit non-Gaussian tissue intensity distributions. One advantage of the method is that it is intuitively appealing and computationally simple. To avoid performance degradation caused by intensity inhomogeneity, different methods for correcting bias fields were applied prior to image segmentation, and their correction effects on the segmentation results were examined in the empirical study. The partitions of brain tissues (i.e., gray and white matter) resulting from the method were validated and evaluated against manual segmentation results based on 38 real T1-weighted image volumes from the internet brain segmentation repository, and 18 simulated image volumes from BrainWeb. The Jaccard and Dice similarity indexes were computed to evaluate the performance of the proposed approach relative to the expert segmentations. Empirical results suggested that the proposed segmentation method yielded higher similarity measures for both gray matter and white matter as compared with those based on the traditional segmentation using the Gaussian mixtures approach.

Dynamic positron emission tomography data-driven analysis using sparse Bayesian learning.

A method is presented for the analysis of dynamic positron emission tomography (PET) data using sparse Bayesian learning. Parameters are estimated in a compartmental framework using an over-complete exponential basis set and sparse Bayesian learning. The technique is applicable to analyses requiring either a plasma or reference tissue input function and produces estimates of the system's macro-parameters and model order. In addition, the Bayesian approach returns the posterior distribution which allows for some characterisation of the error component. The method is applied to the estimation of parametric images of neuroreceptor radioligand studies.

HBM functional imaging analysis contest data analysis in wavelet space.

An analysis of the Functional Imaging Analysis Contest (FIAC) data is presented using spatial wavelet processing. This technique allows the image to be filtered adaptively according to the data itself, rather than relying on a predetermined filter. This adaptive filtering leads to better estimation of the parameters and contrasts in terms of mean squared error. It will be shown that by introducing a slight bias into the estimation, a large reduction in the variance can be achieved, leading to better overall mean squared error estimates. As no single filter needs to be preselected, results containing many scales of information can be found. In the FIAC data, it is shown that both small-scale and large-scale (smoother, more dispersed) effects occur. The combination of small- and large-scale effects detected in the FIAC data would be easy to miss using conventional single filter analysis.