A machine learning-based choledocholithiasis prediction tool to improve ERCP decision making: a proof-of-concept study.

BACKGROUND: Previous studies demonstrated limited accuracy of existing guidelines for predicting choledocholithiasis, leading to overutilization of endoscopic retrograde cholangiopancreatography (ERCP). More accurate stratification may improve patient selection for ERCP and allow use of lower-risk modalities. METHODS: A machine learning model was developed using patient information from two published cohort studies that evaluated performance of guidelines in predicting choledocholithiasis. Prediction models were developed using the gradient boosting model (GBM) machine learning method. GBM performance was evaluated using 10-fold cross-validation and area under the receiver operating characteristic curve (AUC). Important predictors of choledocholithiasis were identified based on relative importance in the GBM. RESULTS: 1378 patients (mean age 43.3 years; 61.2% female) were included in the GBM and 59.4% had choledocholithiasis. Eight variables were identified as predictors of choledocholithiasis. The GBM had accuracy of 71.5% (SD 2.5%) (AUC 0.79 [SD 0.06]) and performed better than the 2019 American Society for Gastrointestinal Endoscopy (ASGE) guidelines (accuracy 62.4% [SD 2.6%]; AUC 0.63 [SD 0.03]) and European Society of Gastrointestinal Endoscopy (ESGE) guidelines (accuracy 62.8% [SD 2.6%]; AUC 0.67 [SD 0.02]). The GBM correctly categorized 22% of patients directed to unnecessary ERCP by ASGE guidelines, and appropriately recommended as the next management step 48% of ERCPs incorrectly rejected by ESGE guidelines. CONCLUSIONS: A machine learning-based tool was created, providing real-time, personalized, objective probability of choledocholithiasis and ERCP recommendations. This more accurately directed ERCP use than existing ASGE and ESGE guidelines, and has the potential to reduce morbidity associated with ERCP or missed choledocholithiasis.

Rethinking recovery in adolescent concussions: Network-level functional connectivity alterations associated with motor deficits.

Adolescents who are clinically recovered from concussion continue to show subtle motor impairment on neurophysiological and behavioral measures. However, there is limited information on brain-behavior relationships of persistent motor impairment following clinical recovery from concussion. We examined the relationship between subtle motor performance and functional connectivity of the brain in adolescents with a history of concussion, status post-symptom resolution, and subjective return to baseline. Participants included 27 adolescents who were clinically recovered from concussion and 29 never-concussed, typically developing controls (10-17 years); all participants were examined using the Physical and Neurologic Examination of Subtle Signs (PANESS). Functional connectivity between the default mode network (DMN) or dorsal attention network (DAN) and regions of interest within the motor network was assessed using resting-state functional magnetic resonance imaging (rsfMRI). Compared to controls, adolescents clinically recovered from concussion showed greater subtle motor deficits as evaluated by the PANESS and increased connectivity between the DMN and left lateral premotor cortex. DMN to left lateral premotor cortex connectivity was significantly correlated with the total PANESS score, with more atypical connectivity associated with more motor abnormalities. This suggests that altered functional connectivity of the brain may underlie subtle motor deficits in adolescents who have clinically recovered from concussion. More investigation is required to understand the persistence and longer-term clinical relevance of altered functional connectivity and associated subtle motor deficits to inform whether functional connectivity may serve as an important biomarker related to longer-term outcomes after clinical recovery from concussion.

Longitudinal regression of covariance matrix outcomes.

In this study, a longitudinal regression model for covariance matrix outcomes is introduced. The proposal considers a multilevel generalized linear model for regressing covariance matrices on (time-varying) predictors. This model simultaneously identifies covariate-associated components from covariance matrices, estimates regression coefficients, and captures the within-subject variation in the covariance matrices. Optimal estimators are proposed for both low-dimensional and high-dimensional cases by maximizing the (approximated) hierarchical-likelihood function. These estimators are proved to be asymptotically consistent, where the proposed covariance matrix estimator is the most efficient under the low-dimensional case and achieves the uniformly minimum quadratic loss among all linear combinations of the identity matrix and the sample covariance matrix under the high-dimensional case. Through extensive simulation studies, the proposed approach achieves good performance in identifying the covariate-related components and estimating the model parameters. Applying to a longitudinal resting-state functional magnetic resonance imaging data set from the Alzheimer's Disease (AD) Neuroimaging Initiative, the proposed approach identifies brain networks that demonstrate the difference between males and females at different disease stages. The findings are in line with existing knowledge of AD and the method improves the statistical power over the analysis of cross-sectional data.

Identifying brain hierarchical structures associated with Alzheimer's disease using a regularized regression method with tree predictors.

Brain segmentation at different levels is generally represented as hierarchical trees. Brain regional atrophy at specific levels was found to be marginally associated with Alzheimer's disease outcomes. In this study, we propose an ℓ1 -type regularization for predictors that follow a hierarchical tree structure. Considering a tree as a directed acyclic graph, we interpret the model parameters from a path analysis perspective. Under this concept, the proposed penalty regulates the total effect of each predictor on the outcome. With regularity conditions, it is shown that under the proposed regularization, the estimator of the model coefficient is consistent in ℓ2 -norm and the model selection is also consistent. When applied to a brain sMRI dataset acquired from the Alzheimer's Disease Neuroimaging Initiative (ADNI), the proposed approach identifies brain regions where atrophy in these regions demonstrates the declination in memory. With regularization on the total effects, the findings suggest that the impact of atrophy on memory deficits is localized from small brain regions, but at various levels of brain segmentation. Data used in preparation of this paper were obtained from the ADNI database.

Covariate Assisted Principal regression for covariance matrix outcomes.

In this study, we consider the problem of regressing covariance matrices on associated covariates. Our goal is to use covariates to explain variation in covariance matrices across units. As such, we introduce Covariate Assisted Principal (CAP) regression, an optimization-based method for identifying components associated with the covariates using a generalized linear model approach. We develop computationally efficient algorithms to jointly search for common linear projections of the covariance matrices, as well as the regression coefficients. Under the assumption that all the covariance matrices share identical eigencomponents, we establish the asymptotic properties. In simulation studies, our CAP method shows higher accuracy and robustness in coefficient estimation over competing methods. In an example resting-state functional magnetic resonance imaging study of healthy adults, CAP identifies human brain network changes associated with subject demographics.

B-value and empirical equivalence bound: A new procedure of hypothesis testing.

In this study, we propose a two-stage procedure for hypothesis testing, where the first stage is conventional hypothesis testing and the second is an equivalence testing procedure using an introduced empirical equivalence bound (EEB). In 2016, the American Statistical Association released a policy statement on P-values to clarify the proper use and interpretation in response to the criticism of reproducibility and replicability in scientific findings. A recent solution to improve reproducibility and transparency in statistical hypothesis testing is to integrate P-values (or confidence intervals) with practical or scientific significance. Similar ideas have been proposed via the equivalence test, where the goal is to infer equality under a presumption (null) of inequality of parameters. However, the definition of scientific significance/equivalence can sometimes be ill-justified and subjective. To circumvent this drawback, we introduce the B-value and the EEB, which are both estimated from the data. Performing a second-stage equivalence test, our procedure offers an opportunity to improve the reproducibility of findings across studies.

Identification of the Somatomotor Network from Language Task-based fMRI Compared with Resting-State fMRI in Patients with Brain Lesions.

Background Resting-state functional MRI (rs-fMRI) is a potential alternative to task-based functional MRI (tb-fMRI) for somatomotor network (SMN) identification. Brain networks can also be generated from tb-fMRI by using independent component analysis (ICA). Purpose To investigate whether the SMN can be identified by using ICA from a language task without a motor component, the sentence completion functional MRI (sc-fMRI) task, compared with rs-fMRI. Materials and Methods The sc-fMRI and rs-fMRI scans in patients who underwent presurgical brain mapping between 2012 and 2016 were analyzed, using the same imaging parameters (other than scanning time) on a 3.0-T MRI scanner. ICA was performed on rs-fMRI and sc-fMRI scans with use of a tool to separate data sets into their spatial and temporal components. Two neuroradiologists independently determined the presence of the dorsal SMN (dSMN) and ventral SMN (vSMN) on each study. Groups were compared by using t tests, and logistic regression was performed to identify predictors of the presence of SMNs. Results One hundred patients (mean age, 40.9 years ± 14.8 [standard deviation]; 61 men) were evaluated. The dSMN and vSMN were identified in 86% (86 of 100) and 76% (76 of 100) of rs-fMRI scans and 85% (85 of 100) and 69% (69 of 100) of sc-fMRI scans, respectively. The concordance between rs-fMRI and sc-fMRI for presence of dSMN and vSMN was 75% (75 of 100 patients) and 53% (53 of 100 patients), respectively. In 10 of 14 patients (71%) where rs-fMRI did not show the dSMN, sc-fMRI demonstrated it. This rate was 67% for the vSMN (16 of 24 patients). Conclusion In the majority of patients, independent component analysis of sentence completion task functional MRI scans reliably demonstrated the somatomotor network compared with resting-state functional MRI scans. Identifying target networks with a single sentence completion scan could reduce overall functional MRI scanning times by eliminating the need for separate motor tasks. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Field and Birn in this issue.

Multimodal neuroimaging data integration and pathway analysis.

With advancements in technology, the collection of multiple types of measurements on a common set of subjects is becoming routine in science. Some notable examples include multimodal neuroimaging studies for the simultaneous investigation of brain structure and function and multi-omics studies for combining genetic and genomic information. Integrative analysis of multimodal data allows scientists to interrogate new mechanistic questions. However, the data collection and generation of integrative hypotheses is outpacing available methodology for joint analysis of multimodal measurements. In this article, we study high-dimensional multimodal data integration in the context of mediation analysis. We aim to understand the roles that different data modalities play as possible mediators in the pathway between an exposure variable and an outcome. We propose a mediation model framework with two data types serving as separate sets of mediators and develop a penalized optimization approach for parameter estimation. We study both the theoretical properties of the estimator through an asymptotic analysis and its finite-sample performance through simulations. We illustrate our method with a multimodal brain pathway analysis having both structural and functional connectivity as mediators in the association between sex and language processing.

Long-term kidney function and survival in recipients of allografts from living kidney donors with hypertension: a national cohort study.

Allografts from living kidney donors with hypertension may carry subclinical kidney disease from the donor to the recipient and, thus, lead to adverse recipient outcomes. We examined eGFR trajectories and all-cause allograft failure in recipients from donors with versus without hypertension, using mixed-linear and Cox regression models stratified by donor age. We studied a US cohort from 1/1/2005 to 6/30/2017; 49 990 recipients of allografts from younger (<50 years old) donors including 597 with donor hypertension and 21 130 recipients of allografts from older (≥50 years old) donors including 1441 with donor hypertension. Donor hypertension was defined as documented predonation use of antihypertensive therapy. Among recipients from younger donors with versus without hypertension, the annual eGFR decline was -1.03 versus -0.53 ml/min/m2 (P = 0.002); 13-year allograft survival was 49.7% vs. 59.0% (adjusted allograft failure hazard ratio [aHR] 1.23; 95% CI 1.05-1.43; P = 0.009). Among recipients from older donors with versus without hypertension, the annual eGFR decline was -0.67 versus -0.66 ml/min/m2 (P = 0.9); 13-year allograft survival was 48.6% versus 52.6% (aHR 1.05; 95% CI 0.94-1.17; P = 0.4). In secondary analyses, our inferences remained similar for risk of death-censored allograft failure and mortality. Hypertension in younger, but not older, living kidney donors is associated with worse recipient outcomes.

Neuroconductor: an R platform for medical imaging analysis.

Neuroconductor (https://neuroconductor.org) is an open-source platform for rapid testing and dissemination of reproducible computational imaging software. The goals of the project are to: (i) provide a centralized repository of R software dedicated to image analysis, (ii) disseminate software updates quickly, (iii) train a large, diverse community of scientists using detailed tutorials and short courses, (iv) increase software quality via automatic and manual quality controls, and (v) promote reproducibility of image data analysis. Based on the programming language R (https://www.r-project.org/), Neuroconductor starts with 51 inter-operable packages that cover multiple areas of imaging including visualization, data processing and storage, and statistical inference. Neuroconductor accepts new R package submissions, which are subject to a formal review and continuous automated testing. We provide a description of the purpose of Neuroconductor and the user and developer experience.

Machine learning to predict transplant outcomes: helpful or hype? A national cohort study.

An increasing number of studies claim machine learning (ML) predicts transplant outcomes more accurately. However, these claims were possibly confounded by other factors, namely, supplying new variables to ML models. To better understand the prospects of ML in transplantation, we compared ML to conventional regression in a "common" analytic task: predicting kidney transplant outcomes using national registry data. We studied 133 431 adult deceased-donor kidney transplant recipients between 2005 and 2017. Transplant centers were randomly divided into 70% training set (190 centers/97 787 recipients) and 30% validation set (82 centers/35 644 recipients). Using the training set, we performed regression and ML procedures [gradient boosting (GB) and random forests (RF)] to predict delayed graft function, one-year acute rejection, death-censored graft failure C, all-cause graft failure, and death. Their performances were compared on the validation set using -statistics. In predicting rejection, regression (C = 0.601 0.6110.621 ) actually outperformed GB (C = 0.581 0.5910.601 ) and RF (C = 0.569 0.5790.589 ). For all other outcomes, the C-statistics were nearly identical across methods (delayed graft function, 0.717-0.723; death-censored graft failure, 0.637-0.642; all-cause graft failure, 0.633-0.635; and death, 0.705-0.708). Given its shortcomings in model interpretability and hypothesis testing, ML is advantageous only when it clearly outperforms conventional regression; in the case of transplant outcomes prediction, ML seems more hype than helpful.

Sparse Principal Component based High-Dimensional Mediation Analysis.

Causal mediation analysis aims to quantify the intermediate effect of a mediator on the causal pathway from treatment to outcome. When dealing with multiple mediators, which are potentially causally dependent, the possible decomposition of pathway effects grows exponentially with the number of mediators. An existing approach incorporated the principal component analysis (PCA) to address this challenge based on the fact that the transformed mediators are conditionally independent given the orthogonality of the principal components (PCs). However, the transformed mediator PCs, which are linear combinations of original mediators, can be difficult to interpret. A sparse high-dimensional mediation analysis approach is proposed which adopts the sparse PCA method to the mediation setting. The proposed approach is applied to a task-based functional magnetic resonance imaging study, illustrating its ability to detect biologically meaningful results related to an identified mediator.

Improved state change estimation in dynamic functional connectivity using hidden semi-Markov models.

The study of functional brain networks has grown rapidly over the past decade. While most functional connectivity (FC) analyses estimate one static network structure for the entire length of the functional magnetic resonance imaging (fMRI) time series, recently there has been increased interest in studying time-varying changes in FC. Hidden Markov models (HMMs) have proven to be a useful modeling approach for discovering repeating graphs of interacting brain regions (brain states). However, a limitation lies in HMMs assuming that the sojourn time, the number of consecutive time points in a state, is geometrically distributed. This may encourage inaccurate estimation of the time spent in a state before switching to another state. We propose a hidden semi-Markov model (HSMM) approach for inferring time-varying brain networks from fMRI data, which explicitly models the sojourn distribution. Specifically, we propose using HSMMs to find each subject's most probable series of network states and the graphs associated with each state, while properly estimating and modeling the sojourn distribution for each state. We perform a simulation study, as well as an analysis on both task-based fMRI data from an anxiety-inducing experiment and resting-state fMRI data from the Human Connectome Project. Our results demonstrate the importance of model choice when estimating sojourn times and reveal their potential for understanding healthy and diseased brain mechanisms.

Modular preprocessing pipelines can reintroduce artifacts into fMRI data.

The preprocessing pipelines typically used in both task and resting-state functional magnetic resonance imaging (rs-fMRI) analysis are modular in nature: They are composed of a number of separate filtering/regression steps, including removal of head motion covariates and band-pass filtering, performed sequentially and in a flexible order. In this article, we illustrate the shortcomings of this approach, as we show how later preprocessing steps can reintroduce artifacts previously removed from the data in prior preprocessing steps. We show that each regression step is a geometric projection of data onto a subspace, and that performing a sequence of projections can move the data into subspaces no longer orthogonal to those previously removed, reintroducing signal related to nuisance covariates. Thus, linear filtering operations are not commutative, and the order in which the preprocessing steps are performed is critical. These issues can arise in practice when any combination of standard preprocessing steps including motion regression, scrubbing, component-based correction, physiological correction, global signal regression, and temporal filtering are performed sequentially. In this work, we focus primarily on rs-fMRI. We illustrate the problem both theoretically and empirically through application to a test-retest rs-fMRI data set, and suggest remedies. These include (a) combining all steps into a single linear filter, or (b) sequential orthogonalization of covariates/linear filters performed in series.

High-dimensional multivariate mediation with application to neuroimaging data.

Mediation analysis is an important tool in the behavioral sciences for investigating the role of intermediate variables that lie in the path between a treatment and an outcome variable. The influence of the intermediate variable on the outcome is often explored using a linear structural equation model (LSEM), with model coefficients interpreted as possible effects. While there has been significant research on the topic, little work has been done when the intermediate variable (mediator) is a high-dimensional vector. In this work, we introduce a novel method for identifying potential mediators in this setting called the directions of mediation (DMs). DMs linearly combine potential mediators into a smaller number of orthogonal components, with components ranked based on the proportion of the LSEM likelihood each accounts for. This method is well suited for cases when many potential mediators are measured. Examples of high-dimensional potential mediators are brain images composed of hundreds of thousands of voxels, genetic variation measured at millions of single nucleotide polymorphisms (SNPs), or vectors of thousands of variables in large-scale epidemiological studies. We demonstrate the method using a functional magnetic resonance imaging study of thermal pain where we are interested in determining which brain locations mediate the relationship between the application of a thermal stimulus and self-reported pain.

Mixed effect machine learning: A framework for predicting longitudinal change in hemoglobin A1c.

Accurate and reliable prediction of clinical progression over time has the potential to improve the outcomes of chronic disease. The classical approach to analyzing longitudinal data is to use (generalized) linear mixed-effect models (GLMM). However, linear parametric models are predicated on assumptions, which are often difficult to verify. In contrast, data-driven machine learning methods can be applied to derive insight from the raw data without a priori assumptions. However, the underlying theory of most machine learning algorithms assume that the data is independent and identically distributed, making them inefficient for longitudinal supervised learning. In this study, we formulate an analytic framework, which integrates the random-effects structure of GLMM into non-linear machine learning models capable of exploiting temporal heterogeneous effects, sparse and varying-length patient characteristics inherent in longitudinal data. We applied the derived mixed-effect machine learning (MEml) framework to predict longitudinal change in glycemic control measured by hemoglobin A1c (HbA1c) among well controlled adults with type 2 diabetes. Results show that MEml is competitive with traditional GLMM, but substantially outperformed standard machine learning models that do not account for random-effects. Specifically, the accuracy of MEml in predicting glycemic change at the 1st, 2nd, 3rd, and 4th clinical visits in advanced was 1.04, 1.08, 1.11, and 1.14 times that of the gradient boosted model respectively, with similar results for the other methods. To further demonstrate the general applicability of MEml, a series of experiments were performed using real publicly available and synthetic data sets for accuracy and robustness. These experiments reinforced the superiority of MEml over the other methods. Overall, results from this study highlight the importance of modeling random-effects in machine learning approaches based on longitudinal data. Our MEml method is highly resistant to correlated data, readily accounts for random-effects, and predicts change of a longitudinal clinical outcome in real-world clinical settings with high accuracy.

Improved estimation of subject-level functional connectivity using full and partial correlation with empirical Bayes shrinkage.

Reliability of subject-level resting-state functional connectivity (FC) is determined in part by the statistical techniques employed in its estimation. Methods that pool information across subjects to inform estimation of subject-level effects (e.g., Bayesian approaches) have been shown to enhance reliability of subject-level FC. However, fully Bayesian approaches are computationally demanding, while empirical Bayesian approaches typically rely on using repeated measures to estimate the variance components in the model. Here, we avoid the need for repeated measures by proposing a novel measurement error model for FC describing the different sources of variance and error, which we use to perform empirical Bayes shrinkage of subject-level FC towards the group average. In addition, since the traditional intra-class correlation coefficient (ICC) is inappropriate for biased estimates, we propose a new reliability measure denoted the mean squared error intra-class correlation coefficient (ICCMSE) to properly assess the reliability of the resulting (biased) estimates. We apply the proposed techniques to test-retest resting-state fMRI data on 461 subjects from the Human Connectome Project to estimate connectivity between 100 regions identified through independent components analysis (ICA). We consider both correlation and partial correlation as the measure of FC and assess the benefit of shrinkage for each measure, as well as the effects of scan duration. We find that shrinkage estimates of subject-level FC exhibit substantially greater reliability than traditional estimates across various scan durations, even for the most reliable connections and regardless of connectivity measure. Additionally, we find partial correlation reliability to be highly sensitive to the choice of penalty term, and to be generally worse than that of full correlations except for certain connections and a narrow range of penalty values. This suggests that the penalty needs to be chosen carefully when using partial correlations.

An M-estimator for reduced-rank system identification.

High-dimensional time-series data from a wide variety of domains, such as neuroscience, are being generated every day. Fitting statistical models to such data, to enable parameter estimation and time-series prediction, is an important computational primitive. Existing methods, however, are unable to cope with the high-dimensional nature of these data, due to both computational and statistical reasons. We mitigate both kinds of issues by proposing an M-estimator for Reduced-rank System IDentification ( MR. SID). A combination of low-rank approximations, ℓ1 and ℓ2 penalties, and some numerical linear algebra tricks, yields an estimator that is computationally efficient and numerically stable. Simulations and real data examples demonstrate the usefulness of this approach in a variety of problems. In particular, we demonstrate that MR. SID can accurately estimate spatial filters, connectivity graphs, and time-courses from native resolution functional magnetic resonance imaging data. MR. SID therefore enables big time-series data to be analyzed using standard methods, readying the field for further generalizations including non-linear and non-Gaussian state-space models.