Bayesian and influence function-based empirical likelihoods for inference of sensitivity in diagnostic tests.

In medical diagnostic studies, a diagnostic test can be evaluated based on its sensitivity under a desired specificity. Existing methods for inference on sensitivity include normal approximation-based approaches and empirical likelihood (EL)-based approaches. These methods generally have poor performance when the specificity is high, and some require choosing smoothing parameters. We propose a new influence function-based empirical likelihood method and Bayesian empirical likelihood methods to overcome such problems. Numerical studies are performed to compare the finite sample performance of the proposed approaches with existing methods. The proposed methods are shown to perform better in terms of both coverage probability and interval length. A real data set from Alzheimer's Disease Neuroimaging Initiative (ANDI) is analyzed.

Relative efficiency of using summary versus individual data in random-effects meta-analysis.

Meta-analysis is a statistical methodology for combining information from diverse sources so that a more reliable and efficient conclusion can be reached. It can be conducted by either synthesizing study-level summary statistics or drawing inference from an overarching model for individual participant data (IPD) if available. The latter is often viewed as the "gold standard." For random-effects models, however, it remains not fully understood whether the use of IPD indeed gains efficiency over summary statistics. In this paper, we examine the relative efficiency of the two methods under a general likelihood inference setting. We show theoretically and numerically that summary-statistics-based analysis is at most as efficient as IPD analysis, provided that the random effects follow the Gaussian distribution, and maximum likelihood estimation is used to obtain summary statistics. More specifically, (i) the two methods are equivalent in an asymptotic sense; and (ii) summary-statistics-based inference can incur an appreciable loss of efficiency if the sample sizes are not sufficiently large. Our results are established under the assumption that the between-study heterogeneity parameter remains constant regardless of the sample sizes, which is different from a previous study. Our findings are confirmed by the analyses of simulated data sets and a real-world study of alcohol interventions.

Early detection and staging of chronic liver diseases with a protein MRI contrast agent.

Early diagnosis and noninvasive detection of liver fibrosis and its heterogeneity remain as major unmet medical needs for stopping further disease progression toward severe clinical consequences. Here we report a collagen type I targeting protein-based contrast agent (ProCA32.collagen1) with strong collagen I affinity. ProCA32.collagen1 possesses high relaxivities per particle (r1 and r2) at both 1.4 and 7.0 T, which enables the robust detection of early-stage (Ishak stage 3 of 6) liver fibrosis and nonalcoholic steatohepatitis (Ishak stage 1 of 6 or 1 A Mild) in animal models via dual contrast modes. ProCA32.collagen1 also demonstrates vasculature changes associated with intrahepatic angiogenesis and portal hypertension during late-stage fibrosis, and heterogeneity via serial molecular imaging. ProCA32.collagen1 mitigates metal toxicity due to lower dosage and strong resistance to transmetallation and unprecedented metal selectivity for Gd3+ over physiological metal ions with strong translational potential in facilitating effective treatment to halt further chronic liver disease progression.

Investigation of Key Variables Impacting ICM Sensing Using Computer Simulations.

The insertable cardiac monitors (ICMs) used for diagnosing and managing abnormal heart activities can falsely detect heart rhythms due to respiration, device rotation/orientation, device position, device flipping, and body mass that alter the amplitudes and morphologies. The objective of this paper is to investigate the effects of these key variables on ICM sensing by using computer simulations and virtual human family. We observed in these simulations that sensing amplitudes can vary greatly depending on device flipping, orientation/rotation, and migration; change significantly due to respiration effect; and are most sensitive to it when body mass is large. Those findings support identification of the key variables impacting clinical false detections.

Transfer Functions of a Spinal Cord Stimulation System in Mixed Media and Homogeneous Media for Estimation of RF Heating during MRI Scans.

There is a risk for patients with spinal cord stimulation (SCS) implants that MRI-RF induced heating may cause harm to spinal cord (SC) or spinal nerves. The ISO/TS 10974 Clause 8, Tier 3 based evaluation of RF heating has been commonly used for active implantable manufactures through regulatory body world-wide. Generating TFs for the lead and SCS system is an essential component in Tier 3. The objective of this paper is to investigate the effects of media on transfer functions (TFs) and dissipated power near the distal electrode along clinical pathways of SCS implants. The simulation results indicate that SCS TF measurement and validation should be performed in HCM (0.47 S/m) since it has the closest match with SC mixed media. With the SC mixed media model, the compound tissue effect from tissues around the system was seen and TF follows that of the HCM TFs closely. These trends were confirmed through power calculation of TF equation by integrals of electrical fields along clinical SCS implant pathways in an adult male model, with a 2%-4% underestimation from HCM while LCM overestimates by 72%- 74%. This study indicates appropriate section of media for deriving TFs could lead to more clinically relevant RF heating predictions during MRI scans.

Mechanisms of Dorsal Root Ganglion Stimulation in Pain Suppression: A Computational Modeling Analysis.

OBJECTIVE: The mechanisms of dorsal root ganglion (DRG) stimulation for chronic pain remain unclear. The objective of this work was to explore the neurophysiological effects of DRG stimulation using computational modeling. METHODS: Electrical fields produced during DRG stimulation were calculated with finite element models, and were coupled to a validated biophysical model of a C-type primary sensory neuron. Intrinsic neuronal activity was introduced as a 4 Hz afferent signal or somatic ectopic firing. The transmembrane potential was measured along the neuron to determine the effect of stimulation on intrinsic activity across stimulation parameters, cell location/orientation, and membrane properties. RESULTS: The model was validated by showing close correspondence in action potential (AP) characteristics and firing patterns when compared to experimental measurements. Subsequently, the model output demonstrated that T-junction filtering was amplified with DRG stimulation, thereby blocking afferent signaling, with cathodic stimulation at amplitudes of 2.8-5.5 × stimulation threshold and frequencies above 2 Hz. This amplified filtering was dependent on the presence of calcium and calcium-dependent small-conductance potassium channels, which produced a hyperpolarization offset in the soma, stem, and T-junction with repeated somatic APs during stimulation. Additionally, DRG stimulation suppressed somatic ectopic activity by hyperpolarizing the soma with cathodic or anodic stimulation at amplitudes of 3-11 × threshold and frequencies above 2 Hz. These effects were dependent on the stem axon being relatively close to and oriented toward a stimulating contact. CONCLUSIONS: These results align with the working hypotheses on the mechanisms of DRG stimulation, and indicate the importance of stimulation amplitude, polarity, and cell location/orientation on neuronal responses.

THE SCREENING AND RANKING ALGORITHM FOR CHANGE-POINTS DETECTION IN MULTIPLE SAMPLES.

The chromosome copy number variation (CNV) is the deviation of genomic regions from their normal copy number states, which may associate with many human diseases. Current genetic studies usually collect hundreds to thousands of samples to study the association between CNV and diseases. CNVs can be called by detecting the change-points in mean for sequences of array-based intensity measurements. Although multiple samples are of interest, the majority of the available CNV calling methods are single sample based. Only a few multiple sample methods have been proposed using scan statistics that are computationally intensive and designed toward either common or rare change-points detection. In this paper, we propose a novel multiple sample method by adaptively combining the scan statistic of the screening and ranking algorithm (SaRa), which is computationally efficient and is able to detect both common and rare change-points. We prove that asymptotically this method can find the true change-points with almost certainty and show in theory that multiple sample methods are superior to single sample methods when shared change-points are of interest. Additionally, we report extensive simulation studies to examine the performance of our proposed method. Finally, using our proposed method as well as two competing approaches, we attempt to detect CNVs in the data from the Primary Open-Angle Glaucoma Genes and Environment study, and conclude that our method is faster and requires less information while our ability to detect the CNVs is comparable or better.

Modified screening and ranking algorithm for copy number variation detection.

MOTIVATION: Copy number variation (CNV) is a type of structural variation, usually defined as genomic segments that are 1 kb or larger, which present variable copy numbers when compared with a reference genome. The screening and ranking algorithm (SaRa) was recently proposed as an efficient approach for multiple change-points detection, which can be applied to CNV detection. However, some practical issues arise from application of SaRa to single nucleotide polymorphism data. RESULTS: In this study, we propose a modified SaRa on CNV detection to address these issues. First, we use the quantile normalization on the original intensities to guarantee that the normal mean model-based SaRa is a robust method. Second, a novel normal mixture model coupled with a modified Bayesian information criterion is proposed for candidate change-point selection and further clustering the potential CNV segments to copy number states. Simulations revealed that the modified SaRa became a robust method for identifying change-points and achieved better performance than the circular binary segmentation (CBS) method. By applying the modified SaRa to real data from the HapMap project, we illustrated its performance on detecting CNV segments. In conclusion, our modified SaRa method improves SaRa theoretically and numerically, for identifying CNVs with high-throughput genotyping data. AVAILABILITY AND IMPLEMENTATION: The modSaRa package is implemented in R program and freely available at http://c2s2.yale.edu/software/modSaRa. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Modeling dermatome selectivity of single-and multiple-current source spinal cord stimulation systems.

A recently published computational modeling study of spinal cord stimulation (SCS) predicted that a multiple current source (MCS) system could generate a greater number of central points of stimulation in the dorsal column (DC) than a single current source (1 CS) system. However, the clinical relevance of this finding has not been established. The objective of this work was to compare the dermatomal zone selectivity of MCS and 1 CS systems. A finite element method (FEM) model was built with a representation of the spinal cord anatomy and a 2 × 8 paddle electrode array. Using a contact configuration with two aligned tripoles, the FEM model was used to solve for DC field potentials across incremental changes in current between the two cathodes, modeling the MCS and 1 CS systems. The activation regions within the DC were determined by coupling the FEM output to a biophysical nerve fiber model, and coverage was mapped to dermatomal zones. Results showed marginal differences in activated dermatomal zones between 1 CS and MCS systems. This indicates that a MCS system may not provide incremental therapeutic benefit as suggested in prior analysis.

Computational modeling analysis of a spinal cord stimulation paddle lead reveals broad, gapless dermatomal coverage.

Spinal cord stimulation (SCS) is an effective therapy for treating chronic pain. The St. Jude Medical PENTA(TM) paddle lead features a 4 × 5 contact array for achieving broad, selective coverage of dorsal column (DC) fibers. The objective of this work was to evaluate DC activation regions that correspond to dermatomal coverage with use of the PENTA lead in conjunction with a lateral sweep programming algorithm. We used a two-stage computational model, including a finite element method model of field potentials in the spinal cord during stimulation, coupled to a biophysical cable model of mammalian, myelinated nerve fibers to determine fiber activation within the DC. We found that across contact configurations used clinically in the sweep algorithm, the activation region shifted smoothly between left and right DC, and could achieve gapless medio-lateral coverage in dermatomal fiber tract zones. Increasing stimulation amplitude between the DC threshold and discomfort threshold led to a greater area of activation and number of dermatomal zones covered on the left and/or right DC, including L1-2 zones corresponding to dermatomes of the lower back. This work demonstrates that the flexibility in contact selection offered by the PENTA lead may enable patient-specific tailoring of SCS.

A Bayesian hierarchical model of nontraumatic lower-extremity amputation rates.

A Bayesian hierarchical generalized linear model is used to estimate the risk of lower-extremity amputations (LEA) among diabetes patients from different counties in the state of Missouri. The model includes fixed age effects, fixed gender effect, random geographic effects, and spatial correlations between neighboring counties. The computation is done by Gibbs sampling using OPENBUGS. DIC (Deviance Information Criterion) is used as a criterion of goodness of fit to examine age effects, gender effect, and spatial correlations among counties in the risks of having LEAs. The Bayesian estimates are also shown to be quite robust in terms of choices of hyper-parameters.

Estimation of the optimal VV delay by an IEGM-based method in cardiac resynchronization therapy.

Determination of the optimal interventricular (VV) delay in cardiac resynchronization therapy currently relies on costly, time-consuming echocardiographic (ECHO) methods. This study evaluated the performance of a new intracardiac electrogram (IEGM)-based VV method compared to the aortic velocity time integral (AVTI) method of VV delay optimization. The study included two patient groups. Eleven patients enrolled by a single center in the Rhythm II ICD trial underwent prospective comparisons of the AVTI at the VV interval determined by the IEGM VV method versus the maximum AVTI at the echocardiographically determined optimal VV delay. In 61 patients enrolled in the RHYTHM VV trial, the same testing methods were compared retrospectively. In the prospective study, the maximum AVTI by the ECHO-based method (24.3 +/- 7.9 cm), was closely correlated with maximum AVTI by the IEGM-based method (23.9 +/- 7.9 cm; concordance correlation coefficient = 0.99; 95% confidence, lower limit of 98%. Likewise, in the retrospective analysis, the ECHO-determined maximum AVTI (22.1 +/- 8.2 cm) was similar to that determined by the IEGM-based method (20.9 +/- 8.3 cm; concordance correlation coefficient = 0.98; 95% confidence, lower limit of 97%).