Browsing by Author "Lin, Guang"

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    A-Optimal Subsampling For Big Data General Estimating Equations
    (2019-08) Cheung, Chung Ching; Peng, Hanxiang; Rubchinsky, Leonid; Boukai, Benzion; Lin, Guang; Al Hasan, Mohammad
    A significant hurdle for analyzing big data is the lack of effective technology and statistical inference methods. A popular approach for analyzing data with large sample is subsampling. Many subsampling probabilities have been introduced in literature (Ma, \emph{et al.}, 2015) for linear model. In this dissertation, we focus on generalized estimating equations (GEE) with big data and derive the asymptotic normality for the estimator without resampling and estimator with resampling. We also give the asymptotic representation of the bias of estimator without resampling and estimator with resampling. we show that bias becomes significant when the data is of high-dimensional. We also present a novel subsampling method called A-optimal which is derived by minimizing the trace of some dispersion matrices (Peng and Tan, 2018). We derive the asymptotic normality of the estimator based on A-optimal subsampling methods. We conduct extensive simulations on large sample data with high dimension to evaluate the performance of our proposed methods using MSE as a criterion. High dimensional data are further investigated and we show through simulations that minimizing the asymptotic variance does not imply minimizing the MSE as bias not negligible. We apply our proposed subsampling method to analyze a real data set, gas sensor data which has more than four millions data points. In both simulations and real data analysis, our A-optimal method outperform the traditional uniform subsampling method.
  • Thumbnail Image
    Item
    Feature Selection Techniques for a Machine Learning Model to Detect Autonomic Dysreflexia
    (Frontiers Media, 2022-08-10) Suresh, Shruthi; Newton, David T.; Everett, Thomas H., IV; Lin, Guang; Duerstock, Bradley S.; Medicine, School of Medicine
    Feature selection plays a crucial role in the development of machine learning algorithms. Understanding the impact of the features on a model, and their physiological relevance can improve the performance. This is particularly helpful in the healthcare domain wherein disease states need to be identified with relatively small quantities of data. Autonomic Dysreflexia (AD) is one such example, wherein mismanagement of this neurological condition could lead to severe consequences for individuals with spinal cord injuries. We explore different methods of feature selection needed to improve the performance of a machine learning model in the detection of the onset of AD. We present different techniques used as well as the ideal metrics using a dataset of thirty-six features extracted from electrocardiograms, skin nerve activity, blood pressure and temperature. The best performing algorithm was a 5-layer neural network with five relevant features, which resulted in 93.4% accuracy in the detection of AD. The techniques in this paper can be applied to a myriad of healthcare datasets allowing forays into deeper exploration and improved machine learning model development. Through critical feature selection, it is possible to design better machine learning algorithms for detection of niche disease states using smaller datasets.
  • Thumbnail Image
    Item
    Modeling the Potential of Treg-Based Therapies for Transplant Rejection: Effect of Dose, Timing, and Accumulation Site
    (Frontiers Media, 2022-04-11) Lapp, Maya M.; Lin, Guang; Komin, Alexander; Andrews, Leah; Knudson, Mei; Mossman, Lauren; Raimondi, Giorgio; Arciero, Julia C.; Mathematical Sciences, School of Science
    Introduction: The adoptive transfer of regulatory T cells (Tregs) has emerged as a method to promote graft tolerance. Clinical trials have demonstrated the safety of adoptive transfer and are now assessing their therapeutic efficacy. Strategies that generate large numbers of antigen specific Tregs are even more efficacious. However, the combinations of factors that influence the outcome of adoptive transfer are too numerous to be tested experimentally. Here, mathematical modeling is used to predict the most impactful treatment scenarios. Methods: We adapted our mathematical model of murine heart transplant rejection to simulate Treg adoptive transfer and to correlate therapeutic efficacy with Treg dose and timing, frequency of administration, and distribution of injected cells. Results: The model predicts that Tregs directly accumulating to the graft are more protective than Tregs localizing to draining lymph nodes. Inhibiting antigen-presenting cell maturation and effector functions at the graft site was more effective at modulating rejection than inhibition of T cell activation in lymphoid tissues. These complex dynamics define non-intuitive relationships between graft survival and timing and frequency of adoptive transfer. Conclusion: This work provides the framework for better understanding the impact of Treg adoptive transfer and will guide experimental design to improve interventions.
  • Thumbnail Image
    Item
    Regional 4D Cardiac Magnetic Resonance Strain Predicts Cardiomyopathy Progression in Duchenne Muscular Dystrophy
    (medRxiv, 2023-11-08) Earl, Conner C.; Jauregui, Alexa M.; Lin, Guang; Hor, Kan N.; Markham, Larry W.; Soslow, Jonathan H.; Goergen, Craig J.; Pediatrics, School of Medicine
    Background: Cardiomyopathy (CMP) is the leading cause of death in Duchenne muscular dystrophy (DMD). Characterization of disease trajectory can be challenging, especially in the early stage of CMP where onset and clinical progression may vary. Traditional metrics from cardiovascular magnetic resonance (CMR) imaging such as LVEF (left ventricular ejection fraction) and LGE (late gadolinium enhancement) are often insufficient for assessing disease trajectory. We hypothesized that strain patterns from a novel 4D (3D+time) CMR regional strain analysis method can be used to predict the rate of DMD CMP progression. Methods: We compiled 115 short-axis cine CMR image stacks for n=40 pediatric DMD patients (13.6±4.2 years) imaged yearly for 3 consecutive visits and computed regional strain metrics using custom-built feature tracking software. We measured regional strain parameters by determining the relative change in the localized 4D endocardial surface mesh using end diastole as the initial reference frame. Results: We first separated patients into two cohorts based on their initial CMR: LVEF≥55% (n=28, normal cohort) and LVEF<55% (n=12, abnormal cohort). Using LVEF decrease measured two years following the initial scan, we further subclassified these cohorts into slow (ΔLVEF%≤5) or fast (ΔLVEF%>5) progression groups for both the normal cohort (n=12, slow; n=15, fast) and the abnormal cohort (n=8, slow; n=4, fast). There was no statistical difference between the slow and fast progression groups in standard biomarkers such as LVEF, age, or LGE status. However, basal circumferential strain (Ecc) late diastolic strain rate and basal surface area strain (Ea) late diastolic strain rate magnitude were significantly decreased in fast progressors in both normal and abnormal cohorts (p<0.01, p=0.04 and p<0.01, p=0.02, respectively). Peak Ea and Ecc magnitudes were also decreased in fast progressors, though these only reached statistical significance in the normal cohort (p<0.01, p=0.24 and p<0.01, p=0.18, respectively). Conclusion: Regional strain metrics from 4D CMR can be used to differentiate between slow or fast CMP progression in a longitudinal DMD cohort. These results demonstrate that 4D CMR strain is useful for early identification of CMP progression in patients with DMD. Clinical Perspective: Cardiomyopathy is the number one cause of death in Duchenne muscular dystrophy, but the onset and progression of the disease are variable and heterogeneous. In this study, we used a novel 4D cardiovascular magnetic resonance regional strain analysis method to evaluate 40 pediatric Duchenne patients over three consecutive annual visits. From our analysis, we found that peak systolic strain and late diastolic strain rate were early indicators of cardiomyopathy progression. This method offers promise for early detection and monitoring, potentially improving patient outcomes through timely intervention and management.