Advanced information technologies promise a massive influx of individual-specific medical data. These rich sources offer great potential for an increased understanding of disease mechanisms and for providing evidence-based and personalized clinical decision support. However, the size, complexity, and biases of the data pose new challenges, which make it difficult to transform the data to useful and actionable knowledge using conventional statistical analysis. The so-called “Big Data” era has created an emerging and urgent need for scalable, computer-based data mining methods that can turn data into useful, personalized decision support knowledge in a flexible, cost-effective, and productive way. The goal of my Ph.D. research is to address some key challenges in current clinical deci-sion support, including (1) the lack of a flexible, evidence-based, and personalized data mining tool, (2) the need for interactive interfaces and visualization to deliver the decision support knowledge in an accurate and effective way, (3) the ability to generate temporal rules based on patient-centric chronological events, and (4) the need for quantitative and progressive clinical predictions to investigate the causality of targeted clinical outcomes. The problem statement of this dissertation is that the size, complexity, and biases of the current clinical data make it very difficult for current informatics technologies to extract individual-specific knowledge for clinical decision support. This dissertation addresses these challenges with four overall specific aims: Evidence-Based and Personalized Decision Support: To develop clinical decision support systems that can generate evidence-based rules based on personalized clinical conditions. The systems should also show flexibility by using data from different clinical settings. Interactive Knowledge Delivery: To develop an interactive graphical user interface that expedites the delivery of discovered decision support knowledge and to propose a new visualiza-tion technique to improve the accuracy and efficiency of knowledge search. Temporal Knowledge Discovery: To improve conventional rule mining techniques for the discovery of relationships among temporal clinical events and to use case-based reasoning to evaluate the quality of discovered rules. Clinical Casual Analysis: To expand temporal rules with casual and time-after-cause analyses to provide progressive clinical prognostications without prediction time constraints. The research of this dissertation was conducted with frequent collaboration with Children’s Healthcare of Atlanta, Emory Hospital, and Georgia Institute of Technology. It resulted in the development and adoption of concrete application deliverables in different medical settings, including: the neuroARM system in pediatric neuropsychology, the PHARM system in predictive health, and the icuARM, icuARM-II, and icuARM-KM systems in intensive care. The case studies for the evaluation of these systems and the discovered knowledge demonstrate the scope of this research and its potential for future evidence-based and personalized clinical decision support.

Predicting the condition of a degrading dynamic system is critical for implementing successful control and designing the optimal operation and maintenance strategies throughout the lifetime of the system. In many situations, especially in manufacturing, systems experience multiple degradation cycles, failures, and maintenance events throughout their lifetimes. In such cases, historical records of sensor readings observed during the lifecycle of a machine can yield vital information about degradation patterns of the monitored machine, which can be used to formulate dynamic models for predicting its future performance. Besides the ability to predict equipment failures, another major component of cost effective and high-throughput manufacturing is tight control of product quality. Quality control is assured by taking periodic measurements of the products at various stages of production. Nevertheless, quality measurements of the product require time and are often executed on costly measurement equipment, which increases the cost of manufacturing and slows down production. One possible way to remedy this situation is to utilize the inherent link between the manufacturing equipment condition, mirrored in the readings of sensors mounted on that machine, and the quality of products coming out of it. The concept of Virtual Metrology (VM) addresses the quality control problem by using data-driven models that relate the product quality to the equipment sensors, enabling continuous estimation of the quality characteristics of the product, even when physical measurements of product quality are not available. VM can thus bring significant production benefits, including improved process control, reduced quality losses and higher productivity. In this dissertation, new methods are formulated that will combine long-term performance prediction of sensory signatures from a degrading manufacturing machine with VM quality estimation, which enables integration of predictive condition monitoring (prediction of sensory signatures) with predictive manufacturing process control (predictive VM model). The recently developed algorithm for prediction of sensory signatures is capable of predicting the system condition by comparing the similarity of the most recent performance signatures with the known degradation patterns available in the historical records. The method accomplishes the prediction of non-Gaussian and non-stationary time-series of relevant performance signatures with analytical tractability, which enables calculations of predicted signature distributions with significantly greater speeds than what can be found in literature. VM quality estimation is implemented using the recently introduced growing structure multiple model system paradigm (GSMMS), based on the use of local linear dynamic models. The concept of local models enables representation of complex, non-linear dependencies with non-Gaussian and non-stationary noise characteristics, using a locally tractable model representation. Localized modeling enables a VM that can detect situations when the VM model is not adequate and needs to be improved, which is one of the main challenges in VM. Finally, uncertainty propagation with Monte Carlo simulation is pursued in order to propagate the predicted distributions of equipment signatures through the VM model to enable prediction of distributions of the quality variables using the readily available sensor readings streaming from the monitored manufacturing machine. The newly developed methods are applied to long-term production data coming from an industrial plasma-enhanced chemical vapor deposition (PECVD) tool operating in a major semiconductor manufacturing fab. / text

BACKGROUND: There are many studies that have documented the increasing impact of stress and anxiety on an individual’s health and well-being. Everyone handles stress and anxiety differently with these conditions having varying physiological effects. To better recognize whether or not a person may need help in tackling these conditions, scholars have developed reliable validated instruments. Two prominent instruments that effectively assess stress and anxiety levels are the Perceived Stress (PSS) and Generalized Anxiety Disorder (GAD-7) scales. Furthermore, the literature has shed light onto the importance of the carotid intima-media thickness (c-IMT) measurement as a tool in evaluating the risk of cardiovascular disease. After all, heart disease has been reported as being the number one killer of Americans in recent years. The specific aims of this study were to determine if there was an association between perceived stress / generalized anxiety and c-IMT (static association), and also if higher levels of perceived stress / generalized anxiety result in a significant increase in c-IMT (changes over time). METHODS: Data was collected on about 700 participants comprised of employees from Emory University in Atlanta, Georgia. At baseline, six, twelve, and twenty-four months, the largest number of participants had completed and calculated their scores on the PSS and GAD-7 scales. At these same time points, participants had their IMT measured and recorded for the left and right common carotid arteries by a trained sonographer of the Emory Predictive Health Institute. Due to incomplete measurements and scores, only 228 participants were included for statistical analyses. This was still considered a suitable sample size given that this study only involved four measurement time points. Various statistical models were fitted for the data. All variables in the models were treated as categorical except for time which was continuous. Four separate models were built that included the variables perceived stress, age group, gender and time. In a similar manner, four models were built that included the variables generalized anxiety, age group, gender and time. AIC values, -2 log-likelihoods, partial correlations, p-values, and other relevant information were reported for these models. All statistical analyses were performed using the Statistical Analysis System (SAS), version 9.2. RESULTS: The mean c-IMT measurements for the Emory participants were higher than established normal ranges. A strong correlation existed between the PSS and GAD-7 two-year averages when treated as continuous variables (.7316, p <.0001). Likewise, a meaningful relationship existed when both scales were categorical (.4154, p < .0001). The analyses revealed that the left and right mean IMT measurements for the common carotid arteries modeled a linear trend with an unstructured covariance the best. The partial correlations for perceived stress and generalized anxiety revealed weak, but significant positive associations with the mean c-IMT measurement. Although the slope coefficients were not significant for perceived stress, an increase from below average to above average perceived stress level still resulted in an increase in mean c-IMT measurement. Conversely, mild generalized anxiety was found to be statistically significant in the regression model of the left mean c-IMT. This was after controlling for age group and gender. The p-value for mild generalized anxiety was 0.0258, and the slope coefficient was 0.04856. IMT measurements were consistently higher for males on both sides compared to females. They were also higher on the left side compared to the right. CONCLUSIONS: Failure to control anxiety could lead to c-IMT soaring to dangerous levels resulting in a myocardial infarction and/or cerebrovascular accident. Individuals should engage in healthy lifestyle practices that lower stress and anxiety levels to decrease the chances of cardiovascular disease. Based on this study’s findings, a person can certainly use their c-IMT readings, as well as their perceived stress and generalized anxiety scores, as indicators that lifestyle modifications may be needed.

Context. Foreign Accent Syndrome is a rare neurological disorder, where among other symptoms of the patient’s emotional speech is affected. As FAS is one of the mildest speech disorders, there has not been much research done on the cost-effective biomarkers which reflect recovery of competences speech. Objectives. In this pilot study, we implement the Emotional Temperature biomarker and check its validity for assessing the FAS. We compare the results of implemented biomarker with another biomarker based on the global distances for FAS and identify the better one. Methods. To reach the objective, the emotional speech data of two patients at different phases of the treatment are considered. After preprocessing, experiments are performed on various window sizes and the observed correctly classified instances in automatic recognition are used to calculate Emotional temperature. Further, we use the better biomarker for tracking the recovery in the patient’s speech. Results. The Emotional temperature of the patient is calculated and compared with the ground truth and with that of the other biomarker. The Emotional temperature is calculated to track the emergence of compensatory skills in speech. Conclusions. A biomarker based on the frame-view of speech signal has been implemented. The implementation has used the state of art feature set and thus is an unproved version of the classical Emotional Temperature. The biomarker has been used to automatically assess the recovery of two patients diagnosed with FAS. The biomarker has been compared against the global view biomarker and has advantages over it. It also has been compared to human evaluations and captures the same dynamics.

The U.S. health care system is considered costly and highly inefficient, devoting substantial resources to the treatment of acute conditions in a hospital setting rather than focusing on prevention and keeping patients out of the hospital. The potential for cost savings is large; in the U.S. more than $30 billion are spent each year on hospitalizations deemed preventable, 31% of which is attributed to heart diseases and 20% to diabetes. Motivated by this, our work focuses on developing centralized and distributed learning methods to predict future heart- or diabetes- related hospitalizations based on patient Electronic Health Records (EHRs). We explore a variety of supervised classification methods and we present a novel likelihood ratio based method (K-LRT) that predicts hospitalizations and offers interpretability by identifying the K most significant features that lead to a positive prediction for each patient. Next, assuming that the positive class consists of multiple clusters (hospitalized patients due to different reasons), while the negative class is drawn from a single cluster (non-hospitalized patients healthy in every aspect), we present an alternating optimization approach, which jointly discovers the clusters in the positive class and optimizes the classifiers that separate each positive cluster from the negative samples. We establish the convergence of the method and characterize its VC dimension. Last, we develop a decentralized cluster Primal-Dual Splitting (cPDS) method for large-scale problems, that is computationally efficient and privacy-aware. Such a distributed learning scheme is relevant for multi-institutional collaborations or peer-to-peer applications, allowing the agents to collaborate, while keeping every participant's data private. cPDS is proved to have an improved convergence rate compared to existing centralized and decentralized methods. We test all methods on real EHR data from the Boston Medical Center and compare results in terms of prediction accuracy and interpretability.