Diagnosis

In the current world of fast-paced data production, statistics and machine learning tools are essential for interpreting and utilizing the full potential of this data. This dissertation comprises three studies employing statistical analysis and Convolutional Neural Network models. First, the research investigates the genetic evolution of the SARS-CoV-2 RNA molecule, emphasizing the role of epistasis in the RNA virus’s ability to adapt and survive. Through statistical tests, this study validates the significant impacts of genetic interactions and mutations on the virus’s structural changes over time, offering insights into its evolutionary dynamics. Secondly, the dissertation explores medical diagnosis by implementing Convolutional Neural Networks to differentiate between lung CT-scans of COVID-19 and non-COVID patients. This portion of the research demonstrates the capability of deep learning to enhance diagnostic processes, thereby reducing time and increasing accuracy in clinical settings. Lastly, we delve into gravitational wave detection, an area of astrophysics requiring precise data analysis to identify signals from cosmic events such as black hole mergers. Our goal is to utilize Convolutional Neural Network models in hopes of improving the sensitivity and accuracy of detecting these difficult to catch signals, pushing the boundaries of what we can observe in the universe. The findings of this dissertation underscore the utility of combining statistical methods and machine learning models to solve problems that are not only varied but also highly impactful in their respective fields.

This study aimed to examine the relationship between licensed clinical social workers' (LCSWs) skepticism and knowledge about dissociative identity disorder (DID) and their accurate diagnosis of the disorder as moderated by specialized training about DID and/or postmaster’s experience with diagnosis and treatment of the disorder. The following research questions guided the study.
• RQ1a. Is there a relationship between LCSWs’ level of skepticism and level of knowledge about DID.
• RQ1b. Is there a relationship between LCSWs’ level of skepticism and accurate diagnosis of the disorder?
• RQ1c. Is there a relationship between LCSWs’ level of knowledge about DID and accurate diagnosis of the disorder?
• RQ2a. Does specialized training about DID affect the diagnostic accuracy of the disorder?
• RQ2b. Does post-master’s clinical experience with diagnosing and treating DID affect the diagnostic accuracy of the disorder?
Using a cross-sectional research design and informed by philosophical underpinnings of epistemology and skepticism and Kahneman’s model of diagnostic reasoning (Kahneman, 2011), the data for this study were collected via an online survey of randomly selected LCSWs (N=85) in Florida. The survey consisted of a diagnostic vignette with a very short answer (VSA) response, the Skepticism and Knowledge Scales (SKS) comprising 11 items that assess skepticism, six items that assess knowledge, and 13 distractor items (Hayes & Mitchell, 1994), and a demographic questionnaire with 12 items, two of which quantified specialized training about and clinical experience with DID.

Plasma-based diagnostics are ideal for detecting a variety of diseases because they offer a method of detection that is minimally invasive, readily available, and easy to use for monitoring patients as they progress through a disease or respond to treatment. The only serum marker for PDAC is CA19-9 which lacks specificity, has limited sensitivity, and is unreliable for early detection. It is therefore of great importance to develop a diagnostic that is viable for screening and early detection. Exosomal miRNA were determined via bioinformatics analyses and then examined in PDAC cell lines to identify markers with greatest potential. These markers were then examined in plasma from PDAC patients and control groups. Four markers, miR-93-5p, miR-339-3p, miR-425-5p, and miR-425-3p, emerged as the most viable biomarker panel with the ability to detect PDAC in 100% of the early stages (N=5) compared to CA19-9 which showed increased levels in only one patient with early stage PDAC. Additionally, the diagnostic has a specificity of 80% and a sensitivity of 94.7%, making it comparable to CA19-9, and may even be beneficial for use in conjunction with CA19-9.
A plasma-based diagnostic was also developed for multi-strain HIV-1 detection utilizing the loop-mediated isothermal amplification (LAMP) method. LAMP primers were developed against the integrase and vpr regions of the HIV-1 genome. They were tested first in cultured HIV samples and then examined for their ability to amplify HIV-1 subtypes A-G. The integrase primer set provided a reliable means of diagnosing all 55 strains and isolates in under 30 minutes, whereas vpr was inconsistent and exhibited high variability in detecting the HIV subtypes. Our limit of detection for B-subtype with integrase was 30 viral copies/reaction. This could provide the basis for a novel, point-of-care diagnostic for use in underdeveloped regions.

Two skills necessary for the execution of proficient calculation, retrieving arithmetic facts from memory and accessing number magnitude information, were studied in a group of patients diagnosed with Alzheimer's disease (AD), mild cognitive impairment (MCI), and healthy controls to try to elucidate the locus of impairment in AD-related calculation deficits. This was achieved through the use of an arithmetic production task and a number-matching task as measures of explicit and implicit retrieval of arithmetic facts, and a numerical Stroop task that assesses automatic access to number magnitude representation. AD patients, but not MCI patients, showed high response latencies and a high number of errors when performing multiplications in the production task, and reduced automatic retrieval of arithmetic task in the number-matching task. All participants showed the classic problem-size effect often reported in the mathematical cognition literature. Performance on the numerical Stroop task suggests that access to number magnitude information is relatively resistant to cognitive impairment. ... Results for the AD group are consistent with a pattern of preserved and impaired cognitive processes that might mediate the reported calculation deficits in AD.

In the medical science, the receiving operationg characteristic (ROC) curve is a graphical representation to evaluate the accuracy of a medical diagnostic test for any cut-off point. The area under the ROC curve (AUC) is an overall performance measure for a diagnostic test. There are two parts in this dissertation. In the first part, we study the properties of bi-Exponentiated Weibull models. FIrst, we derive a general moment formula for single Exponentiated Weibull models. Then we move on to derive the precise formula of AUC and study the maximus likelihood estimation (MLE) of the AUC. Finally, we obtain the asymptotoc distribution of the estimated AUC. Simulation studies are used to check the performance of MLE of AUC under the moderate sample sizes. The second part fo the dissertation is to study the estimation of AUC under the crossing model, which extends the AUC formula in Gonen and Heller (2007).

Though several clinical monitoring ways exist and have been applied to detect cardiac atril fibrillation (A-Fib) and other arrhythmia, these medical interventions and the ensuing clinical treatments are after the fact and costly. Current portable healthcare monitoring systems come in the form of Ambulatory Event Monitors. They are small, battery-operated electrocardiograph devices used to record the heart's rhythm and activity. However, they are not energy-aware ; they are not personalized ; they require long battery life, and ultimately fall short on delivering real-time continuous detection of arrhythmia and specifically progressive development of cardiac A-Fib. The focus of this dissertation is the design of a class of adaptive and efficient energy-aware real-time detection models for monitoring, early real-time detection and reporting of progressive development of cardiac A-Fib.... The design promises to have a greater positive public health impact from predicting A-Fib and providing a viable approach to meeting the energy needs of current and future real-time monitoring, detecting and reporting required in wearable computing healthcare applications that are constrained by scarce energy resources.

MicroRNAs (miRNAs) may serve as diagnostic and predictive biomarkers for cancer. The aim of this study was to identify novel cancer biomarkers from miRNA datasets, in addition to those already known. Three published miRNA cancer datasets (liver, breast, and brain) were evaluated, and the performance of the entire feature set was compared to the performance of individual feature filters, an ensemble of those filters, and a support vector machine (SVM) wrapper. In addition to confirming many known biomarkers, the main contribution of this study is that seven miRNAs have been newly identified by our ensemble methodology as possible important biomarkers for hepatocellular carcinoma or breast cancer, pending wet lab confirmation. These biomarkers were identified from miRNA expression datasets by combining multiple feature selection techniques (i.e., creating an ensemble) or by the SVM-wrapper, and then classified by different learners. Generally speaking, creating a subset of features by selecting only the highest ranking features (miRNAs) improved upon results generated when using all the miRNAs, and the ensemble and SVM-wrapper approaches outperformed individual feature selection methods. Finally, an algorithm to determine the number of top-ranked features to include in the creation of feature subsets was developed. This algorithm takes into account the performance improvement gained by adding additional features compared to the cost of adding those features.

Mild traumatic brain injuries (MTBI) are the leading type of head injuries with appreciable risque of sequelae leading to functional and psychological deficits. Although mild traumatic brain injuries are frequently underdiagnosed by conventional imaging modalities, rapidly evolving techniques such as diffusion tensor imaging (DTI) reveal subtle changes in white matter integrity as a result of head trauma and play an important role in refining diagnosis, therapeutic interventions and management of MTBI. In this dissertation we use diffusion tensor imaging to detect the microstructural changes induced by axonal injuries and to monitor their evolution during the recovery process. DTI data were previously acquired from 11 subjects, football players of age 19-23 years (median age 20 years). Three players had suffered a mild traumatic brain injury during the season and underwent scanning within 24 hours after the injury with follow-ups after one and two weeks. A set of diffusion indices, such as fractional anisotropy, axial, radial and mean diffusivity were derived from the diffusion tensor. Changes in diffusion indices in concussed subjects were analyzed based on two different approaches: whole brain analysis, using tract-based spatial statistics (TBSS) and region of interest analysis (ROI). In both approaches we use a voxelwise analysis to examine group differences in diffusion indices between five controls and three concussed subjects for all DTI scans. Additional statistical analysis was performed between control groups consisting of five and three non-injured players. Both analyses demonstrated that the MTBI group reveals increase in fractional anisotropy and decreases in transversal and mean diffusivity in cortical and subcortical areas within 24 hours after the injury.

Tinnitus is a conscious perception of phantom sounds in the absence of external acoustic stimuli, and masking is one of the popular ways to treat it. Due to the variation in the perceived tinnitus sound from patient to patient, the usefulness of masking therapy cannot be generalized. Thus, it is important to first determine the feasibility of masking therapy on a particular patient, by quantifying the tinnitus sound, and then generate an appropriate masking signal. This paper aims to achieve this kind of individual profiling by developing interactive software -Tinnitus Analyzer, based on clinical approach. The developed software has been proposed to be used in place of traditional clinical methods and this software (as a part of the future work) will be implemented in the practical scenario involving real tinnitus patients.