41656 research outputs found
Sort by
Machine Learning Based Medical Ultrasound Image Classification and Grad-Cam Interpretation
Get PDFThis work takes a step in creating a diagnostic tool for the classification decision process of Achilles tendinopathy using ultrasound images. An attention-based multiple instance learning model is developed to classify the images. Typically, doctors capture multiple ultrasound images of the Achilles tendon during a study to determine a complete diagnosis. Multiple instance models adopt this behavior by providing a single label for a set of instances (images). The images are grouped into ”bags” at the study level and passed into the model. The MIL model then uses its attention property to assign an importance score to each image to make its final classification decision. While all bags of images are composed of ultrasound images of the Achilles tendon, its important to note that different ultrasound machines produce images with different qualities. In the diagnostic process, it is important to ensure the model can be used on ultrasound images captured from any machine. As a result, this work explores the robustness of a multiple instance learning model to different image qualities.
To interpret the classification decision of the model, a post-processing step is explored to explain the decision-making process of the model. Using Gradient-weighted Class Activation Mapping, the classification decision of the model is visualized via a heatmap. Being able to interpret the decision-making process gives doctors confidence that the diagnostic classification system is accurate
Effectiveness of Electrosynthesized Hydrogen Peroxide in Urease Inactivation for Stabilization of Source-separated Urine to Recover Urea
Get PDFThe practice of separating urine at a building scale, rather than mixing with domestic wastewater to be treated at water resource recovery facilities, allows for concentrated urine to be directly treated with the aim of resource recovery. However, source separation and treatment present challenges during separation, storage, and transport. This study evaluated the use of hydrogen peroxide-producing electrochemical cells, which represent a technology that uses electricity as an input, to stabilize urine through cathodically produced peroxide to enable downstream water and nutrient recovery. The electrochemical cells have been previously shown to stabilize urine with electrochemically produced peroxide as an effective biocide that ceases enzymatic and biological activity in source-separated urine at concentrations in a few hundred to thousands of mg/L. Electrochemically treated urine with residual peroxide may be stable for a longer duration, thus allowing for processes for both water and urea recovery to follow. The effectiveness of urease inactivation by the electrochemically produced peroxide is important to understand, as it will form the basis for scaling up the technology and determining process operational parameters. The concentration of peroxide collected in urine is affected by: 1) the retention time of urine within the cell, and 2) the magnitude of the current density applied to the cathode, where peroxide is produced. Thus, this research evaluates the effect of these two parameters via systematic use of operational conditions to determine the effectiveness of peroxide at inactivating urease. These two conditions can then be tuned for a given expected concentration of urease in the urine. The results presented confirm iv that the concentration of peroxide is directly related to the effectiveness of inactivating urease, where the amount of urease present corresponds with the concentration of peroxide needed for effective stabilization. Additionally, the higher the amount of current density applied through the electrochemical cell, the better the urea stabilization. The challenge of increased retention time in the urine processing system that comes with storage and transportation was further studied through introducing urease into stabilized urine after different storage times, and it was determined that the longer the urine is stored, the more the current density that will be necessary, since the electrochemically produced peroxide will start to degrade. These results thus provide further understanding of the process that will lead to the development of large-scale treatment processes
An Empirical Analysis of the Pitch Clock on Major League Baseball
Get PDFThe pitch clock was instituted in baseball to speed up the pace of play and make the game more watchable. While it has achieved this effect, it may have affected other areas of the sport, like the balance between offense and defense and pitching injuries. This analysis explores the results of a difference-in-difference regression comparing MLB and AAA 2015-2024 seasons. Overall, it finds that pitching statistics were not significantly affected, including elbow injuries. This includes games, ERA, and times a batter is hit by a pitch. In fact, the number of home runs and walks both went down significantly by four home runs and eleven walks per season. However, batting statistics generally suffered. Batting average, OPS, and home runs all went down significantly, by .004, .046, and 1.648 per year, respectively. This shows that batters were negatively affected more than pitchers were
Revolutionizing Digital Privacy Education for Older Adults: Enhanced Interventions and AI-Assisted Learning Strategies
Get PDFAs older adults increasingly engage with digital platforms, they face unique privacy risks stemming from limited digital literacy, reduced trust in AI technologies, and constrained access—especially in rural or underserved communities. While digital tools offer benefits like social connection and information access, current privacy education efforts often neglect the needs of older adults. This dissertation addresses this gap by developing, testing, and refining digital privacy education interventions tailored for older adults, with a focus on trust, personalization, and AI-assisted learning.
Study 1 evaluates multiple instructional modalities across age groups, revealing older adults prefer structured videos and interactive tutorials, while younger adults favor chatbots and infographics. Studies 2a and 2b extend this work with experimental assessments of learning, enjoyment, and behavioral change, highlighting the importance of modality-personalization and user characteristics like rurality, motivation, and digital literacy.
Study 3a explores trust in AI vs. human instructors, revealing older and middle-aged women as the most trusted. Study 3b introduces trust transfer, demonstrating that AI instructors are better received when introduced by trusted humans—enhancing trust without overreliance.
Finally, Study 4 integrates an interactive AI assistant with a human-led introduction, showing significant gains in engagement, learning, and trust calibration. Results suggest older adults are not inherently resistant to AI, but require designs grounded in trust and usability.
Together, these studies offer theory-driven, evidence-based strategies for inclusive AI-assisted education. This work advances research in HCI, privacy literacy, and AI acceptance, supporting a future where older adults are empowered to engage safely and confidently with digital technologies
Replacement Optimization for Offshore Wind Turbine Farms
Get PDFThis dissertation is concerned with devising optimal replacement policies for offshore wind turbines with a focus on minimizing the costs associated with major component replacements and production losses due to downtime. Like their onshore counterparts, offshore wind turbines are subject to progressive degradation due to normal operations, as well as the influence of dynamic environmental conditions that influence their rate of degradation. Due to their proximity, wind farm turbines share common environmental conditions, as well as specialized maintenance resources. Their common exposure to the environment and need to share resources introduce both stochastic and economic dependence between the wind turbines. The primary objective of this research is to optimally prescribe the timing of major component replacements in offshore wind farms so that long-run expected costs are minimized.
First, we examine structured optimal replacement policies subject to these dependencies using a Markov decision process (MDP) framework. We find that a degradation-based threshold policy exists and is optimal. Structural properties of the cost function and optimal replacement policies are proved analytically, and these results offer practical guidance for operators seeking to balance cost-effectiveness while accounting for both stochastic and economic dependencies. While the MDP framework is helpful, it may be difficult (or impossible) to obtain optimal replacement policies for realistically-sized wind farms due to the curses of dimensionality. To address this computational challenge, near-optimal policies are obtained by devising an approximate linear programming (ALP) formulation that can be efficiently solved using a column generation algorithm. This approach facilitates computation of value function bounds and achieves near-optimal solutions exhibiting an optimality gap of approximately 1%. We also establish sufficient conditions under which the optimal policy can be retrieved from the approximate policy and establish a performance bound to evaluate its accuracy. Numerically, we explore how setup costs, the number of environment states, and the number of turbines in the wind farm influence replacement policies. Finally, the MDP framework is extended to consider the problem of jointly optimizing replacement and jack-up vessel (JUV) requesting decisions in a wind farm containing both a fixed-base and a floating offshore wind (FOW) turbine. Environment states in shallow and deep waters are modeled as a (weakly) positively correlated bivariate Markov chain. We formulate an infinite-horizon MDP model that captures stochastic dependence from the correlated degradation processes and economic dependence from shared JUV usage. We derive structural properties of the optimal policy and show that both replacement and JUV requesting decisions exhibit a threshold-type structure in the degradation levels. Additionally, we establish sufficient conditions for the policy to be monotone in the environment states. These results provide practical insights for coordinated maintenance planning across heterogeneous turbine types in offshore settings
Resilient Control Framework for EV Motor Drive System Subject to Cyber-Physical Security
Get PDFThe electric drive system (EDS) in electric vehicles (EVs) is one of the key safety-critical components. As IoT-enabled communication infrastructure for modern cyber-physical automotive systems continues to evolve, the importance of securing EDS against cyber threats along with physical faults, has become increasingly prominent. Among physical faults, power switches are particularly vulnerable and exhibit the highest susceptibility to open-circuit faults (OCFs). A compromised EDS, whether due to cyber threats or physical issues, can lead to excessive mechanical vibrations, increased thermal stress, fluctuations in electromagnetic torque, and elevated total harmonic distortion. These factors can substantially undermine traction control stability and jeopardize occupant safety. Most existing data-driven diagnostic schemes for motor drives solely rely on residual signals and are prone to misclassification due to overlapping data features associated with these anomalies. These methods concentrate on detecting either cyberattacks or OCFs independently, lacking a unified approach that can be universally applied to both types of anomalies. Moreover, following the detection and classification of cyber and physical anomalies (CPAs), it is essential to implement a robust control mechanism that not only localizes faults but also mitigates cyber threats. This mechanism should act in response to the detection signals generated by the CPAs, thereby enhancing the overall security and operational reliability of the EDS.
So to address these issues in this dissertation, an in-depth analysis is initially conducted on how these anomalies affect the electrical, mechanical, and thermal characteristics of the EDS. Subsequently, a unified physics-informed machine learning (PIML) approach is introduced, leveraging non-residual stator voltage features to effectively detect and differentiate between cyberattacks and OCFs. Moreover, the performance of the proposed PIML framework is compared against traditional residual-based methods using various machine learning models. When an OCF detection signal is identified through the PIML model, the corresponding resistive-loss profile for each semiconductor device, derived from the active thermal management-based model predictive control of the EDS, is used to accurately localize the OCF. Moreover, a reliability score quantification criterion is employed for machine learning model selection, which considers both traditional performance metrics and timing characteristics. Conversely, upon detecting a cyberattack, an AI-based reference tracking model, comprising signal regulation and prediction components, is deployed to mitigate its impact on the normal operation of the EDS. Thus, the integral components of detection, differentiation, localization, and mitigation of CPAs form a robust and comprehensive Resilient Control Framework (RCF). This framework is designed to enhance the reliability of EDS operations, ensuring optimal performance even under various abnormal conditions.
Finally, a trust-based monitoring system is proposed to evaluate the performance integrity of the EDS by emulating human-like decision-making. To achieve this, a model-based trust evaluation framework is developed to compute a trust score based on the severity and impact of potential cyber and physical disturbances. This framework aims to quantify the system’s performance under both normal and anomalous operating conditions
“They’re All Our Students!” Using Data-Based Individualization to Increase Reading Performance for Students With Disabilities in the Resource Setting
Get PDFStudents with Disabilities (SWDs) are not making progress in reading, and this is a problem that occurs across the nation. The National Assessment of Educational Progress scores showed that students without disabilities are increasing in reading growth, while SWDs showed a decrease in reading (Stevens et al., 2024; Vaughn & Wanzek, 2014). On average, SWDs are about three years behind their peers in reading (Gilmour et al., 2015). All students, including those with disabilities, need to learn how to read because it leads to success in academics and future careers as an adult (Sullivan et al., 2016). If SWDs do not make progress in reading, it can lead to many deficits in the future, including continued low performance in reading in future grades (Kearns et al., 2018). Poor reading performance can also be linked to dropping out of high school, low pay in a future job, and unemployment, as well as obesity (Danielson & Rosenquist, 2014; Kearns et al., 2018). Data-based Individualization (DBI) is a research-based process that has shown to be successful for students receiving special education services (Bruhn et al., 2023; Filderman et al., 2018; Jung et al., 2018; Powell et al., 2022). It requires special education teachers to come up with an intervention, monitor progress, analyze data, collaborate with DBI team members, and come up with a plan to adjust instruction if needed so the student can be successful (Danielson & Rosenquist, 2014; Lemons et al., 2014; Kearns et al., 2018; NCII, n.d.). DBI is an ongoing process and may take many cycles and extra work but has shown to give impressive results to SWDs (Lemons et al., 2014). This study shows how a school in South Carolina uses DBI to help solve a problem, increasing reading performance for SWDs in the resource setting. Data involving reading performance, student confidence, and teacher mindset were studied and analyzed. Pre- and post-tests were given for reading performance and confidence, while a teacher focus group took place that analyzed teacher mindset towards teaching SWDs. Data from this study showed that DBI was successful in helping SWDs achieve reading success, as well as helping to shift the mindset and confidence for teachers when teaching SWDs in their general education classroom setting
Studying the Impact of Co-teaching on Students With Disabilities Self-Efficacy in an Algebra 1 Course.
Get PDFABSTRACT
Students with disabilities (SWD) at Large High School in Rural School District have consistently underperformed in Algebra 1 as compared to their non-disabled peers and SWD across South Carolina. This performance gap contributes to the school’s designation as an Additional Targeted Support and Improvement (ATSI) site, putting it at risk for state intervention. This dissertation investigates whether implementing a co-teaching model in Algebra 1 can positively impact SWD students’ self-efficacy in mathematics, with a secondary focus on potential gains in academic performance in Algebra 1 on the state End of Course (EOC).
Grounded in improvement science methodology, the study uses a Plan-Do-Study-Act (PDSA) cycle to test the intervention’s effectiveness.Two co-taught Algebra 1 sections were created for the Spring 2025 semester. Each class included both a general education math teacher and a special education teacher who collaboratively planned and delivered instruction. Students enrolled in these classes completed 5 self-efficacy surveys every 4 weeks, while academic progress was measured using formative assessments and the EOC.
The study analyzed trends in student self-efficacy over time by analyzing student responses on the surveys given in the class. Benchmark assessments were given during the course and the results were compared to non-disabled peers at the site. Results were also evaluated against historical performance data. Findings suggest that co-teaching had a positive impact on student performance and perceptions in mathematics. Notably, students in co-taught settings reported increased feelings of self-efficacy, and the proportion of SWD achieving proficiency on the Algebra 1 EOC rose compared to previous years.
This research contributes to the field by providing evidence that co-teaching may be an effective, scalable intervention to improve both self-efficacy and performance in mathematics for SWD. The study highlights the importance of collaborative instruction, ongoing and intentional professional development, scheduled planning, and access to real-time student data as key drivers of successful co-teaching. These results offer a promising path forward not only for Large High School, but for similar rural and underperforming schools facing accountability systems. Future implementation and longitudinal study could further expound on the long-term effects of co-teaching on academic outcomes for students with disabilities
Transforming Childhood Adversity Into Executive Excellence: Impact on Firm Performance
Get PDFMany successful business leaders have overcome difficult childhoods, yet little research has examined how early hardships might actually contribute to leadership effectiveness. This study investigated whether CEOs who experienced childhood adversity can transform these difficult experiences into strengths that improve their companies\u27 performance.
The research surveyed 165 business owners across two groups: general business owners and automotive dealership owners. Participants answered questions about their childhood experiences, their ability to grow from trauma, and various psychological traits like grit, mental toughness, resilience, and growth mindset. Company performance data was also collected.
The findings challenge conventional wisdom that difficult childhoods necessarily harm leadership ability. Instead, the study found that childhood adversity can lead to better company performance when leaders experience what psychologists call post-traumatic growth —the positive changes that can emerge from processing difficult experiences. Leaders who developed strength, wisdom, and deeper relationships from their hardships were more likely to run successful companies.
The study also found that certain psychological traits help leaders transform adversity into growth. Traits like grit (perseverance toward long-term goals), mental toughness, resilience, and believing that abilities can be developed all helped leaders turn negative childhood experiences into positive outcomes. These effects were stronger for independent business owners compared to those operating within more structured corporate environments.
These results suggest a fundamental shift in how we think about leadership development. Rather than viewing difficult childhoods as barriers to success, organizations should recognize that many effective leaders have grown stronger through adversity. This has important implications for how companies select, develop, and support their leaders.
The research provides hope for individuals who experienced childhood difficulties and offers practical guidance for organizations. It suggests that leadership development programs should help executives process their past experiences constructively, focus on building psychological strengths, and create supportive environments where leaders can leverage their unique backgrounds for organizational benefit.
This work represents the first comprehensive study linking childhood adversity to business performance through growth mechanisms, offering a new perspective on what makes leaders effective
Exploring Teacher Decision-Making Based on 4K Children’s Emotional Observations and Narratives: A Multiple Case Study
Get PDFThe benefits of quality social-emotional instruction in early childhood programs are well established. The crux of delivering these benefits lies in teacher social-emotional decision-making in early childhood classrooms. While there is scant research on the topic, there is evidence that decentralizing social-emotional decision-making can help teachers achieve better outcomes by including families and other early childhood professionals, using data to inform decisions, and reviewing research. The current study is a multi-case qualitative analysis of two 4K teachers from a child development center in the South Carolina Upstate. It describes how the two subjects use social-emotional decision-making to support their students’ development. The researcher took observational notes on six children, three from each teacher, and interviewed them about their social-emotional knowledge, valuing their voice and extracting usable data from their responses. The researcher then interviewed the two teachers about the observations and child interviews to investigate how they would make social-emotional decisions to support each student’s development. The researcher used constant comparative analysis to conclude how the 4K teachers made social-emotional decisions. He derived a pyramid model that illustrates how stress level influences the teachers’ formality in decisions. At the base of the pyramid, stress is low, decisions affect all students principally because they involve preparing the learning environment, and they are formal because they are pre-planned. Stress is more present in the second tier, but decisions are less formal as they are characterized by more spontaneity and less decentralization. There is a targeted number of students in this tier. The peak of the pyramid includes a few children. It is characterized by high stress, which drives the level of formality in the decisions upward through more planning/less spontaneity, and more decentralization. Lastly, the researcher offers alternative models that include pivoting decision-making around student needs instead of stress and including a model that increases accountability measures