Doctoral Dissertations
Date of Award
5-2024
Degree Type
Dissertation
Degree Name
Doctor of Philosophy
Major
Mechanical Engineering
Major Professor
Xiaopeng Zhao
Committee Members
Subhadeep Chakraborty, Caleb Rucker, Eun Jin Paek, Devin Casenhiser
Abstract
Alzheimer's Disease (AD) and related dementias (PwADRD) often lead to memory issues, language difficulties, and cognitive impairments. These symptoms can cause social isolation, anxiety, depression, accelerating cognitive decline, and negatively impacting quality of life. Traditional diagnostic and rehabilitation methods, relying on Artificial Intelligence (AI) or clinical observations, sometimes lack transparency and explainability for the growing population affected by AD. My research aims to address these challenges by developing AI systems that analyze affective states and cognitive deficits in PwADRD interactions. This involves creating more precise diagnostic tools and rehabilitation methods, as well as trustworthy AI agents to improve PwADRD's quality of life.
Language serves as a crucial early indicator of AD, but the effectiveness of self-screening and understanding the impact of AD on speech are often compromised by patient-specific and external factors. To overcome these challenges, I designed an AI-integrated language screening tool using the referential communication task (RCT), a model emphasizing clear socio-communicative exchanges. This system uses universally recognized symbols in RCT to reduce efficacy-influencing factors. Through human-human experiments, we applied natural language processing (NLP) and visual-language (VL) models to RCT-derived speech, achieving a 98.7\% accuracy rate in distinguishing PwADRD from healthy individuals. Additionally, VL analysis revealed enhanced visual-semantic skills in healthy participants during RCT.
Integrating these insights, the system combines real-time communication with the VL model and automated screening through the NLP model. This AI-enhanced AD screening tool was further incorporated into a social robot for RCT, with its usability confirmed in a study involving college students. This approach not only promises to refine diagnosis and rehabilitation for PwADRD but also fosters the development of AI agents that can significantly enhance their quality of life.
Recommended Citation
Liu, Ziming, "Uniting Cognitive Models and AI: Early Alzheimer’s Screening through Language Analysis. " PhD diss., University of Tennessee, 2024.
https://trace.tennessee.edu/utk_graddiss/10137