Zifan Jiang
BME PhD Proposal Presentation

Date: 2023-01-24
Time: 4:15pm - 5:15pm
Location / Meeting Link: https://zoom.us/j/8950734241?pwd=ckhrTEVVbnYxWWhZQmFwd2llR2dsUT09

Committee Members:
Gari D. Clifford, D.Phil (Advisor); Rishi Kamaleswaran, Ph.D.; Eva L. Dyer, Ph.D..


Title: Understanding and assessing neuropsychiatric disorders by modeling behavior and physiology with multimodal machine learning

Abstract:
The World Health Organization estimated in 2019 that 13% of the world population, or close to one billion people around the world, were living with a mental disorder, 5% of the world’s elderly population is affected by dementia, and 16% by Mild Cognitive Impairment, where most people do not have access to adequate care. Since the COVID-19 pandemic, those numbers have been rising rapidly, and the pandemic also continues to impede access to already underserved neuropsychiatric health services. In the United States alone, this crisis of neuropsychiatric disorders translates to an economic burden of about 600 billion $ every year. To reduce the high yearly cost and to delay the transition into often chronic or life-long neuropsychiatric conditions, it is critical to gain a better understanding and to provide accurate, fast, and accessible detection of those disorders to enable early and effective interventions. However, the current diagnosis and phenotyping of neurological and psychiatric disorders fail to satisfy this dire need. The emergence of digitally administered neuropsychological assessments has been one of the most promising approaches that may address the challenges encountered by traditional methods via objective, fast and noninvasive quantification of cognitive processes and physical status. In this work, I propose to improve the digitally administered neuropsychological assessments further. This work aims to get a better understanding and provide a better assessment of neuropsychiatric disorders by modeling and investigating both behavioral and physiological signals using multimodal machine learning. I will develop a bias-aware and privacy-preserving multimodal machine learning framework that assesses multiple cognitive processes with behavioral and physiological signals and their interactions. In aim one, I will investigate whether cognitive states, such as emotion, memory, and pain, can be reliably estimated with behavioral and physiological signals. In aim two, I will extend results from aim one and explore whether the detection and severity of various disorders, including psychiatric disorders like major depressive disorder, schizophrenia, and PTSD, and neurological disorders like dementia and mild cognitive impairment, could be classified or quantified. In aim three, I aim to answer whether simultaneously modeling multiple behavioral and physiological modalities brings improvement in the evaluations and a better understanding of cognitive states and neuropsychiatric disorders.