Most mental illness originates in youth: 50% of mental illness is diagnosed by the age of 14, and 75% by 24. Manifest illness emerges after a period of ~8-10 years. Remarkably, functional MRI can detect disrupted patterns of brain function associated with these vulnerable trajectories. However, the emergence of manifest mental illness reflects a complex and dynamic interplay of many biological, developmental, social and environmental factors that may be risky or protective, shared across illness categories, co-vary, and currently lack specificity. Moreover, the ability of fMRI and serial assessment to detect clinically-relevant shifts in dynamic risk trajectories of mental illness is poorly defined. Charting mental illness trajectories to optimize intervention is a critical need. A major obstacle is the identification of specific neural-behavioral risk pathways that shape vulnerable developmental trajectories into various categories of manifest mental illness. The long-term goal of my research is to develop computational approaches to advance our understanding of risk trajectories in the mental health of youth. Consequently, the overall objective of this study is to leverage advanced machine learning methods to illuminate dynamic trajectories of mental illness progression, to characterize sustained mental health and to quantify the incremental predictive value of serial and functional MRI assessment. My central hypothesis is that specific risk pathways that have specific brain functional correlates characterize discrete peri-adolescent trajectories in major mental illness categories. Further, that functional MRI metrics and serial assessment significantly improve case prediction. This study will leverage a new, unprecedented opportunity to dissect risk trajectories by applying deep machine learning to ?big data? from 20,000 diverse youth, including longitudinal data. I will characterize the behavior and specificity of dynamic risk trajectories and their functional neural correlates in 5 major peri-adolescent mental illness categories and determine the incremental predictive value of adding functional MRI to bio-psycho-social data and continuing assessment after age 10, the developmental inflection point for mental illness. Further innovation will accrue from the use of a cloud computing infrastructure to perform the research. My rationale is that elucidating specific risk trajectories and how to best monitor them in peri-adolescence will stimulate more refined prevention and early intervention strategies in youth mental health, known to improve outcomes and reduce resource use. Concomitantly, I will obtain capstone training in advanced machine learning methods, programming and cloud computing infrastructure, building on my KL2 training in machine learning, data science and the acquisition and analysis of fMRI and bio-psycho-social data. An interdisciplinary mentorship team of international experts in data science, developmental psychopathology, bioinformatics, epidemiology and biostatistics at the University of Washington will prepare me for my transition to independence as a physician-scientist working at the interface of computational science and population youth mental health, and as a child and adolescent psychiatrist specializing in neurodevelopmental disorders.
Charting mental illness trajectories in community populations of diverse youth to optimize early intervention is a critical need in population mental health. The goal of this study is to leverage advanced machine learning methods in unique, new ?big data? from 20,000 diverse youth to illuminate dynamic trajectories of mental illness progression, to characterize sustained mental health and to quantify the incremental predictive value of serial and functional MRI assessment. The resulting body of work will elucidate specific dynamic risk trajectories and suggest how to monitor them in peri-adolescence and stimulate more refined prevention and early intervention strategies in youth mental health, known to improve outcomes and reduce resource use. !
