The onset of most mental health issues tends to occur during adolescence and is influenced by a complex blend of neurobiological and environmental factors. A new technique in manifold learning has been developed to model the interactions between the brain and its environment, significantly enhancing the detection of current mental health issues and improving predictions for future symptoms compared to traditional methods. The research, published in Biological Psychiatry: Cognitive Neuroscience and Neuroimaging by Elsevier, emphasizes the significance of integrating the development of the adolescent brain with its surrounding environment.
The onset of most mental health issues tends to occur during adolescence and is influenced by a complex blend of neurobiological and environmental factors. A new technique in manifold learning has been developed to model the interactions between the brain and its environment, significantly enhancing the detection of current mental health issues and improving predictions for future symptoms compared to traditional methods. The research published in Biological Psychiatry: Cognitive Neuroscience and Neuroimaging by Elsevier emphasizes the significance of integrating the development of the adolescent brain with its surrounding environment.
The need for more intricate and nuanced models of human biology and behavior, particularly in understanding the emergence of mental health symptoms, is becoming increasingly apparent. Despite acknowledging this challenge, much of the research continues to examine the brain and environmental factors separately or in simple, linear terms.
May I. Conley, MS, MPhil, a PhD candidate at Yale University’s Department of Psychology and co-lead author, states, “For a long time, developmental scientists have struggled to test theories that, in many ways, are hiding in plain sight. From neighborhoods to families, we understand that young people’s experiences in their environments and their neurobiology both impact emotional and behavioral growth. However, past methods have failed to accurately capture the complexity of these interactions.”
To tackle this issue, the researchers employed manifold learning, a set of algorithms designed for revealing hidden structures in high-dimensional biomedical data such as functional magnetic resonance imaging (fMRI). They created the exogenous PHATE (E-PHATE) algorithm to explore brain-environment interactions. Utilizing data from the Adolescent Brain and Cognitive Development (ABCD) study, funded by the National Institutes of Health and other federal entities, they applied E-PHATE algorithms to study participants’ brain activation during emotional and cognitive tasks to forecast individual differences in cognition and emotional and behavioral symptoms, both at a specific point in time and over a longer duration.
A key discovery in the research was the impact of incorporating various environmental factors into the E-PHATE model. The researchers found increased correlations between brain activity and mental health symptoms when they included either neighborhood or family factors in E-PHATE. Moreover, when these factors were combined with others, the model continued to improve its representation, indicating that adding environmental context is essential, rather than merely increasing the number of variables tested. This reinforces the necessity of considering the various environments that youth experience alongside how their brains process information from those surroundings.
First author Erica L. Busch, MS, MPhil, a PhD candidate at Yale University, added, “I was thrilled to discover that the modeling principles I was working on for basic science questions could be adapted for clinical purposes, resulting in such remarkable outcomes and insights. This experience also showcased the invaluable nature of interdisciplinary collaboration. My colleague May Conley and her advisor Dr. Baskin-Sommers are experts in the biopsychosocial models of mental health, and together with my computational skills, we each played vital roles in defining the goals and methodologies of this project.”
This research highlights the potential clinical applications of new machine learning and signal processing methods while underscoring the intricate and multifaceted relationship between adolescent brains and environments concerning emotional and behavioral symptoms. The researchers introduce a general-purpose method applicable in both clinical and non-clinical settings.
Cameron S. Carter, MD, Editor-in-Chief of Biological Psychiatry: Cognitive Neuroscience and Neuroimaging at the University of California Irvine, remarked, “Years of developmental research indicate that both neurobiology and the surrounding context contribute to the formation of mental health symptoms. This study illustrates the effectiveness of computational techniques like manifold learning for modeling intricate developmental data across multiple modalities, showing great promise for advancing our understanding of adolescent emotional and behavioral issues.”
The current study introduces innovation along three primary dimensions:
- By defining both neural and environmental data as multivariate measurements.
- By viewing their interaction as nonlinear and lower-dimensional, echoing the complexity of real-world data.
- By enabling both hypothesis-driven and data-driven discoveries related to these signals.
Senior author Arielle Baskin-Sommers, PhD, from Yale University’s Department of Psychology, concluded, “It is crucial that our field enhances its ability to capture the intricate relationships between individuals and their environments. However, to accurately measure these relationships, new approaches are required to integrate various data types and assess their interactions within individuals. The methodology arising from this interdisciplinary research exemplifies how we can better understand these complex dynamics.”
