1270
Symposium
High expectations have been placed on understanding the brain changes associated with mental ill-health. Neuroimaging has emerged as a promising non-invasive way to identify and monitor these changes. However, the existing evidence on how neurobiological information can contribute to clinical practice is limited. If we can predetermine who benefits from specific therapies, we have the potential to prevent the chronicity of mental health disorders, improving treatment (time) and quality of life. With these objectives in mind, the symposium aims to: (1) describe how to apply normative models of brain morphometry and age to gain insights into (mental) health and disease, (2) evaluate whether normative models of brain morphometry and age can prognosticate treatment responses and track treatment outcomes, a crucial step toward their translation into future clinical practice, and (3) contribute to our understanding of whether normative models of brain morphometry and age are living up to its promise in the field of psychiatry. The analysis of high-dimensional neuroimaging data demands advanced modeling to harness its predictive power effectively. For this symposium, we have gathered four internationally recognized experts that have examined state-of-the-art methods to derive metrics from different imaging data modalities in adult and youth populations. During our discussions, we will present our findings on age-related brain signatures within the context of pharmacological interventions and non-pharmacological interventions. Importantly, we will delve into whether individualized approaches have the potential to enhance clinical outcomes, offering new insights into the role of neuroimaging in psychiatry research.
Symposium attendees will be provided with the knowledge to:
1. Understand the concept of normative modelling and brain age, and basic principles of machine learning prediction using neuroimaging phenotypes
2. Understand how normative models derived from large-scale training data can be applied to smaller samples to gain insights of age-related brain signatures in health and (psychiatric) disease
3. Understand past, current, and future challenges for clinical translation
Our symposium is designed for neuroscientists and methodologists keen to explore recent advances in brain morphometry and cutting-edge machine learning models, as well as for (clinical) researchers interested in applying publicly available models and algorithms to their local datasets with the goal of potential clinical translation.
Presentations
Over the past decades, a major goal in biological psychiatry has been to identify brain biomarkers of psychiatric disorders. However, numerous barriers, including significant clinical and biological heterogeneity, have hindered the progress toward identifying clinically relevant neuroimaging markers and translating neuroimaging findings into clinical practice. Traditional approaches often overlook this variability, treating clinical cohorts as homogeneous. In this talk, I will introduce normative modeling of brain morphometry, a statistically rigorous framework for making inferences at the individual level. Unlike traditional group-level analyses, normative modelling captures variability by estimating a normative distribution for a given phenotype (e.g., brain volume) based on relevant demographic characteristics (e.g., age, sex). Individual-level deviations from these norms can then be measured to better understand clinical and biological variability without assuming shared patterns of pathology or predefined clusters. Over the past decade, various research groups have developed several tools and frameworks to obtain normative models of brain morphometry, such as the PCNToolkit, CentileBrain, and BrainCharts. I will outline the intuition behind these normative modeling approaches/methods, emphasizing its potential to advance our understanding of the neurobiology of psychiatric disorders and improving treatment outcomes. I will then discuss some of the most recent applications across a variety of clinical samples (neurodegenerative, neurodevelopmental, and psychiatric disorders) in different clinical contexts (treatment outcomes, illness stage, transdiagnostic samples) using various neuroimaging correlates (gray matter volume, cortical thickness, fMRI, surface area). I will conclude by illustrating how normative modelling encourages us to reframe variability as a feature rather than a limitation in psychiatric neuroimaging research by presenting recent work combining normative models of brain morphology with genetics to more precisely identify aspects of brain dysfunction that may have casual influences on psychiatric phenotypes, and demonstrating how multiscale approaches might be useful to improve our understanding of the neurobiology of psychiatric disorders.
Presenter
Ashlea Segal, Yale University New Haven, CT
United States
Neuroimaging-derived brain age metrics may potentially aid in treatment selection. This talk explores the utility of brain age gaps (BAG) in affective disorders - specifically depression, anxiety, and bipolar disorder - where symptom overlap, treatment response heterogeneity, and different treatment options (e.g., psychotherapy, different medications and TMS and ECT) pose clinical challenges. Higher BAG could potentially predict differential treatment outcomes, e.g., greater benefits from medication compared to psychotherapy and increased risk of cognitive side effects from interventions like ECT. I will begin by presenting on the relevance of studying brain age in affective disorders, followed by an overview of current methodological approaches to BAG estimation. This will include an examination of various models and input features derived from structural neuroimaging (MRI and PET) as well as functional neuroimaging (fMRI and EEG). I will conclude this part by addressing neuro-assessment (i.e., cognition), and exploring how brain age models could be scaled and implemented in future routine practice. I will also look at how different multivariate neuroimaging approaches may differentially predict various symptom domains and clinical outcomes, highlighting the importance of matching brain age estimation methods to specific treatment targets and looking beyond mood improvements. The last focus is the often-overlooked role of BAG in predicting treatment side effects, particularly cognitive dysfunction and tolerability across different interventions. Finally, I will present findings on the prognostic potential of BAG in the treatment of depression and anxiety, including both pharmacological (e.g., SSRIs) and non-pharmacological interventions (e.g., running therapy).
Presenter
Kristian Jensen, Copenhagen University Hospital Copenhagen
Denmark
Large-scale studies have linked a higher brain age gap (BAG) in adults to poorer mental health and unfavorable lifestyle factors, but its practical utility in clinical settings remains uncertain. While it is suggested that individual variations in the BAG reflect an ongoing process of neurobiological aging, current evidence is mostly based on cross-sectional data and there is a notable scarcity of studies examining repeated brain age measurements over time. Our understanding of how the BAG relates to health and disease is still evolving, and a key open question is whether the BAG is stable or dynamic. For example, it is crucial to determine whether recovery from psychiatric symptoms following treatment is accompanied by a corresponding beneficial change of the BAG. Such covariance is a necessary, but not sufficient condition for causality, and will add to our understanding of whether the BAG could be a useful target for treatment. In this talk, I will review the current state of the field regarding longitudinal BAG studies, with a particular focus on their application in clinical trials. I will explore potential (methodological) reasons for the lack of current evidence demonstrating dynamic changes in the BAG over time, both in general and specifically in response to treatment. Finally, I will end by presenting findings from two clinical trials with repeated scans (i.e., pre- and post-treatment) in which brain age was estimated using an algorithm trained on the largest and most diverse dataset to date (N = 53,542 individuals). This algorithm was applied to both young (15-26 years) and adult (21-66 years) individuals diagnosed with mood disorders—such as major depressive disorder, anxiety, and bipolar disorder —who received either pharmacological treatments (antidepressants, mood stabilizers, antipsychotics) or non-pharmacological treatments (running therapy).
Presenter
Laura Han, Amsterdam UMC Amsterdam, Noord-Holland
Netherlands
This presentation examines how combining Brain Age Gap (BAG) analysis with normative modelling of brain morphometry can enhance predictions of treatment responses in Major Depressive Disorder (MDD). Brain age provides a global quantitative representation of an individual’s brain health, but often lacks detailed regional information. In contrast, normative modelling offers precise regional brain deviations but lacks a standardized summary method across all regions. Together, these methods can provide a comprehensive and complementary assessment of brain health. Utilizing the CentileBrain Model, trained on a comprehensive dataset of over 37,000 individuals from 20 countries, we analyzed structural MRI data from the International Study to Predict Optimized Treatment in Depression (iSPOT-D). This dataset includes pre- and post-treatment scans of 224 MDD patients and 271 healthy controls. MDD patients were treated with one of three pharmacological treatments commonly prescribed as first-line medication treatments for depression. The model integrates extensive MRI metrics such as cortical thickness, surface area, and subcortical volumes to assess brain ages and regional deviations from the normative model. I will present our findings on both BAG and regional deviations for characterizing case-control differences, and predicting responses to the three treatments used in the study. I will discuss the implications of integrating brain age with normative models for clinical psychiatry. This approach not only refines our predictions but also provides deeper insights into the neurobiological mechanisms influencing treatment response. The presentation aims to show how these predictive tools can be used to tailor therapeutic interventions more precisely, potentially transforming treatment strategies based on individual age-related brain profiles.
Presenter
Yu-Chi Chen, University of Sydney Sydney
Australia