Facebook
X
LinkedIn
ForwardPredictive Models in Neuroimaging
Overview
Predictive models play a crucial role in advancing precision psychiatry by enabling data-driven insights into individual differences in brain function and clinical outcomes. Unlike traditional group-level analyses, predictive modeling leverages machine learning to forecast future states or identify risk patterns at the individual level. This approach has the potential to support early detection, tailor interventions, and improve prognosis in mental health by integrating neuroimaging biomarkers into clinical decision-making frameworks.
This 5-hour workshop provides an introduction to cutting-edge predictive modeling techniques applied to neuroimaging data, with a focus on understanding developmental trajectories in neuropsychopathology. Participants will explore how functional activation and resting-state functional connectivity derived from fMRI data can be harnessed to forecast cognitive and clinical outcomes in youth and adolescent populations.
Dataset/s:
Participants will work with the datasets below to explore how connectivity patterns relate to behavioral and clinical measures. Participants will use their own computers for this workshop’s hands-on segments. Prior to the workshop, they will receive instructions to download all of the software and data that will be needed during this segment of the workshop.
- 1000 Functional Connectomes Project (FCP); this resting-state dataset is publicly available at https://fcon_1000.projects.nitrc.org/ (and it can also be directly downloaded from CONN).
- Transdiagnostic Connectome Project (TCP); from this dataset, we will use an already-preprocessed/denoised/parcellated data and clinical measures available at https://openneuro.org/datasets/ds005237.
Please note that additional details, including a breakdown of the workshop structure, information about course instructors, and supplementary workshop materials, will be provided.
Pricing
MDs/Doctoral-Level Professionals/Other Professionals: $300.00
Students/Fellows/Trainees/Interns: $150.00
*Note: You DO NOT need to be registered for the 5th Annual Conference on Precision Psychiatry: Innovation to Implementation on Sept 25 - Sept 26 to register for a workshop.
This workshop is NOT for CME.
This workshop is NOT eligible for discounts.
Workshop Demo Session/Q&A:
Workshop registrants are invited to participate in an optional demonstration session on Tuesday, September 16, 2025, from 10:00 AM - 12:00 PM ET. The session will provide a demonstration on downloading and installing the necessary datasets and tools, along with a live Q&A to address participant questions.
Target Audience
This workshop is ideal for graduate students, researchers, and clinicians with a basic understanding of neuroimaging and an interest in machine learning applications in neuroscience. No prior experience in predictive modeling or fMRI is required.
Learning Objectives
- Understanding the principles of predictive modeling in the context of fMRI data.
- Exploring the role of functional activation and functional connectivity in forecasting developmental and clinical outcomes.
- Gaining hands-on experience with preprocessing and feature extraction from resting-state fMRI data.
- Applying predictive modeling techniques to real-world neuroimaging datasets using CONN.
Outcomes:
By the end of the workshop, participants will be equipped with foundational knowledge and practical skills to start incorporating predictive modeling techniques into their own neuroimaging research.
Contact:
For any questions on registration, please email mghctrprecisionpsych@mgh.harvard.edu. For workshop-specific questions, please email Dr. Alfonso Nieto Castañón at alfnie@bu.edu or Dr. Susan Whitfield-Gabrieli at sgabrieli@mgh.harvard.edu
Program
- Overview and Applications of predictive modeling in neuroimaging (1 hour)
Topics include the technical challenges of processing and interpreting resting state networks, such as the default mode network (DMN), the value of large-scale databases containing high-resolution functional and structural MRI-based and behavioral data, as well as the importance of single-subject deep phenotyping. Key concepts will be illustrated through research studies on the relationship between brain and behavior, with particular emphasis on clinical applications in precision psychiatry. Such applications include concepts such as neuroprevention: neuroimaging as used for early identification of mental illness; neuroprediction: neuroimaging as used to predict treatment response and stratification; and neuromodulation: neuroimaging as used for neurofeedback in precision network therapeutics.
- Predictive modeling methods in neuroimaging (1 hour)
- Functional activation and connectivity: concepts and extraction methods.
- Connectome-based predictive modeling: methods and applications at between- and within-subjects levels
- Feature selection, dimensionality reduction, cross-validation, permutation testing
- Challenges and considerations (overfitting, interpretability)
- Questions / Break (30 minutes)
- Hands-on practice with datasets (2.5 hours)
- Introduction. Using publicly available resting-state fMRI datasets (15 minutes)
- Preprocessing and feature generation with CONN (30 minutes). Preprocessing and denoising pipelines. Quality Control. Basic functional connectivity metrics.
- Using fc-MVPA to explore functional heterogeneity across subjects (30 minutes). Eigenpatterns and their interpretation. Associations with behavioral metrics (discuss issues of comorbidity/confounds/interpretation). Brain-wide association analyses.
- Building and evaluating predictive models using CONN’s multivariate tools (1 hour). Forward- and Backward- models. Cross-validation and model performance metrics. Interpretation of model parameters.
- Group discussion and Q&A (15 minutes)
Faculty
Dr. Alfonso Nieto-Castanon: Dr. Nieto-Castanon is a computational neuroscientist with a background in Cognitive and Neural systems (Ph.D. in Cognitive and Neural Systems, Boston University) and engineering (B.S./M.S. in Telecommunication Engineering, Universidad de Valladolid). His areas of specialization are modeling and statistics, brain imaging methods, and machine learning, and his main research interest is the understanding and characterization of human brain dynamics underlying mental function. He is the director of the Computational Neuroscience Research Lab and the lead developer of CONN
Dr. Susan Whitfield-Gabrieli: Dr. Susan Whitfield-Gabrieli is the Tommy Fuss Endowed Chair in Precision Psychiatry, Associate Director of the Center for Precision Psychiatry, Research Director for the Center for Comprehensive Healing, Director of the EPIC Lab, and a member of the Harvard Medical School faculty. Her primary mission is to understand the brain basis of neurodevelopmental and psychiatric disorders and to promote translation of this knowledge into clinical practice. Towards this end, she employs multimodal neuroimaging techniques to investigate the neural underpinnings of typical and atypical development as well as the pathophysiology of psychiatric disorders such as schizophrenia, depression, bipolar disorder, anxiety, and ADHD. Her ultimate mission is to discover biomarkers, derived from functional and anatomical brain networks, which may be utilized for (a) prediction of therapeutic response, geared towards precision medicine, (b) early detection, which potentiates early interventions designed to mitigate symptom progression, and (c) precision network therapeutics (e.g., real-time fMRI neurofeedback) with the hope of improving, or augmenting, currently available treatments.
Dr. Aaron Kucyi: Aaron Kucyi, Ph.D., is an Assistant Professor of Psychological & Brain Sciences at Drexel University, where he directs the Dynamic Brain and Mind Lab. His research employs functional neuroimaging, electrophysiology, and predictive modeling methods to study the neural bases of spontaneous thought, rumination, and attentional fluctuations in health and illness. Dr. Kucyi’s research has been funded by the National Institute of Mental Health (NIMH), the National Science Foundation (NSF), the Brain and Behavior Research Foundation (BBRF), and the Canadian Institutes of Health Research (CIHR).
Dr. Francesca Morfini: Francesca Morfini is a Postdoctoral Research Fellow in the Treatment & Etiology of Depression in Youth Laboratory at McLean Hospital and Harvard Medical School. Her research integrates multimodal neuroimaging and machine learning to advance the understanding and treatment of depression and anxiety in adolescents. As part of her research, she also investigates methodological factors that influence MRI study outcomes. She holds a Ph.D. in Cognitive Psychology from Northeastern University and Bachelor’s and Master’s degrees in Clinical Psychology and Neuroscience from San Raffaele University (Milan, Italy). She is also a licensed clinical psychologist in Italy.
Register/Take course
Price
This optional workshop supplements the 5th Annual Conference on Precision Psychiatry: Innovation to Implementation. Space is limited, and the cost is in addition to the registration fees to attend the conference.
This workshop is NOT eligible for further discounting than what is indicated above.
This workshop is NOT for CME.