Decoding Caregiver Mental Health from Children's Amygdala Functional Connectivity

[maegancalvert]

Research question(s)

1. Can we use a multivariate data-driven approach to define child amygdala resting-state functional connectivity signatures predictive of caregiver mental illness?

Description

This question is an extension of the work I completed in the 2021 ABCD-ReproNim class (see Development of Amygdala Functional Connectivity Issue #19).

Amygdala functional connectivity has been associated with mental and behavioral health problems in children, adolescents, and adults; however, the strength and direction of these associations are variable. Possible explanations for seemingly disparate findings maybe related to developmental effects, study (i.e., cross sectional, longitudinal) and task (task-fMRI, resting-state, structural) types, and underpowered studies paired with small effect sizes. Machine learning may provide researchers with a tool by which to distinguish individuals within groups and also generalize these findings with greater accuracy than univariate methods. This study aims to utilize support vector regression to learn features of amygdala functional connectivity to predict caregiver mental illness among children in the ABCD study. Additionally, we aim to evaluate feature importance of resting-state connectivity.

Tools and algorithms to be used

Pre-registration
Singularity/containers for reproducibility
Git and Github use for version control and open science practice
Python
Support Vector Regression
Shapley Additive exPlanations (SHAP)

Skills we could use help with (optional)

ABCD preferred pre-registration instructions.

Link to analysis plan (optional)

The analysis plan will be forth coming.

Suggested keywords/tags

Caregiver mental health, resting-state connectivity, machine learning