Predicting adult functional outcomes in childhood-onset attention-deficit/hyperactivity disorder using multimodal MRI and machine learning: A prospective follow-up study.
Authors
Affiliations (3)
Affiliations (3)
- Peking University Sixth Hospital, Peking University Institute of Mental Health, National Clinical Research Center for Mental Disorders, (Peking University Sixth Hospital), NHC Key Laboratory of Mental Health (Peking University), 51 Huayuan Bei Road, Haidian District, Beijing 100191, China; Beijing Key Laboratory for Big Data Innovative Application of Child and Adolescent Mental Disorders, Beijing, China.
- Peking University Sixth Hospital, Peking University Institute of Mental Health, National Clinical Research Center for Mental Disorders, (Peking University Sixth Hospital), NHC Key Laboratory of Mental Health (Peking University), 51 Huayuan Bei Road, Haidian District, Beijing 100191, China; Beijing Key Laboratory for Big Data Innovative Application of Child and Adolescent Mental Disorders, Beijing, China. Electronic address: [email protected].
- Peking University Sixth Hospital, Peking University Institute of Mental Health, National Clinical Research Center for Mental Disorders, (Peking University Sixth Hospital), NHC Key Laboratory of Mental Health (Peking University), 51 Huayuan Bei Road, Haidian District, Beijing 100191, China; Beijing Key Laboratory for Big Data Innovative Application of Child and Adolescent Mental Disorders, Beijing, China. Electronic address: [email protected].
Abstract
Attention-deficit/hyperactivity disorder (ADHD) is a childhood-onset neurodevelopmental disorder that often persists into adulthood, leading to extensive functional impairments for individuals with ADHD. However, the predictors for adult functional outcomes of children with ADHD remain unclear. This prospective follow-up study aimed to establish a predictive model using clinical characteristics and multimodal neuroimaging features in children with ADHD for functional outcomes in adulthood. Finally, 104 children with ADHD who accepted baseline magnetic resonance imaging (MRI) scan and clinical assessment completed followed up into adulthood, with a mean follow-up duration of 8.2 years. Functional outcomes assessed in adulthood adopted the Global Assessment of Functioning scale. Random forest models were applied to predict functional outcomes in adults with ADHD, including clinical characteristics, structural MRI (sMRI) and resting state functional MRI (rs-fMRI) as input features. Model performances were evaluated by area under the curve (AUC) and related metrics, with interpretability assessed using SHapley Additive exPlanations (SHAP). The model based on clinical information showed limited predictive performance (AUC = 0.589). Among unimodal models, rs-fMRI outperformed sMRI (AUC = 0.778 vs. 0.611). The multimodal features model demonstrated superior performance (AUC = 0.816). Predictors with the highest predictive value were low-frequency fluctuations (ALFF) in the left middle occipital gyrus, volumes of the right medial orbitofrontal cortex and right supramarginal gyrus, and degree centrality (DC) of the right cerebellum Crus I. Integrating childhood clinical information with multimodal MRI substantially improves prediction of adult functional outcomes in ADHD. Larger prospective study is needed to establish utility for risk stratification.