Predicting brain age using functional connectivity in attention-deficit/hyperactivity disorder
Poster No:
425
Submission Type:
Abstract Submission
Authors:
Yingxue Gao1, Zilin Zhou1, Weijie Bao1, Xinyue Hu1, Hailong Li1, Lianqing Zhang1, Xiaoqi Huang1
Institutions:
1Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University, Chengdu, China
First Author:
Yingxue Gao
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Co-Author(s):
Zilin Zhou
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Weijie Bao
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Xinyue Hu
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Hailong Li
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Lianqing Zhang
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Xiaoqi Huang
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University
Chengdu, China
Introduction:
Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental disorder with apparent roots in atypical brain development. The concept of 'brain age' derived from neuroimaging data holds the potential for a quantitative biomarker, reflecting the level of brain maturation along with neural development. Previous studies of brain age prediction all used structural MRI measures and found either older (Kaufmann et al., 2019) or younger appearing brains in individuals with ADHD (Kurth et al., 2022). But functional connectivity patterns are also crucial in characterizing brain maturity (Dosenbach et al., 2010) and exhibit abnormalities in ADHD (Gao et al., 2019). Therefore, in this study, we quantified the brain age using functional connectivity of brain resting-state networks (RSN) and a well-validated machine learning algorithm to enhance our understanding in abnormal developmental mechanism of ADHD.
Methods:
We recruited 360 male participants aged 7 to 18 years old [187 patients with ADHD and 173 typically developing controls (TDC)] from three sites. Resting-state fMRI and T1-weighted images were obtained on 3T MRI scanner and were preprocessed using the standardized pipeline in DPARSF. We used the Power atlas with 231 spherical ROIs that assigned to 11 large-scale RSN to construct functional connectivity matrices. The connectivity matrices were harmonized using Block-ComBat (Chen et al., 2022). Then, the within- and between-network connectivity were estimated for each subject.
The brain age prediction model was constructed using the linear support vector regression (SVR) based on obtained network connectivity features (Figure 1). The model was trained on TDC using a nested cross-validation (leave-one-out cross-validation for outer loop and 10-fold cross-validation for inner loop). Data from ADHD patients were used for testing using the model trained in all TDC subjects. We used the mean squared error (MSE), mean absolute error (MAE) and Pearson correlation coefficient (r) between the predicted and actual age to assess the prediction performance.
We calculated the brain age gap to represent the difference between the predicted age and the chronological age for each subject. A covariance analysis was applied to compare the difference in brain age gap between ADHD patients and TDC with chronological age as covariates.
The brain age prediction model was constructed using the linear support vector regression (SVR) based on obtained network connectivity features (Figure 1). The model was trained on TDC using a nested cross-validation (leave-one-out cross-validation for outer loop and 10-fold cross-validation for inner loop). Data from ADHD patients were used for testing using the model trained in all TDC subjects. We used the mean squared error (MSE), mean absolute error (MAE) and Pearson correlation coefficient (r) between the predicted and actual age to assess the prediction performance.
We calculated the brain age gap to represent the difference between the predicted age and the chronological age for each subject. A covariance analysis was applied to compare the difference in brain age gap between ADHD patients and TDC with chronological age as covariates.
Results:
The mean brain age gap for the TDC was 0.04 years (MSE = 5.14 years, MSE = 1.77 years, r = 0.370, p < 0.0001). Application of the brain age model to the patients with ADHD yielded a mean brain age gap of -0.06 years (MSE = 7.03 years, MSE = 2.23 years, r = 0.228, p = 0.0017). The top 10 features contributing the most to the brain age prediction model were shown in Figure 2B. There was no significant difference in the brain age gap between patients with ADHD and TDC (p = 0.869). However, when we divided the participants into child and adolescent groups, we found that the brain age gap of adolescents with ADHD (-2.06 years) was significantly lower than that of healthy adolescents (-1.24 years) (p = 0.021), but it did not significantly differ in children (p = 0.084, Figure 2D).
Conclusions:
This study represents the first attempt to predict brain age in patients with ADHD using functional network connectivity. We found that the brain age was significantly lower than chronological age in adolescents with ADHD, while no significant difference was observed in brain age gap between children with ADHD and TDC. This suggests that apparent developmental delays in brain functional networks of ADHD boys may not manifest until adolescence, providing new insights into the neurodevelopment mechanisms of ADHD.
Disorders of the Nervous System:
Neurodevelopmental/ Early Life (eg. ADHD, autism) 1
Lifespan Development:
Early life, Adolescence, Aging 2
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
Keywords:
Attention Deficit Disorder
Development
FUNCTIONAL MRI
1|2Indicates the priority used for review
Provide references using author date format
Dosenbach NU, Nardos B, Cohen AL, et al. Prediction of individual brain maturity using fMRI [published correction appears in Science. 2010 Nov 5;330(6005):756]. Science. 2010;329(5997):1358-1361.
Gao Y, Shuai D, Bu X, et al. Impairments of large-scale functional networks in attention-deficit/hyperactivity disorder: a meta-analysis of resting-state functional connectivity. Psychol Med. 2019;49(15):2475-2485.
Kaufmann T, van der Meer D, Doan NT, et al. Common brain disorders are associated with heritable patterns of apparent aging of the brain [published correction appears in Nat Neurosci. 2020 Feb;23(2):295]. Nat Neurosci. 2019;22(10):1617-1623.
Kurth F, Levitt JG, Gaser C, et al. Preliminary evidence for a lower brain age in children with attention-deficit/hyperactivity disorder. Front Psychiatry. 2022;13:1019546.