Rezaei Masoud, Zare Hoda, Hakimdavoodi Hamidreza, Nasseri Shahrokh, Hebrani Paria
Department of Medical Physics, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran.
Medical Physics Research Center, Mashhad University of Medical Sciences, Mashhad, Iran.
Front Hum Neurosci. 2022 Aug 18;16:948706. doi: 10.3389/fnhum.2022.948706. eCollection 2022.
The study of brain functional connectivity alterations in children with Attention-Deficit/Hyperactivity Disorder (ADHD) has been the subject of considerable investigation, but the biological mechanisms underlying these changes remain poorly understood. Here, we aim to investigate the brain alterations in patients with ADHD and Typical Development (TD) children and accurately classify ADHD children from TD controls using the graph-theoretical measures obtained from resting-state fMRI (rs-fMRI).
We investigated the performances of rs-fMRI data for classifying drug-naive children with ADHD from TD controls. Fifty six drug-naive ADHD children (average age 11.86 ± 2.21 years; 49 male) and 56 age matched TD controls (average age 11.51 ± 1.77 years, 44 male) were included in this study. The graph measures extracted from rs-fMRI functional connectivity were used as features. Extracted network-based features were fed to the RFE feature selection algorithm to select the most discriminating subset of features. We trained and tested Support Vector Machine (SVM), Random Forest (RF), and Gradient Boosting (GB) using Peking center data from ADHD-200 database to classify ADHD and TD children using discriminative features. In addition to the machine learning approach, the statistical analysis was conducted on graph measures to discover the differences in the brain network of patients with ADHD.
An accuracy of 78.2% was achieved for classifying drug-naive children with ADHD from TD controls employing the optimal features and the GB classifier. We also performed a hub node analysis and found that the number of hubs in TD controls and ADHD children were 8 and 5, respectively, indicating that children with ADHD have disturbance of critical communication regions in their brain network. The findings of this study provide insight into the neurophysiological mechanisms underlying ADHD.
Pattern recognition and graph measures of the brain networks, based on the rs-fMRI data, can efficiently assist in the classification of ADHD children from TD controls.
注意缺陷多动障碍（ADHD）儿童脑功能连接改变的研究一直是大量研究的主题，但这些变化背后的生物学机制仍知之甚少。在此，我们旨在研究ADHD患者和正常发育（TD）儿童的脑改变，并使用静息态功能磁共振成像（rs-fMRI）获得的图论测量方法，将ADHD儿童与TD对照组准确分类。
我们研究了rs-fMRI数据对未用药的ADHD儿童与TD对照组进行分类的性能。本研究纳入了56名未用药的ADHD儿童（平均年龄11.86±2.21岁；49名男性）和56名年龄匹配的TD对照组儿童（平均年龄11.51±1.77岁，44名男性）。从rs-fMRI功能连接中提取的图测量值用作特征。将提取的基于网络的特征输入RFE特征选择算法，以选择最具区分性的特征子集。我们使用来自ADHD-200数据库的北京中心数据训练和测试支持向量机（SVM）、随机森林（RF）和梯度提升（GB），以使用判别性特征对ADHD儿童和TD儿童进行分类。除了机器学习方法外，还对图测量进行了统计分析，以发现ADHD患者脑网络的差异。
使用最优特征和GB分类器，从未用药的ADHD儿童与TD对照组中分类的准确率达到了78.2%。我们还进行了枢纽节点分析，发现TD对照组和ADHD儿童中的枢纽数量分别为8个和5个，表明ADHD儿童的脑网络中关键通信区域存在紊乱。本研究结果为ADHD潜在的神经生理机制提供了见解。
基于rs-fMRI数据的脑网络模式识别和图测量可以有效地帮助将ADHD儿童与TD对照组进行分类。
