@misc{oai:repo.qst.go.jp:00072572,
 author = {Yahata, Noriaki and Morimoto, Jun and Hashimoto, Ryuichiro and Lisi, Giuseppe and Shibata, Kazuhisa and Kawakubo, Yuki and Kuwahara, Hitoshi and Kuroda, Miho and Yamada, Takashi and Megumi, Fukuda and Imamizu, Hiroshi and E., Náñez and Sr José and Takahashi, Hidehiko and Okamoto, Yasumasa and Kasai, Kiyoto and Kato, Nobumasa and Sasaki, Yuka and Watanabe, Takeo and 八幡 憲明 and 高橋 英彦},
 month = {Nov},
 note = {Autism spectrum disorder (ASD) is a major neurodevelopmental disorder characterized by deficits in reciprocal social interactions and communication, and by repetitive and restricted behaviors. Despite the significance of this disorder, its underlying neural mechanism remains unclear. Recently, using resting-state functional-connectivity magnetic resonance imaging (rs-fcMRI) techniques, attempts have been made to develop classifiers of ASD and typically developed (TD) individuals, and thereby to identify the abnormality of functional connections (FCs) in ASD. However, none of the previous classifiers has ever been successfully validated for an independent cohort because of over-fitting and the interferential effects of nuisance variables (NVs) such as measurement conditions and demographic distributions. Here, using a multiple-site data set from Japan, we developed an ASD classifier by focusing on abnormal FCs in ASD as revealed by rs-fcMRI [1]. To overcome the difficulties associated with over-fitting and the effects of NVs, we developed a novel machine-learning algorithm that identified a small number of abnormal FCs in ASD (0.2% of all FCs considered). The resultant classifier attained high accuracy for a Japanese discovery cohort [85%, area under the curve (AUC) = 0.93], and furthermore, demonstrated a remarkable degree of site generalization for two independent validation cohorts in the US ABIDE Project (75%, AUC = 0.76) and in Japan (70%, AUC = 0.77). The identified FCs predicted socio-communicative scores of ASD individuals and constituted the neural substrates of ASD (ADOS A; r = 0.44, P = 0.001). Collectively, we have established a reliable rs-fcMRI-based biomarker of ASD that elucidates a direct link between the underlying neural mechanisms and the behavioral characteristics of ASD. We also suggest that the selected FCs in the biomarker could be a novel target of therapies for ASD such as neurofeedback., Real-time Functional Imaging and Neurofeedback Conference 2017},
 title = {A small number of abnormal functional connections in the brain predicts adult autism spectrum disorder},
 year = {2017}
}