Qiyong Gong

With the AI technical development, psychoradiology is primed to assist clinician for improving the clinical care of the psychiatric patients.

Conor Liston

Shi Tang¹, Yanlin Wang¹, Yongbo Hu¹, Lu Lu¹, Lianqing Zhang¹, Xuan Bu¹, Hailong Li¹, Yingxue Gao¹, Lingxiao Cao¹, Xinyue Hu¹, Jing Liu¹, Xinyu Hu¹, Weihong Kuang², Qiyong Gong¹, and Xiaoqi Huang^1
1Huaxi MR Research Center (HMRRC), Functional and molecular imaging Key Laboratory of Sichuan Province, Department of Radiology, West China Hospital, Sichuan University, Chengdu 610041, China, Chengdu, China, ²Department of psychiatry, West China Hospital, Sichuan University, Chengdu 610041, China, Chengdu, China

Functional connectivity/network analyses using fMRI data have been applied to characterize diagnostic biomarkers in MDD. However, the association between brain connection and dimensional symptoms of this heterogeneous syndrome still remains unclear. In this work, we focused on first-episode and unmedicated MDD patients, firstly using unsupervised clustering analysis differentiated them into two subgroups on the basis of clinical features. Also, we compared the brain connectivity among subgroups plus healthy people. Then we used multivariate methods identified which clinical symptoms are significantly influenced by which brain connectivity. Our results may provide neurobiological mechanisms of MDD symptoms and serve as effective diagnostic biomarkers.

Yarong Wang¹ and Lei Wang^2
1Department of Radiology, The First Affiliated Hospital of Xi'an Jiaotong University, Xi'an, China, ²Department of nuclear medicine, Tangdu Hospital of Air Force Medical University of PLA, Xi'an, China

The objective of this study is to identify the heroin dependents undertaking stable methadone maintenance treatment (MMT patients) at high risk for opioid relapse prospectively. First, a self-defined addiction-related brain network was constructed with 10 hubs of several circuits associated with addiction and their degree centrality. Next, sixty male MMT patients was classified into different subgroups through grouping their addiction-related network into distinct neuronal activity patterns by K-means clustering algorithm.  By comparing relapse rate between subgroups with distinct network pattern, the one at high risk for relapse was identified. This finding implicated a novel strategy for improving MMT therapeutic effect.

Yael Jacob¹, Laurel S Morris¹, Kuang-Han Huang¹, Molly Schneider¹, Gaurav Verma¹, James W Murrough¹, and Priti Balchandani^1
1Icahn School of Medicine at Mount Sinai, New York, NY, United States

Currently, diagnosis for major depressive disorder (MDD) is largely reliant on self-reported symptoms. The ability to identify MDD without self-report is greatly needed. Implementing a graph-theoretical analysis on resting state fMRI (rsfMRI), we tested whether whole-brain network topology can be used as predictors of MDD using a machine learning algorithm. We found that MDD patients exhibit aberrant network centrality measures within the right hippocampus, supramarginal and parsopercularis. Using these as predictors in a machine learning algorithm we were able to classify MDD and controls with total accuracy of 81%, demonstrating the applicability of rsfMRI for diagnostics of MDD.

Chao Li^1,2,3, Yue Cui^1,2,3, Yongfeng Yang⁴, Jing Sui^1,2,3, Luxian Lv⁴, and Tianzi Jiang^1,2,3
1Brainnetome Center, Institute of Automation, Chinese Academy of Sciences, Beijing, China, ²National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing, China, ³University of Chinese Academy of Sciences, Beijing, China, ⁴Department of Psychiatry, Henan Mental Hospital, The Second Affiliated Hospital of Xinxiang Medical University, Xinxiang, China

Siamese Network is an artificial neural network that has been used in small sample sets multi-class classification studies. This study identified major depressive disorder (MDD), bipolar disorder (BD), and schizophrenia (SZ) based on combined gray matter, white matter and cerebrospinal fluid using Siamese Network. The participants included four groups: MDD (n = 102), BD (n = 44), SZ (n = 114), and healthy controls (n = 103). We found Siamese Network achieved improved performance than the multilayer perception network with different numbers of features. We achieved a classification accuracy of 46.06% and Macro F1 of 41.47% for this multi-class identification.