基于大脑形态学特征的机器学习方法预测帕金森病患者的抑郁症状

目的 研究脑形态学特征预测帕金森病(PD)患者抑郁症状严重程度的效果.方法 对106例PD患者进行头部MRI检查,采集患者的MRI结构像数据.将患者的数据分为训练集和测试集,采用Vertex-base和广义线性模型方法分析皮层厚度与抑郁症状严重程度的相关性,通过支持向量机方法进行症状预测.结果 本组患者的抑郁症状严重程...

Changes of brain morphology in patients with depressive disorder complicated by Parkinson's disease and prediction via machine learning

Objective To study the efficacy of brain morphological features in predicting the severity of depressive symptoms in Parkinson's disease(PD)patients.Methods The brain MRI was performed on 106 PD patients,and MRI structural phases were collected.The data of patients were assigned to training sets and test sets,and vertex-base analysis and generalized linear models were performed to investigate the correlation between cortical thickness and depression severity.And support vector machine was conducted to predict the symptom in test set.Results The changes in cortical thickness of multiple regions were related to depressive disorder severity in these patients.The Pearson correlation coefficient between the predicted Hamilton Depression Scale(HAMD)score and the real HAMD score was 0.394(P=0.0312),and the determination coefficient was 0.155.Conclusion The depressive disorder severity of PD patients is related to the cortical thickness,and these features could be used to predict depression symptom severity.