Objective To explore the value of diffusion-tensor imaging （DTI） of different gray matter nuclei in the diagnosis and assessment of prodromal Parkinson's disease （pPD） under logistic regression model. Methods A total of 20 patients with pPD were collected as case group and 28 healthy people as control group （HC group）. All patients were examined by MRI plain scan and DTI. DSI studio was used to post-process the DTI images of all patients. Parameters （FA, MD, AD, RD） of basal ganglia, midbrain and brainstem of patients with pPD and HC group were automatically extracted and statistically analyzed. Logistic regression analysis was used to draw the Receiver Operating Characteristic （ROC） curve to analyze and compare the diagnostic efficacy of individual diagnosis and combined diagnosis of each parameter. And the correlation between the parameters of each group and MMSE score was analyzed. Results There were statistical differences in basal ganglia, midbrain and brain stem in PPD and HC group （P < 0.05）. Under Logistic regression equation model, when the optimal threshold was 0.63, the AUC of PPD was 0.964. The sensitivity and specificity of differential diagnosis were 85.0% and 100% respectively （P < 0.001）. There was correlation between DTI parameters and MMSE score in locus coeruleus in PPD group （P < 0.05）, and the correlation coefficient of FA value in locus coeruleus （r = -0. 646, P = 0.002） was the highest. Conclusions The lesions of basal ganglia, midbrain and brainstem correlated gray matter nuclei in pPD were extensive and different in degree. AD value of locus coeruleus was valuable for quantitative diagnosis of pPD, FA value of locus coeruleus could be used as a characteristic sensitive index for recognition of the severity of dysfunction in pPD patients. Multi-parameter combined diagnosis of DTI under Logistic regression model could effectively improve the diagnostic efficiency, and provide valuable reference for early diagnosis and intervention of pPD.