The Study Of Prosodic Parameters And Fist-Edge-Palm Test In Detecting Cognitive Impairment In Parkinson’s Disease

Background:Dementia is a frequent and tough non-motor symptom in Parkinson’s disease（PD）which takes a lot of doctor’s time to diagnose and manage it.It is hard to do cognitive test in community or PD specific outpatient department.Fist-edge-palm（FEP）test is a brief cognitive test which is developed for detecting Frontal temporal dementia（FTD）and is sensitive for detecting frontal cortical dysfunction.FEP has been studied in Alzheimer’s disease（AD）but seldomly studied in PD.Acoustic analysis of speech prosody in detecting mild cognitive impairment（MCI）and early dementia has some inspiring results which could lead to automatic machine diagnosis in AD but there are few studies in PD cognitive impairment detecting.Method:In part 1 study,we included continuous cases of Shanghai General Hospital.Assessed FEP,Mini-Mental State Examination（MMSE）and Clinical dementia rating scale（CDR）.Linear regression and area under curve（AUC）were used to study the sensitivity and specificity of FEP to diagnose cognitive impairment in PD.In part 2 study,we recruited30 patients with PD（PWP）with normal cognition（PD-NC）,30 PD patients with mild cognitive impairment（PD-MCI）and 30 PD patients with dementia（PDD）.Subjects were assessed by cognitive assessment battery.Gender was matched in three groups and age was from 65y to 75y.Acoustic prosodic data were taken from“Supermarket”reading task and“Cookie theft”picture describing task.Sound was recorded by Dr.Speech and Sound forge.Prosodic parameters including speech rate,articulation rate,pause position and pitch（F₀and F₀SD）was analyzed by Praat.One way ANOVA and Pearson correlation were used in SPSS 26 to analyze the characteristics and relationships of the parameters.Acoustic model for detecting cognitive impairment in PWP was done by Least absolute shrinkage and selection operator（LASSO）and leave one out cross validation（LOOCV）.In the data set,80%of the samples were randomly collected from the PD-NC and PD-MCI/PDD group to form the train set.The rest 20%datasets worked as test set.Present procedure was repeated for 15 times to evaluate the model’s robustness.Result:The AUC of FEP to detect cognitive impairment in PWP（PD-CI）is 0.763,the AUC of FEP to detect PD-MCI is 0.65,the AUC of FEP to detect PDD is 0.876.The cut-off point of 2-3 score to detect PD-MCI has a sensitivity of 71.4%,specificity of 51.9%;the cut-off point of 1-2 score to detect PDD has a sensitivity of 85.7%,specificity of 84.6%.The sum score of speech in MDS-UPDRS is more closely related to cognitive state in PWP than HY staging score,MDS-UPDRS-III score,PIGD score and AR score.The analysis of acoustic parameters showed that rhythm of speech prosody in picture describing task is significantly related to cognitive state and is not significantly related to motor symptom of PWP.Acoustic parameters model could detect PD-CI with an AUC of 0.853,could detect PD-MCI with an AUC of 0.724 and with an AUC of 0.92 to detect PDD.Conclusion:Present study shows that both FEP and acoustic parameters are efficient in detecting PDD,but they are not specific enough in detecting PD-MCI.Further study and parameters’optimization is needed.