P 129 Quantitative EEG and neuropsychological tests to differentiate between Parkinson's disease patients and healthy controls with Random Forest algorithm. Issue 10 (October 2017)

• Record Type: Journal Article
• Authors: Chaturvedi, M.; Hatz, F.; Gschwandtner, U.; Meyer, A.; Cozac, V.V.; Bogaarts, J.G.; Roth, V.; Fuhr, P.
• Is Part Of: Clinical neurophysiology. Volume 128:Issue 10(2017:Oct.)
• Page Start: e391
• Page End: e392
• Publication Date: 2017-10
• Subjects: Neurophysiology -- Periodicals; Electroencephalography -- Periodicals; Electromyography -- Periodicals; Neurology -- Periodicals; 612.8
• DOI: 10.1016/j.clinph.2017.06.202
• Languages: English
• ISSNs: 1388-2457

Abstract

Abstract : Background and bbjectives: studies have shown that quantitative EEG (QEEG) and neuropsychological parameters are associated with Parkinson's disease (PD). We investigated the differences between PD patients and healthy controls (HC) in high-resolution QEEG measures, and analyzed the prediction accuracy. We also wanted to see if a combination of QEEG and neuropsychological factors could increase the prediction accuracy of the model in comparison with QEEG parameters alone. Methods: high-resolution 256-channel EEG were recorded in 66 PD patients and 59 HC. Neuropsychological assessment of the patients covered five cognitive domains: attention, working memory, executive functions, memory and visuo-spatial functions (18 cognitive tests). An average score for each domain was calculated along with an overall cognitive score, resulting in 6 additional scores. EEG data were processed to calculate the relative power in alpha, theta, delta, beta frequency bands across 10 regions of the brain. Alpha1/theta ratios were also calculated, resulting in a total of 77 QEEG frequency measures. Random Forest algorithm was applied to the data to check for change in prediction accuracy. Results: using the QEEG measures alone for classification, Area-under-the-Curve (AUC) value of 0.819 was obtained along with Positive and Negative predictive values (PPV, NPV) of 0.736 and 0.754, respectively. The 6 neuropsychological domain scores, when used alone, resulted in an AUC of 0.82, PPV of 0.71 and NPV of 0.8. On combining the QEEG measures and the 6 neuropsychological scores, an AUC value of 0.859 was obtained along with a PPV of 0.729 and NPV of 0.76. A slight increase in the AUC was observed on combining the QEEG and 6 neuropsychological measures, in comparison to using them alone while the PPV and NPV values did not have much difference. However, on combining the QEEG measures with all 24 available neuropsychological scores instead of using the average domain scores and overall cognitive scores alone, the AUC value increased to 0.88 while the PPV and NPV values increased to 0.785 and 0.8. Conclusion: QEEG measures can be useful in distinguishing Parkinson's disease patients from healthy controls with a considerable accuracy. This accuracy can be significantly improved by combining the QEEG measures with distinct neuropsychological test scores.