558 - Multimodal EEG-MRI in the diagnosis of mild cognitive impairment with lewybodies

Abstract

Background:Differentiating mild cognitive impairment with Lewy bodies (MCI-LB) from mild cognitive impairment due to Alzheimer’s disease (MCI-AD) is challenging due to an overlap of symptoms. Quantitative EEG analyses have shown varying levels of diagnostic accuracy, while visual assessment of EEG may be a promising diagnostic method. Additionally, a multimodal EEG-MRI approach may have greater diagnostic utility than individual modalities alone.Research Objective:To evaluate the utility of (1) a structured visual EEG assessment and (2) a machine learning multimodal EEG-MRI approach to differentiate MCI-LB from MCI-AD.Method:300 seconds of eyes-closed, resting-state EEG from 37 MCI-LB and 36 MCI-AD patients were analysed. EEGs were visually assessed for the presence of diffuse, focal, and epileptiform abnormalities, overall grade of abnormalities and focal rhythmic delta activity (FIRDA). Random forest classifiers to discriminate MCI-LB from MCI-AD were trained on combinations of visual EEG, quantitative EEG and structural MRI features. Quantitative EEG features (dominant frequency, dominant frequency variability, theta/alpha ratio and measures of spectral power in the delta, theta, prealpha, alpha and beta bands) and structural MRI features (hippocampal and insular volumes) were obtained from previous analyses of our dataset.Results:Most patients had abnormal EEGs on visual assessment (MCI-LB = 91.9%, MCI-AD = 77.8%). Overall grade (Χ2 (73, 2) = 4.416, p = 0.110), diffuse abnormalities Χ2(73,1) = 3.790, p = 0.052, focal abnormalities Χ2 (73,1) = 3.113, p = 0.077 and FIRDA Χ2(73,1) = 0.862, p = 0.353 did not differ between groups. All multimodal classifiers had similar diagnostic accuracy (area underthe curve, AUC = 0.681 - 0.686) to a classifier that used quantitative EEG features only (AUC =0.668). The feature ‘beta power’ had the highest predictive power in all classifiers.Conclusion:Visual EEG assessment was unable to discriminate between MCI-LB and MCI-AD. However, future work with a more sensitive visual assessment score may yield more promising results.A multimodal EEG-MRI approach does not enhance the diagnostic value of quantitative EEG alone in diagnosing MCI-LB.(326 words)