Improved Alzheimer's disease versus frontotemporal lobar degeneration differential diagnosis combining EEG and neurochemical biomarkers

IntroductionDistinguishing between two of the most common causes of dementia, Alzheimer disease (AD) and frontotemporal lobar degeneration (FTLD), on clinical diagnostic criteria alone has poor diagnostic accuracy. Promising tools to increase the diagnostic accuracy of AD are the use of cerebrospinal fluid (CSF) biomarkers and electroencephalography (EEG), which is being investigated as a diagnostic tool for neurodegenerative brain disorders. In this study, we investigated the utility of EEG-based biomarkers in comparison and in addition to established neurochemical biomarkers in the AD versus FTLD differential diagnosis. MethodsOur study cohort comprised 37 AD and 32 FTLD patients, of which 19 AD and 11 FTLD had definite diagnoses. All these participants had CSF biomarker analyses resulting in four neurochemical (NCM) biomarkers (A{beta}1-42, T-tau, P-tau181 and Nf-L) and underwent 19-channel resting-state EEG. From the EEG spectra, dominant frequency peaks (DFP) were extracted in four regions resulting in four dominant frequencies (in left-temporal, frontal, right-temporal and parieto-occipital regions). This yielded a total of eight features (4 NCM + 4 EEG), which we used to train and test a classifier and assess the diagnostic value of the markers separately (using only the NCM or EEG subset) and combined. ResultsThe classification accuracies were much higher when training and testing on the definite subgroup than on the whole group. Furthermore, we found that the NCM feature subset allowed a better accuracy than the EEG feature subset, both when training and testing on the whole group (NCM 82% vs EEG 72%), as on the definite group only (92% vs 86%). Using both feature subsets together increased the accuracy to 86% in the whole group and 94% in the definite subgroups. Another interesting finding was the presence of two spectral peaks in a considerable number of patients in both groups. ConclusionCombining EEG with neurochemical biomarkers resulted in differential diagnostic accuracies of 86% in clinically diagnosed AD and FTD patients and of 94% in patients having a definite diagnosis. Furthermore, we found evidence that the slowing down of the dominant EEG rhythm might be a gradual appearance of a slow rhythm and fading out of the normal ground rhythm, rather than a gradual slowing down of the ground rhythm. Finally, we have discovered two differences in the occurrence of the dominant EEG frequency: people lacking a clear dominant peak almost all had definite AD, while people with two peaks more often had FTLD. These unexpected but interesting findings need to be explored further.