Ibanez Soria, D. | bioRxiv Science

bioRxiv2018

Hypoarousal non-stationary ADHD biomarker based on echo-state networks

Attention-Deficit Hyperactivity Disorder (ADHD) is a neurodevelopmental disorder characterized by inattention, hyperactivity and impulsivity. It is one of the most commonly diagnosed neurodevelopmental and psychiatric disorders of childhood and therefore presents a very high prevalence rate. However the high rate of ADHD misdiagnosis makes the discovery of neurophysiological ADHD biomarkers an important clinical challenge. This study proposes a novel non-stationary ADHD biomarker based on Echo State Networks to quantify EEG dynamical changes between low attention/arousal states (resting with eyes closed, or EC) and normal attention/arousal states (resting with eyes open, or EO). Traditionally, EEG biomarkers have revealed an increase in stationary power in the theta band along with a decrease in beta, with these frequencies largely accepted to be altered in the ADHD population. We successfully verify the hypothesis that measured differences between these two conditions are altered in the ADHD population. Statistically significant differences between a group of ADHD subjects and an aged-matched control population were obtained in theta and beta rhythms. Our network discriminates between EO/EC EEG regimes in the ADHDs better than in controls, suggesting that differences in EEG patterns between low and normal arousal/attention states are larger in the ADHD population.

bioengineering↗

bioRxiv2018

Deep learning using EEG spectrograms for prognosis in idiopathic rapid eye movement behavior disorder (RBD)

REM Behavior Disorder (RBD) is now recognized as the prodromal stage of -synucleinopathies such as Parkinsons disease (PD). In this paper, we describe deep learning models for diagnosis/prognosis derived from a few minutes of eyes-closed resting electroencephalography data (EEG) collected from idiopathic RBD patients (n=121) and healthy controls (HC, n=91). A few years after the EEG acquisition (4 {+/-} 2 years), a subset of the RBD patients eventually developed either PD (n=14) or Dementia with Lewy bodies (DLB, n=13), while the rest remained idiopathic. We describe first a simple convolutional neural network (DCNN) with a five-layer architecture combining filtering and pooling, which we train using stacked multi-channel EEG spectrograms. We treat the data as in audio or image classification problems where deep networks have proven highly successful by exploiting compositional and translationally invariant features in the data. For comparison, we study an even simpler deep recurrent neural network using three stacked Long Short Term Memory network (LSTM) cells or gated-recurrent unit (GRU) cells--with very similar results. The performance of these networks typically reaches 80% ({+/-}1%) classification accuracy in the balanced HC vs. PD-outcome classification problem. In particular, using data from a single EEG channel we obtain an area under the curve (AUC) of 87% ({+/-}1%) while avoiding spectral feature selection. The trained classifier can also be used to generate synthetic spectrograms to study what spectrogram features are relevant for classification, pointing to the presence of theta band bursts and a decrease of power in the alpha band in future PD or DLB patients compared to HCs. We conclude that deep networks may provide a key tool for the analysis of EEG dynamics even from relatively small datasets and enable the delivery of new biomarkers.

bioinformatics↗