BackgroundCurrently, the diagnosis of schizophrenia is made solely based on interviews and behavioral observations by a trained psychiatrist. Technologies such as electroencephalography (EEG) are used for differential diagnosis and not to support the psychiatrist’s positive diagnosis. Here, we show the potential of EEG recordings as biomarkers of the schizophrenia syndrome. EEG (electroencephalography) differences between patients with schizophrenia (SCZ) and controls have been reported. Tasks used are complex and specialized, not necessarily resemble natural stimuli/ environment to which the brain is adapted. We tested if SCZ global cognitive deficits could be described by EEG features using an ecological and simple approach.MethodsWe recorded EEG while schizophrenia patients freely viewed natural scenes, and we analyzed the average EEG activity locked to the image onset. We compared occipital ERPs obtained from 11 subjects with SCZ and 9 aged-‐ matched healthy controls (HC) during free-‐ exploration of images. Image categories included Plain Gray, Pink Noise and Landscapes (n=10 each). ERPs locked to image onset were obtained from occipital electrodes ader ocular artifacts rejection (by ICA decomposition).ResultsWe found significant differences between patients and healthy controls in occipital areas approximately 500 ms after image onset. These differences were used to train a classifier to discriminate the schizophrenia patients from the controls. The best classifier had 81% sensitivity for the detection of patients and specificity of 59% for the detection of controls, with an overall accuracy of 71%. We observed a positive wave after NS (natural scenes) landscape image onset, with late differences between the SZ patients and HCs. After visual inspection of the ERPs from each area (frontal, central, parietal, and occipital), we found significant differences only in the occipital ERP. It had two positive peaks in the HCs but a reduced second peak in the SZ patients. The median ERP at 0.4–0.6 s after image onset for the HCs was 4.14 μ V and 1.55 μ V for the SZ patients. The patients had a significant decrease in their ERP amplitude compared to the HCs (p = 0.01, Z = −2.5, T = 82, WRS test). Only the occipital electrodes showed differences in this period with the NS images. No other differences between the HC and SZ groups were found at other locations or time periods.We found significant differences between HC and SZ groups at the occipital electrodes only for the NS. Neither gray (p = 0.29, Z = −1.06, T = 101, WRS test) nor pink noise images (p = 0.93, Z = −0.07, T = 114, WRS test) showed significant differences between the HCs and SZ patients at any group of electrodes at this or any other time period.With an accuracy of 71% we are able to classified subjects. We performed 1350 cross-‐validation leaving 4 subjects out (two SCZ and two controls). 70.5% of the subjects with schizophrenia were correctly detected.DiscussionThis study shows that EEG features can differentiate between SCZ and HC in a simple, instruction-‐free visual task. Differences in late potentials (>300 ms) and in more complex images suggests deficits in top-‐down (cognitive) rather than bottom-‐up (perception) mechanisms.These results indicate that EEG signals from a free-viewing paradigm discriminate patients from healthy controls and have the potential to become a tool for the psychiatrist to support the positive diagnosis of schizophrenia.
Read full abstract