Alignment of Induced EEG Oscillations Improves Analysis of Autism and ADHD Responses in Facial Categorization Task

Background: Children diagnosed with Autism Spectrum Disorder (ASD) often lack the ability to recognize and properly respond to emotional stimuli. Similarly, emotional deficits characterize children with Attention Deficit/Hyperactive Disorder (ADHD). Theory of Mind (ToM) Impairment may explain the presence of these deficits in patients with ASD, and may also be applicable to other conditions, including ADHD. Simultaneous evaluation of ToM impairment in both conditions may promote a better understanding of the emotional deficits in both conditions, and possible relationships between ASD and ADHD.

The deficits seen in ASD and ADHD may affect the induced electroencephalographic (EEG) gamma oscillations that occur following an emotional stimulus; however, analysis is complicated by the varying latencies of the oscillations, which are not fixed at a definite point in time post-stimulus. A more accurate analysis may be achieved by utilizing a data alignment method, which reduces the attenuation observed in the averaged EEG response. This improved analysis may be used to better evaluate changes in the induced gamma oscillations between ASD, ADHD, and control subjects.

Objectives:  To compare emotional recognition differences in ASD, ADHD, and control subjects by utilizing a data alignment technique programmed in MATLAB. 

Methods:  A forced-choice test was designed where subjects were asked to categorize a picture of a human face into one of two groups: male or female, angry or disgusted, and fearful or sad. EEG data was collected from ASD (n=10), ADHD (n=9) and control (n=11) subjects via a 128 channel EGI system. Data was then processed through a continuous wavelet transform and bandpass filter designed to isolate the gamma frequencies from 35-45 Hz. A MATLAB program was then used to align the trials within each subject x experimental condition x EEG site pairing (e.g. Subject A, Angry/Disgusted, P3) by maximizing the Pearson-product moment correlation coefficient between trials. The power of the induced gamma response for a 400 ms window was then calculated and compared between subject groups, and to an analogous power value obtained from the original, unaligned dataset.

Results: The main effect of alignment was significant in parietal and occipital topographies analyzed (F=995.89, p<0.001). A three-way Experimental Condition (Angry/Disgusted, Gender) x Alignment x Subject Group (ASD, ADHD, Controls) interaction was significant in the parietal and occipital topographies (F=2.68, p=0.030). This three-way interaction was most prevalent in the P3-P4 channels (F=3.43, p=0.048) and P7-P8 channels (F=4.304, p=0.025).

Conclusions:  Data alignment significantly reduced the attenuation of the averaged induced gamma oscillations, which increased the calculated induced gamma power. Alignment improved the differentiation of induced gamma power in the subject groups, revealing significant differences that would have gone unnoticed using traditional analysis methods. Group differences in the aligned dataset were most noticeable in the anger/disgust recognition test, whereas the power in the gender recognition test appeared to be more constant from group to group. This technique may be applied to future induced gamma studies in autism, and other neurodevelopmental conditions.