EEG Findings Processed By Next Generation Artificial Adaptive Systems Can Perfectly Distinguish ASD Children from Typically Developing Children: A Proof of Concept Pilot Study

Thursday, May 12, 2016: 11:30 AM-1:30 PM
Hall A (Baltimore Convention Center)
E. Grossi1, M. P. Buscema2 and C. Olivieri1, (1)Autism Unit, Villa Santa Maria Institute Neuropsychiatric Rehabilitation Center, Tavernerio, Italy, (2)Semeion Research Centre of Sciences of Communication, Roma, Italy

To our knowledge, this is the first study that applies an artificial adaptive system to extract interesting features in computerized EEG that distinguishes ASD children from typically developing ones. The new system, named MS-ROM/I-FAST, belongs to the family of systems developed by The Semeion Research Institute in Rome. MS-ROM/I-FAST is a new, complex algorithm used for blind classification of the original EEG tracing of each subject. This is accomplished by recording and analyzing a few minutes of their EEG without any preliminary pre-processing. A proof of concept study published in The Artificial Intelligence Journal in 2015 showed accuracy values ranging from 94%-98% in discerning subjects with Mild Cognitive Impairment and/or Alzheimer’s Disease from healthy elderly people.

Even if the neuropathology related to autism is markedly different from that of Alzheimer disease, simple reasoning would support the idea that the atypical organization of the cerebral cortex present in autism should result in an EEG signature open to detection through potent analytical systems like ANNs.


The aim of the study is to assess how effective this methodology distinguishes ADS subjects from typically developing ones.


Fifteen definite ASD subjects ( age range 8 -22) and ten typically developing subjects ( age range 7-12 ) were included in the study. Patients received independent Autism diagnoses according to DSM-V criteria, then subsequently confirmed by a qualified psychiatrist at Villa Santa Maria, where the patients reside, using the ADOS scale (overall severity score had a range from a minimum of 4 to a maximum of 10 points, average = 7.9). No autistic child was affected by genetic conditions and/or cerebral malformations documented by neuroimaging and epilepsy.

A continuous segment of artefact-free EEG data lasting 60 s in ASCCI format was used to compute multi-scale entropy values and for subsequent analyses.

A Multi-scale ranked organizing map (MS-ROM), based on the self-organizing map (SOM) neural network, coupled with the TWIST system (an evolutionary system able to select predictive features)  created an invariant features vector input of EEG on which supervised machine learning systems acted as blind classifiers.


After MS-ROM/I-FAST preprocessing, the overall predictive capability of different machine learning systems in deciphering autistic cases from normal ones consistently amounted to 100% (Table 1). These results were obtained at different times in separate experiments performed on the same training and testing subsets. The similarities among the ANN weight matrixes measured with apposite algorithms were not affected by the age of the subjects. This suggests that the ANNs do not read age-related EEG patterns, but rather invariant features related to the brain’s underlying disconnection signature.


This pilot study seems to open up new avenues for the development of non-invasive diagnostic testing for the early detection of ASD.