EEG DATA Processed By Advanced Machine Learning Systems ALLOW an Accurate Differential Diagnosis between ASD Children and Children with Other Neuro-Psychiatric Disorders.

Background: In a previous study the authors have shown the ability of a novel kind of Machine Learning System(MLS) named MS-ROM/I-FAST developed by The Semeion Research Institute in Rome to extract interesting features in computerized EEG that allow an almost perfectly distinction of ASD children from typically developing ones. The proof of concept study, published in 2017 in Computer Method and Programs in Biomedicine showed accuracy values near to 100% using training-testing protocol and to 84% - 92.8% using Leave One Out protocol. The similarities among the MLS weight matrixes measured with apposite algorithms were not affected by the age of the subjects suggesting that the MLS do not read age-related EEG patterns, but rather invariant features related to the brain’s underlying abnormalities.

Objectives: The aim of the study is to assess how effectively this methodology distinguishes ADS subjects from children affected with other neuro-psychiatric disorders .

Methods: Twenty definite ASD subjects and twenty subjects with neuropsychiatric disorders matched for age and gender distribution observed at Tarnow Center for Self-Management, Huston (US) were included in the study. The two groups had the same age range ( 4-14 yrs) and male/female ratio (14/6). ASD patients received independent Autism diagnoses according to DSM-V criteria, subsequently confirmed by a qualified psychiatrist using the ADOS scale. No autistic child was affected by genetic conditions and/or cerebral malformations documented by neuroimaging and epilepsy. In the comparison group the range of primary diagnoses was the following: Attention-Deficit Disorder ( N= 13), Disorder of social functioning( N=3), Anxiety disorders( N= 2), Major depressive disorder(N= 1), Specific developmental disorders of scholastic skills(N=1). A continuous segment of artefact-free EEG data lasting 10 minutes 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.

Results: After MS-ROM/I-FAST preprocessing, twelve features were extracted representing the EEG signature. Acting on these features the overall predictive capability of different machine learning systems in deciphering autistic cases from other NP disorders ranged between 93% and 97.5% (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.

Conclusions: This study demonstrates the value of EEG processing with advanced MLS in the differential diagnosis between ASD and other NP disorders confirming therefore the existence of a specific EEG signature in ASD.