Background: Restrictive Repetitive Behaviors (RRBs) and sensory features are core symptoms of autism spectrum disorder (ASD); their impact on patients’ quality of life warrants extensive literature studying them. Few studies so far have addressed the nature of the interplay between these two traits with a data mining approach.

Objectives: The aim of the present study was to depict the associations between demographic and clinical variables and the full spectrum of RRBs and sensory processing abnormalities in a sample of children with a diagnosis of idiopathic ASD using data mining techniques belonging to artificial intelligence. The expectation is that this approach would be able to uncover important latent connections between RBBs and sensory abnormalities and other clinical/demographic variables considered.

Methods: Forty-five patients under 18 years (range: 3–15, median 6) with ASD diagnosis according to DSM-5 entered the study. Demographic information (gender, age), clinical variables (epilepsy, sleep problems, Intellectual Disability (ID), rehabilitation therapy, pharmacological therapy), and Short Sensory Profile (SSP) subscales scores and Repetitive Behavior Scale-Revised (RBS-R) subscales scores have been collected in anonymized form and recorded in an Excel database. To graphically show the most important connections among the 25 variables we used a fourth-generation artificial neural network (ANN) called Auto-CM, that develops weights that are proportional to the strength of the associations of all variables each other. After the training phase, the weights matrix of the Auto-CM represents the warped landscape of the dataset. Subsequently, a minimum spanning tree filter is applied to the weights matrix of the Auto-CM system to obtain a map of the main connections between the variables of the dataset and the basic semantic of their similarities, defined connectivity map.

Results: Figure 1 show the semantic connectivity map produced by the Auto-CM software. The map (figure 1) highlighted these interesting features: 1) the Low Energy subscale (SSP-VI) has proven to be the central hub of the system; 2) the Self-Injurious Behaviors subscale was directly linked to ID; 3) RRBs was directly linked with sleep disturbance, with Sameness Behaviors (RBS-V) and Ritualistic Behaviors (RBS-IV); 4) sequential links line connecting RBS-R subscales Sameness Behaviors, Restricted Interests, Ritualistic Behaviors, Stereotyped Behaviors, Compulsive Behaviors and the variable male.

Conclusions: The application of ANNs to our dataset confirmed the well-known association between RBBs and sensory abnormalities; moreover, interesting association between demographic/clinical variables and SSP and RBS-R subscales, previously emerged by classical statistical analysis, were confirmed and deepened. Interesting links emerged between the subscale Self-Injurious Behaviors and Intellectual Disability and between Sleep Disturbance and various RRBs. In conclusion, our study provides new insight into the relationship between RRBs and sensory abnormalities in children with ASD by applying ANNs for the first time in this area.

Notes: 

(1)Child Neuropsychiatry, IRCCS ISNB Ospedale Bellaria- Policlinico S. Orsola-Malpighi, Bologna, Emilia-Romagna, Italy, (2)Autism Research Unit, Villa Santa Maria SCS, Tavernerio, Como,