Apr 20, 2018 | Author: neuromimeTICs
The fight of science against autism intensifies. This month of April, which began with the Autism Awareness Day, has closed with the publication of new findings on the genetics of this disorder as a result of an important international research. In parallel, AI applications are proliferating to improve the diagnosis and the quality of life of affected individuals.
April 2018 was the month to remember autism, firstly with the celebration of the World Day of Autism Awareness, then with the celebration of the World Autism Month throughout the full month and, as a culmination, with the publication by Science of a new advance in the understanding of the genetics of the autism spectrum disorder (ASD), which is manifested by an alteration of the communicative and linguistic abilities, deficits in the social interaction and the restriction of the behaviors and interests of the people who suffer them, in different degrees.
This disorder, multifactorial and polygenic, affects one of every 160 children in the world, according to the World Health Organization (WHO). The ASD starts in childhood, but persists in adolescence and adulthood. In fact, patients with ASD are getting older and increasingly there are more mature affected people who have specific needs, not yet well covered by society or public authorities, since up to now the programs of comprehensive attention to the ASD have focused fundamentally in affected children, as has been shown in an event, also celebrated this April in Barcelona.
On his part, science does not stop working to unravel the origin of ASD. Why do some people develop it and others
do not? A recent international research published by Science, under the title “Paternally inherited cis-regulatory structural variants are associated with autism”, provides new clues about the genetics of autism. This research has been carried out in 2,600 families and has focused on studying the rare genetic variants inherited from the parents, while until now the genetic studies carried out have analyzed the de novo genetic variants. According to the experts, if the pathogenic structural variants –which account for the 1.9% of ASD cases- are ruled out, the study reveals that the identified structural variants contribute by 11% of ASD cases, which is an important figure. In addition, half are de novo mutations that affect specific genes and the other half are inherited mutations that alter regulating elements or genes, reports the University of Barcelona (UB). These findings brings the researchers even closer to completing the genetic puzzle that gives rise to autism.
The research has been directed by Jonathan Sebat, Professor of Psychiatry and Molecular and Cellular Medicine at the University of California San Diego (UCSD) and Director of the Beyster Center for Psychiatric Genomics, with the involvement of Carig Venter, promoter of the Human Genome Project that marked the beginning of the new era of sequencing different genomes, and thirty institutions from all over the world and the outstanding participation of several Spanish research groups.
This study, through the technique of Whole-genome sequencing, has made possible to locate rare genetic variants where previously other sequencing technologies have not been able to detect them, due to their small size or because they are located in non-coding regions of the genome; among these rare variants there are deletions, tandem duplications, insertions, inversions or complex structural changes. It has also broken with the general idea that the genetic risk of autism is mainly due to fathers and, to a lesser extent, to mothers, as they are less vulnerable to the development of this disorder.
Thus, the new study indicates that the effects of paternal or maternal inheritance on the genetic risk of autism are more complex than previously thought. These new genetic clues provide some insight about which brain functions may be altered in the autism spectrum disorder and on the points of the genome that could serve as a therapeutic target, according to the UB report.
Artificial Intelligence, a new strategy to combat autism
Another branch of science that has been strongly divided in the battle against autism is AI. Particularly, machine learning is expected to shorten the diagnosis of autism, which, despite having an important genetic component, continues to be based on behavioral tests and questionnaires, such as the Autism-Revised Diagnostic Interview (ADI-R), one of the most common instruments to diagnose ASD and that includes 93 questions that must be answered by a care provider and, which can easily take time, up to 2.5 hours.
In 2012, a research in PLOS ONE implemented a variety of machine learning algorithms, and found one, the Alternate Decision Tree (ADTree), to have high sensitivity and specificity in the classification of individuals with autism controls. The ADTree classifier consisted of only seven questions, 93% less than the total of the ADI-R questionnaire, and was performed with an accuracy greater than 99% when applied to independent populations of individuals with autism, erroneously classifying only one of the 1962 cases used for the validation. The authors concluded that “given the dramatic reduction in numbers of questions without appreciable loss in accuracy, our findings may represent an important step towards making the diagnosis of autism a faster process that enables delivery of therapy at earlier and more impactful stages of child development”.
But a recent review of machine learning in ASD research published this year 2018 in Informatics for Health and Social Care points that studies concerning the use of machine learning in ASD diagnosis and treatment suffer from conceptual, implementation, and data issues such as the way diagnostic codes are used, the type of feature selection employed, the evaluation measures chosen, and class imbalances in data among others. Therefore, there is a need to improve all these aspects. In their article, the authors recommend some paths forward that enhance machine learning use in ASD with respect to conceptualization, implementation, and data to benefit the research of this disorder with machine learning.
From the therapeutic point of view or the comprehensive care of individuals affected with ASD, there are several different AI initiatives consisting of the use of autonomous and mobile robots, of different types, to interact with children within the autism spectrum and to analyze its influence on the improvement of their social and emotional interaction skills.
A recent example is the robotic platform developed by researchers from the University of Tehran for reciprocal interaction with children affected with ASD. As they say, “children with autism show a particular interest toward robots, and facial expression recognition can improve these children’s social interaction abilities in real life”.
Thus, they have developed a robotic platform consisting of two main phases, namely as Non-structured and Structured interaction modes. In the first one, a vision system recognizes the facial expressions of the user through a fuzzy clustering method while, in the second one, “the interaction decision-making unit is combined with a fuzzy finite state machine to improve the quality of human-robot interaction by utilizing the results obtained from the facial expression analysis”, expose the authors. Also, in the interaction mode, the robot has a set of imitation scenarios with eight different posed facial behaviours to interact with the children.
The researchers have tested the acceptability and the effect of the platform among autistic children between 3 and 7 years old in a pilot study. So the scenarios started with a simple facial expressions and became more complicated as they continued. Despite this increase in dificulty, the preliminary acceptance of these children has been around 78%, as they researchers observed in their experimental conditions. These results have been published in International Journal of Social Robotics this year 2018.