By Timothy Boyer for eMaxHealth
The problem with obtaining a diagnosis of whether or not your child may have autism is the fact that trained clinicians are backlogged with requests for autism testing. This is in part due to that the incidence of autism is high (now 1 in 88) and the testing procedures are lengthy requiring hours rather than minutes for a proper evaluation.
Autism is typically diagnosed though a 93-question questionnaire called the “Autism Diagnostic Interview, Revised” (ADI-R) test and/or via a behavior observation evaluation of the child in question with the “Autism Diagnostic Observation Schedule” (ADOS) exam.
The ADOS exam consists of 4 age-dependent modules that contain semi-structured activities designed to measure social interaction, communication, play and imaginative use of materials. Module 1 contains 10 activities and 29 items and is typically used for assessment of younger children.
Both the ADI-R and ADOS exams can take up to 3 hours or more and must be performed by a trained clinician with experience in diagnosing autism. One of the shortcomings aside from time is that the test results are analyzed subjectively and thereby prone to suffer from human error
To remedy the backlog, time spent diagnosing and human error, researchers from Harvard Medical School have found a way using artificial intelligence to more accurately detect autism in children and establish a diagnosis in minutes rather than hours.
In a recent issue of Translational Psychiatry, researchers report their findings that by using computational algorithms that rely on a few questions and a short video of a child, a quick and accurate diagnosis is possible and could lead to earlier than average treatment.
“We believe this approach will make it possible for more children to be accurately diagnosed during the early critical period when behavioral therapies are most effective,” says Dennis Wall an associate professor of pathology and director of computational biology initiative at Harvard University’s Center for Biomedical Informatics.
The computational algorithms described in the published paper are referred to as “machine-learning algorithms”—a form of artificial intelligence where data is analyzed leading to a resulting diagnosis for autism that can be made efficiently, effectively and without the potential for subjective human error.
The algorithms were applied toward large data samples of patients who had previously been diagnosed with autism through the ADOS exam. What the algorithm analysis revealed was that 8 of the 29 items contained in Module 1 of the ADOS exam were sufficient to classify autism with 100% accuracy. The analysis led to the development of an alternating decision tree (ADTree) algorithm that used in conjunction with a short observational video of a child proved to be highly effective in a speedy and accurate diagnostic method for determining autism in young children.
The significance of this research is that current medical opinion tells us that too little, too late is a problem with treating autism. Delays in an accurate diagnosis of autism results in missed windows of opportunity for providing speech and behavioral therapies that would otherwise lessen the effects of autism on a developing child. In fact, the average age of diagnosis for autism in the United States is 5.7 years and an estimated 27% remain undiagnosed by 8 years of age. Therefore, at these relatively late stages in cognitive and behavioral development, the therapeutic opportunities that may have made a difference later in life are not realized.
The research is also significant in that it can bring autism detection to the home at the click of a mouse. According to Wall, “With this mobilized approach, the parent or caregiver will be able to take the crucial first steps to diagnosis and treatment from the comfort of their own home, and in just a few minutes.”
To further test the effectiveness of their artificial intelligence based diagnostic test for easy detection of autism in minutes, Wall and his colleagues offer an online survey and video site free to the parents of children diagnosed with autism asking for their help in contributing to the research project in developing a future online autism diagnostic test.
Image Source: Courtesy of Wikipedia
References:
“Use of machine learning to shorten observation-based screening and diagnosis of autism” Translational Psychiatry" 2012; 2 (4): D P Wall, J Kosmicki, T F DeLuca, E Harstad, V A Fusaro.
Autism is typically diagnosed though a 93-question questionnaire called the “Autism Diagnostic Interview, Revised” (ADI-R) test and/or via a behavior observation evaluation of the child in question with the “Autism Diagnostic Observation Schedule” (ADOS) exam.
The ADOS exam consists of 4 age-dependent modules that contain semi-structured activities designed to measure social interaction, communication, play and imaginative use of materials. Module 1 contains 10 activities and 29 items and is typically used for assessment of younger children.
Both the ADI-R and ADOS exams can take up to 3 hours or more and must be performed by a trained clinician with experience in diagnosing autism. One of the shortcomings aside from time is that the test results are analyzed subjectively and thereby prone to suffer from human error
To remedy the backlog, time spent diagnosing and human error, researchers from Harvard Medical School have found a way using artificial intelligence to more accurately detect autism in children and establish a diagnosis in minutes rather than hours.
In a recent issue of Translational Psychiatry, researchers report their findings that by using computational algorithms that rely on a few questions and a short video of a child, a quick and accurate diagnosis is possible and could lead to earlier than average treatment.
“We believe this approach will make it possible for more children to be accurately diagnosed during the early critical period when behavioral therapies are most effective,” says Dennis Wall an associate professor of pathology and director of computational biology initiative at Harvard University’s Center for Biomedical Informatics.
The computational algorithms described in the published paper are referred to as “machine-learning algorithms”—a form of artificial intelligence where data is analyzed leading to a resulting diagnosis for autism that can be made efficiently, effectively and without the potential for subjective human error.
The algorithms were applied toward large data samples of patients who had previously been diagnosed with autism through the ADOS exam. What the algorithm analysis revealed was that 8 of the 29 items contained in Module 1 of the ADOS exam were sufficient to classify autism with 100% accuracy. The analysis led to the development of an alternating decision tree (ADTree) algorithm that used in conjunction with a short observational video of a child proved to be highly effective in a speedy and accurate diagnostic method for determining autism in young children.
The significance of this research is that current medical opinion tells us that too little, too late is a problem with treating autism. Delays in an accurate diagnosis of autism results in missed windows of opportunity for providing speech and behavioral therapies that would otherwise lessen the effects of autism on a developing child. In fact, the average age of diagnosis for autism in the United States is 5.7 years and an estimated 27% remain undiagnosed by 8 years of age. Therefore, at these relatively late stages in cognitive and behavioral development, the therapeutic opportunities that may have made a difference later in life are not realized.
The research is also significant in that it can bring autism detection to the home at the click of a mouse. According to Wall, “With this mobilized approach, the parent or caregiver will be able to take the crucial first steps to diagnosis and treatment from the comfort of their own home, and in just a few minutes.”
To further test the effectiveness of their artificial intelligence based diagnostic test for easy detection of autism in minutes, Wall and his colleagues offer an online survey and video site free to the parents of children diagnosed with autism asking for their help in contributing to the research project in developing a future online autism diagnostic test.
Image Source: Courtesy of Wikipedia
References:
“Use of machine learning to shorten observation-based screening and diagnosis of autism” Translational Psychiatry" 2012; 2 (4): D P Wall, J Kosmicki, T F DeLuca, E Harstad, V A Fusaro.
No comments:
Post a Comment