Neuroscientists from the Boston Children’s Hospital have developed a machine-learning algorithm that can spot autism even before the symptoms show. That is, by analyzing the heart rate and pupil dilation of a child.
To date, Autism Spectrum Disorder (ASD) and other neurodevelopmental disorders remain undiscovered until affected children have aged a few years. Unfortunately, behavioral interventions and speech or occupational therapies have already become ineffective by the time the disease is diagnosed.
Dr. Michela Fagiolini and Dr. Pietro Artoni, together with their colleagues, wanted to diagnose the disease early and allow affected children to get treatment immediately.
The team has already tested the system, and initial results showed that it could detect if a young girl has Rett Syndrome accurately. Rett is a genetic disorder that affects the cognitive, sensory, motor, and automatic functions of infants aged 6 to 18 months.
With their new AI system, Dr. Fagiolini and Dr. Artoni hope to detect not just Rett Syndrome but other neurodevelopmental disorders as well in the future. Dr. Fagiolini noted:
“We want to have some readout of what’s going on in the brain that is quantitative, objective, and sensitive to subtle changes. More broadly, we are lacking biomarkers that are reflective of brain activity, easy to quantify, and not biased. A machine could measure a biomarker and not be affected by subjective interpretations of how a patient is doing.”
Detecting Autism via Heart Rate and Pupil Dilation
The development of the AI system was based on the idea that people on the autism spectrum have altered behavioral states. Previous studies suggest that the brain’s cholinergic circuits and the altered arousal cause pupil dilation and changes in heart rate.
Using these findings as their basis, Dr. Fagiolini and her team measured the pupil fluctuations in several mouse models of ASD. The models also include those with a mutation that leads to CDKL5 disorder or Rett Syndrome.
The experiment was able to detect autism successfully even before the mouse models show symptoms of ASD.
Using 60 hours of observation from their experiment, the researchers trained their AI system to recognize abnormal pupillary patterns. When tested, the algorithm was able to detect cholinergic dysfunction in mouse models with BTBR, CDKL5, and MeCP2 deficiency.
Dr. Fagiolini and her team used the machine-learning algorithm on 35 young girls with Rett Syndrom and 40 typically developing controls. However, instead of measuring pupil dilation, the team focused on the heart rate fluctuations to measure the altered arousal of the patients.
The algorithm was able to recognize the girls with Rett with 80 percent accuracy during the first and second year of life. Dr. Fagiolini said:
“If we have biomarkers that are non-invasive and easily evaluated, even a newborn baby or non-verbal patient could be monitored across multiple timepoints.”
The team’s study titled “Deep Learning of Spontaneous Arousal Fluctuations Detects Early Cholinergic Defects Across Neurodevelopmental Mouse Models and Patients” is published in the journal PNAS.
Comments (0)
Most Recent