Neurora tracks 100-plus biometrics in autism evaluations. CentralReach’s Thomas Frazier explains what it could mean for clinics, parents, and wait lists.
Key Takeaways
- Long waits persist: CDC data put the median age of autism diagnosis at 47 months, and Frazier said some families wait a year or more just for an evaluation.
- A new FDA-cleared aid: Neurora analyzes video and audio from standard assessments to support, not replace, clinician diagnosis in children 18 months through 12 years.
- Encouraging early evidence: A peer-reviewed study reported 77.8% sensitivity and specificity, with differential diagnosis flagged for further development.
- Clinicians stay in charge: Frazier said tools like Neurora could extend scarce clinicians’ reach, sharpen triage, and may someday support outcome tracking.
For years, at the end of autism evaluations, Dr. Thomas Frazier heard a familiar doubt from parents. They had sat through interviews and questionnaires and watched a clinician observe their child for an hour or so, and now that clinician was ready to offer an answer.
“I can tell you the number of times parents have said to me, ‘Dr. Frazier, I think you’re a great guy. I think you know what you’re doing, but it’s only been an hour or two hours. How are you so confident?’” Frazier told Acuity in a video interview.
That question sits close to the center of a tool the U.S. Food and Drug Administration cleared on October 2: Neurora, software from Argus Cognitive, a Hanover, New Hampshire, company. Neurora analyzes video and audio from a child’s standard behavioral assessment and reports how closely the child’s behavior aligns with patterns seen in children with and without autism. Cleared for ages 18 months through 12 years, it is designed to inform a clinician’s diagnosis rather than make one.
Frazier, Chief Clinical Officer of CentralReach and a Board Member on the Autism Speaks National Board, did not develop Neurora, and he said so at the outset. But as a clinical psychologist who has spent much of his career studying autism assessment, and as the parent of an autistic child, he sees the clearance as addressing confidence as much as speed. “Having that objective tool really increases their confidence,” he said of parents.
How Neurora’s FDA Clearance and AI Biometrics Work
According to the FDA’s 510(k) database, Neurora was found substantially equivalent to a device already on the market. It joins a small category the agency created in 2021, when it granted Cognoa’s request for a pediatric autism diagnosis aid, now sold as Canvas Dx. EarliTec Diagnostics has since received clearances in the same category for its eye-tracking technology.
Argus describes Neurora as a Class II diagnosis aid that quantifies biometrics from recordings of behavioral assessment sessions and is not a stand-alone diagnostic. In the company’s October 8 announcement, Chief Executive Officer Attila Meretei, MD, said Neurora works inside the assessments clinicians already use, without changes to their protocols. Co-Founder and Clinical Lead Latha Valluripalli Soorya, PhD, Board Certified Behavior Analyst, put the division of labor plainly: “The clinician still makes the call.”
Argus says Neurora measures more than 100 vocal, facial, and gaze biometrics but has not said which carry the most weight. “Social attention is a really important aspect of autism spectrum disorder, and various people have demonstrated that over the last 20 years,” Frazier said, pointing to gaze. Audio likely matters too, he added, since some autistic children are minimally verbal and others use idiosyncratic language. In a 2026 study in Frontiers in Psychiatry, Argus researchers found the model’s predictive signal spread across many features, with no single one accounting for as much as 1%.
According to Argus’s clinical research overview, the clinician wears video glasses during a standard ADOS-2 session while a Neurora app captures it, and nothing is required of the child. The output includes an overall score and sub-scores for vocal activity, pitch patterns, vocal rhythm, gaze toward the clinician, facial movements, and smiling patterns, along with a transcript and timestamped recordings tied to the reported behaviors.
Objective Autism Assessment Data and Parent Confidence
Frazier called confidence “a really underappreciated part of what Neurora is trying to do here,” and he began with the clinician. “When you’re working with little kids, especially younger children, you want to be confident in the judgments that you have as a clinician,” he said. “If you don’t have confidence, it becomes very hard to make critical judgments, like whether or not they meet diagnostic criteria.”
Parents, who know their children in ways no clinician can in an afternoon, bring their own doubts to the feedback session, and Frazier said objective measurements layered on top of the clinician’s work can ease that skepticism.
Autism Diagnosis Wait Lists and the ADOS-2 Bottleneck
Most diagnostic evaluations lean on the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2), a structured observation that Frazier called the gold standard. It is also demanding, requiring extensive training, ongoing reliability checks, and supervision by clinicians who are in short supply. “These clinicians are like gold; they’re very hard to find,” Frazier said. State Medicaid rules can narrow that pool further: as of November 1, Kansas Medicaid accepts autism diagnoses for ABA only from doctoral psychologists and qualified physicians.
In his view, Neurora’s practical value lies in extending their reach. “What Neurora is allowing you to do as a clinician is train more people, but use Neurora as a supplement and have more direct access to the patient’s behavior,” he said. The study authors make a similar case, describing the approach as a task-sharing tool that could let technicians or therapists administer an assessment while diagnosticians handle scoring and clinical decisions.
“I wouldn’t say that the software itself expedites the diagnosis per se,” he said. Rather, a supervising clinician with a clearer window into each session can run a more efficient clinic, “and because it becomes more efficient, you reduce wait lists.” Families feel that delay directly. “Patients can wait long, long times for these evaluations,” he said. “Sometimes one year or more just to get evaluated.”
Federal data reflect the lag. The CDC’s most recent Autism and Developmental Disabilities Monitoring Network report, which estimated autism prevalence at 1 in 31 children aged 8 in 2022, put the median age of earliest known diagnosis at 47 months.
Autism Evaluation Triage and Referrals Beyond Autism
The cleared age range covers the vast majority of diagnostic evaluations, Frazier said. If he were still conducting them, a tool like this would let his team bring families in for a quicker initial evaluation and then decide who needs the full multidisciplinary workup: speech and language testing, cognitive testing, the ADOS-2, parent interviews, and a stack of questionnaires.
Better triage serves children who are not autistic, too. “There are a number of patients on these wait lists that do not have autism, and they need to be screened out,” he said. Parents’ concerns are almost always valid, he added, but they may point toward a developmental pediatrician, speech-language therapy, or an ADHD specialty clinic. “You want to get those kids to the right thing for them, even if they don’t have autism spectrum disorder.”
What the Neurora Clinical Study Shows on Accuracy
The peer-reviewed evidence behind Neurora comes primarily from the Frontiers study, which enrolled 546 children ages 2 to 12 at four sites in the United States and Qatar. Of those, 458 had recordings that met quality standards. The researchers reported 77.8% sensitivity and specificity in distinguishing autistic children from non-autistic children, including children with other clinical conditions, in their primary analysis set. Against neurotypical children alone, both figures rose to 82%. On a separate hold-out set of 120 children, the model reached 62.3% sensitivity and 81.4% specificity.
The authors identified differential diagnosis (separating autism from other neurodevelopmental conditions with overlapping features) as a limitation requiring further development, noting that the study included relatively few children with co-occurring conditions. They added that the ADOS-2 also loses accuracy in children with co-occurring psychiatric disorders. Argus says its most recent clinical study was conducted with partners including LSU Health Shreveport, Autism Testing 4 Kids, and Ascend Behavior Partners.
Frazier framed those limits as the reason clinicians remain essential. “None of these are going to be perfect,” he said of the new generation of tools, and clinicians need to recognize when a tool “is giving a very unusual or different answer than you would expect.”
AI in Autism Care and Keeping Clinicians in Charge
“It’s extremely important to have the clinician driving the process,” Frazier said. As tools like Neurora are validated and adopted, he added, clinicians will need training to use them in an evidence-based way, so that no one is “just taking them and saying, ‘Oh, the piece of paper that printed out here after this test said this, so I’m going to go with that.’”
At CentralReach, he said, that principle is built into the company’s own AI products. NoteDraftAI, which drafts session notes from behavioral intervention data, requires clinicians to read and edit each draft, he said, and back-end checks flag providers who appear to be accepting outputs wholesale. “If we’re seeing situations where people are just accepting notes, then we know people need more training, and they need more support,” he said. CentralReach also has an AI governance board and a team that oversees AI ethics, according to Frazier. “You could build technologies that actually really assure that, or at least come as close as you possibly can to assure that,” he said.
The concern behind that approach has surfaced in the field. Acuity reported in August that two private equity-backed ABA providers asked an AI vendor to make assessment reports submittable without a BCBA reading them, a request the vendor declined.
Neurora and Objective Autism Outcome Measurement
Neurora’s clearance covers diagnosis, not treatment monitoring, and Frazier was clear that anything beyond that is speculative. “We don’t know yet how it’s going to factor in,” he said of the assessment and outcomes work that follows a diagnosis. Still, he sees potential for the same measurements to set a baseline before intervention and track progress over time, work the field now does largely through parent reports and clinician ratings. The study authors raise a similar possibility. Payers, meanwhile, have begun asking for a more objective anchor, Jamie Pagliaro argued in a contributed Acuity column this spring; Pagliaro is Chief Executive Officer of EarliPoint Health, which has developed eye-gaze measurement technology.
“We’re not there yet, and I just want to be clear, Neurora is not there yet,” Frazier said. “They would tell you that. But I do think that there’s a lot of promise here in terms of outcome assessment.”
Objective Autism Diagnostic Tools Beyond Neurora
Frazier was candid that he is not a neutral observer of this field. His own work includes AI.Measures, the assessment platform he co-founded and led before CentralReach acquired it in 2025. “There are a bunch of us who are working on this problem, including me,” he said. “So I have a dog in the fight.”
Biological screeners face a steeper climb than behavioral tools, he said. “One thing we’ve learned over the last 20 years is that autism is incredibly heterogeneous, and that’s true at the behavior level, but it’s especially true at the biological level.” Even so, he allowed, “You never say never on something like this.”
Rural Autism Diagnosis Access and Diagnostic Disparities
Frazier closed on families far from a diagnostic center. He said CDC data show that much of the delay in diagnosis falls on historically marginalized and rural groups, some of whom live hours from the nearest hub clinic. Online diagnostics have helped, he said, and tools that shorten wait lists can help further. Policy is not uniform on that front: Virginia’s proposed Medicaid ABA manual would require an in-person diagnostic evaluation before ABA is authorized. “I think the rural group in particular is really not talked about enough,” he said.
For Frazier, that is the larger point of a clearance like Neurora’s. “We’re trying to reduce bias,” he said. “We’re trying to reduce these kinds of disparities, and we need these kinds of tools to do this.”
