Brain Imaging Study Identifies Three Distinct Neurological Subtypes of ADHD

A study published in JAMA Psychiatry in February 2026 claims to have identified three biologically distinct subtypes of ADHD using structural brain imaging and machine learning — findings that could reshape how the condition is diagnosed and treated. The research, led by Nanfang Pan of West China Hospital and involving collaborators at the University of Cincinnati and Monash University, analyzed MRI scans from over 1,000 children and identified brain connectivity patterns that align with recognizable clinical profiles [1][2].

The study arrives at a moment of heightened urgency: ADHD affects an estimated 8% of children globally and 3.1% of adults, yet standard treatment still relies heavily on trial-and-error prescribing of stimulant medications [3][4]. If validated, brain-based subtypes could guide more precise interventions. But the history of neuroimaging biomarkers for psychiatric conditions is littered with findings that failed to replicate — and this study's demographic limitations raise questions about whose brains defined these categories.

The Study: Methodology and Scale

The research team constructed morphometric similarity networks (MSNs) — a neuroimaging technique that maps structural relationships between brain regions rather than examining individual regions in isolation. They applied normative modeling to identify how individual brains deviate from typical development, then used a data-driven clustering algorithm called HYDRA to group those deviations into distinct patterns [1][5].

The discovery cohort included 446 children diagnosed with ADHD and 708 typically-developing controls. Critically, the clustering was performed without any clinical symptom data guiding it — the algorithm worked solely from brain structure [5]. The team then validated their findings in an independent cohort of 554 children with ADHD from the Healthy Brain Network, reporting "strong correlations between the brain deviation patterns" across both samples [5][6].

By neuroimaging standards, this is a substantial sample. Many prior ADHD imaging studies involved fewer than 100 participants. A 2020 review in Frontiers in Integrative Neuroscience noted that small sample sizes and methodological heterogeneity had severely limited comparability and replicability across the field [7]. The use of an independent validation cohort addresses one of the most persistent criticisms of neuroimaging research.

Source: OpenAlex

Data as of Jan 1, 2026CSV

The Three Biotypes

Biotype 1: Severe-Combined with Emotional Dysregulation

The most clinically concerning subtype showed abnormalities in 45 brain regions — nearly twice as many as the other two groups. The disruptions concentrated in connections between the medial prefrontal cortex and pallidum, areas central to emotional regulation [5][8]. Dr. Manpreet K. Singh, a study co-author at Stanford, described this pattern as an "overloaded control center" where emotional systems are "overwhelmed" [8].

Children in this cluster showed the highest levels of both inattention and hyperactivity/impulsivity, combined with severe mood instability, explosive tantrums, and aggression. The study found they carry the highest risk for developing comorbid depression, anxiety, substance abuse, and involvement with the criminal justice system [5][8]. Neurotransmitter mapping associated this biotype with serotonin, dopamine, and acetylcholine systems.

Biotype 2: Predominantly Hyperactive/Impulsive

The second subtype was marked by disruptions between the anterior cingulate cortex and pallidum — circuits governing impulse control. Singh described it as an "impulse circuit jam" where "the accelerator is strong, the brake timing is slightly off, and traffic flows fast, but unpredictably" [8]. Children in this group showed lower inattentiveness than the other subtypes but greater disruption in inhibition circuits. The associated neurochemistry involved glutamate and cannabinoid receptor systems [5].

Biotype 3: Predominantly Inattentive

The third subtype involved disruptions in the superior frontal gyrus, affecting working memory and sustained attention. This presentation is subtler and less disruptive in classroom settings, which may explain why it is more frequently observed in girls and often goes undiagnosed for years [5][8]. A specific serotonin receptor subtype was implicated.

How These Map Onto — and Complicate — Current Diagnostic Categories

The DSM-5 currently classifies ADHD into three "presentations": predominantly inattentive, predominantly hyperactive-impulsive, and combined. These categories are defined entirely by behavioral symptom checklists completed by clinicians, teachers, and parents [9]. The DSM-5-TR, published in 2022, made no changes to ADHD diagnostic criteria [10].

On the surface, biotypes 2 and 3 appear to correspond roughly to the hyperactive-impulsive and inattentive presentations. But biotype 1 — the severe form — does not map cleanly onto the "combined" presentation because it involves emotional dysregulation and neural circuitry that the DSM does not currently incorporate into ADHD criteria at all [5]. This is significant: emotional dysregulation has long been observed clinically in ADHD patients but was excluded from formal diagnostic criteria.

Any formal reclassification would require revision of the DSM — a process that typically takes over a decade, involves extensive field trials, requires regulatory approval from the American Psychiatric Association, and demands evidence of clinical utility beyond statistical significance. The DSM-5 revision process took 13 years [10]. Brain scanning would also need to become cost-effective and accessible enough for routine clinical use — a threshold no psychiatric neuroimaging tool has yet cleared.

Global Prevalence and the Scale of Potential Impact

ADHD is among the most common neurodevelopmental conditions worldwide. Meta-analyses estimate a global prevalence of approximately 8% in children and adolescents and 3.1% in adults [3][4].

Source: World Population Review / Meta-analyses

Data as of Jan 1, 2026CSV

With a global population of roughly 8 billion, these figures translate to hundreds of millions of affected individuals. If the three-biotype framework holds, it implies that a substantial proportion of current ADHD patients may be receiving treatments optimized for a different subtype. As one researcher noted: "If each biotype involves different neural circuits and neurotransmitter systems, then the standard 'diagnose, prescribe stimulant, adjust dose' approach may only be well-suited to one of the three groups" [8].

However, applying these findings to the full global ADHD population requires caution. The study examined only children, and adult ADHD may involve different or additional neural patterns. The biotype proportions observed in the study sample may not generalize to populations with different genetic, environmental, or socioeconomic profiles.

Replication and the Neuroimaging Credibility Problem

The study's internal validation in the Healthy Brain Network cohort is a strength, with correlations of approximately r = 0.58 between discovery and replication datasets [7]. But independent replication by separate research teams has not yet occurred. This distinction matters enormously.

The field of psychiatric neuroimaging has a well-documented credibility problem. A 2019 editorial in the American Journal of Psychiatry titled "The Enigma of Neuroimaging in ADHD" noted that despite decades of research, no single neuroimaging finding had achieved the reliability needed for clinical application [7]. Stephen Faraone, a leading ADHD researcher, has stated explicitly that neuroimaging "detects very small differences between people with and without ADHD that are only detectable when you have hundreds or thousands of brain scans for comparison" and "should not be used to diagnose ADHD" [11][8].

The study authors themselves stress that these are not new diagnostic categories and that ADHD remains a clinical diagnosis based on behavioral evaluation [8]. The biotypes represent statistical clusters in brain data — whether they correspond to meaningfully different clinical trajectories or treatment responses remains to be demonstrated in prospective trials.

Funding and Financial Relationships

The study was funded by the National Natural Science Foundation of China (multiple grants to Drs. Pan, Gong, and Li) and the U.S. National Institute of Mental Health (grants to Drs. DelBello, McNamara, and Singh) [1][6].

The conflict-of-interest disclosures reveal extensive pharmaceutical industry ties among some authors. Dr. Singh reported advisory board or consulting relationships with AbbVie, Alkermes, Alto Neuroscience, Boehringer Ingelheim, Johnson & Johnson, Karuna Therapeutics, and Neumora, among others [6]. Dr. DelBello reported research grants from Johnson & Johnson, Janssen, Shire, Otsuka, Eli Lilly, Allergan, Pfizer, and consulting fees from Supernus, Neuronetics, and others [6].

These disclosures do not invalidate the research, but they are relevant context. If subtype-specific diagnostics become clinically standard, the companies developing neuroimaging tools, genetic tests, or targeted medications would stand to benefit substantially. Companies like Alto Neuroscience, which already markets brain-imaging-guided psychiatric treatment selection, and Supernus Pharmaceuticals, which sells ADHD medications, are positioned in this space [6].

Demographic Gaps and Generalizability Concerns

The study drew data from sites in the United States and China [5]. The specific racial and ethnic composition of the sample has not been fully detailed in available reporting, but the study's limitations section acknowledges restricted geographic representation.

This matters because ADHD diagnosis already shows stark demographic disparities. Large-scale analyses of over 849,000 patients have found that non-Hispanic White individuals are approximately 26% more likely to receive an ADHD diagnosis than non-Hispanic Black individuals [12]. Black females are the most underrepresented group, with 59% lower diagnostic prevalence than White males [12][13].

Source: Scientific Reports / JAMA Network Open

Data as of Oct 1, 2025CSV

Girls are diagnosed at roughly half the rate of boys — partly because the inattentive presentation more common in girls is less disruptive and therefore less likely to trigger referral [13][14]. If the study sample reflects these existing biases — with fewer girls, fewer Black children, and fewer children from low-income families — then the biotype definitions themselves may be calibrated to brains that are already overrepresented in research.

The study also excluded participants with certain comorbidities and was not conducted with medication-naive samples [5]. Since the majority of children diagnosed with ADHD in the U.S. receive medication, the long-term effects of stimulants on brain structure could confound the identification of "intrinsic" subtypes.

The Neurodiversity Critique

Some researchers and advocates argue that carving ADHD into neurological subtypes risks deepening the medicalization of cognitive variation. The neurodiversity framework holds that ADHD characteristics are not universal deficits but rather reflect a mismatch between neurocognitive styles and environments designed for neurotypical functioning [15].

The strongest version of this critique holds that brain-based subtyping could create a two-tier system: those whose brain scans confirm a recognized subtype receive treatment and accommodations, while those who don't fit — perhaps because their presentation wasn't represented in the training data — face new barriers. Given that brain MRI costs $1,000-$3,000 per scan in the U.S. and is not accessible in most clinical settings globally, a diagnostic framework that requires imaging could further disadvantage low-income and rural populations who already face significant barriers to ADHD care [15][16].

There is also the insurance question. If insurers begin requiring neuroimaging confirmation of a specific biotype before authorizing certain treatments, patients with clinically significant symptoms who don't match a defined pattern could find themselves without coverage. This is not hypothetical — analogous dynamics have played out with genetic testing requirements for cancer therapies.

What Comes Next

Dr. DelBello has framed the findings as validating what clinicians already do intuitively: "trying to match the symptoms to the actual most effective treatments that we know" [2]. The study provides "biological validity" for that clinical instinct, she argues.

But biological validity is not the same as clinical utility. The next steps require prospective treatment trials that randomize patients by biotype and measure whether subtype-matched interventions produce better outcomes than standard care. They also require replication in demographically diverse samples, in adults, and across different imaging protocols and scanners.

The field of ADHD neuroimaging has produced over 36,000 papers since 2011, with annual output peaking near 4,700 in 2024 [17]. The challenge has never been generating findings — it has been generating findings that hold up across labs, populations, and time. Whether these three biotypes represent a genuine advance or another entry in a long list of neuroimaging promises depends entirely on what happens in the next several years of independent testing.

The study "Mapping ADHD Heterogeneity and Biotypes by Topological Deviations in Morphometric Similarity Networks" was published in JAMA Psychiatry on February 25, 2026.