ADHD Brain Scan Subtypes: What the 2026 Research Actually Shows (And What It Doesn't)

Why This Is in the News Right Now

In late April and early May of 2026, several major outlets ran headlines about three "types" of ADHD identified by brain scans. The Washington Post, National Geographic, Healthline, and others covered the story within days of each other. Most of the coverage was breathless. Some of it implied that ADHD has now been redefined, that brain scans can identify which type you have, and that treatment will soon be guided by imaging. None of those implications is accurate.

The underlying science, however, is genuinely interesting. For decades, clinicians have known that ADHD is heterogeneous — that two children with the same DSM diagnosis can present very differently and respond differently to the same medication. The newer studies attempt to do something specific and worthwhile: instead of starting with behavioral categories and looking for biology, they start with biology directly and let statistical clustering identify subgroups. This is the right scientific direction. It is not yet a clinical tool.

This article walks through what the two recent landmark studies actually found, what their findings do and do not support, how to think about commercial SPECT services marketed for ADHD diagnosis, and what patients and parents should do today.

The JAMA Psychiatry 2026 Biotypes Study: What It Did

The headline 2026 paper is: Mapping ADHD Heterogeneity and Biotypes by Topological Deviations in Morphometric Similarity Networks. JAMA Psychiatry. 2026.

The methodology was sophisticated and worth understanding because it explains what the findings do and do not mean.

The investigators used structural MRI scans from more than 3,500 children, including those with ADHD diagnoses and unaffected controls. From each brain scan, they constructed a morphometric similarity network — a mathematical representation of how brain regions co-vary in their structural features (cortical thickness, surface area, curvature, and so on). Regions whose anatomical profiles are similar are considered "connected" in the network, even if they are not adjacent in space.

This is a relatively new approach in psychiatric neuroimaging. It captures something that voxel-based analyses miss: the way that distributed brain regions form coordinated systems. Two children might have superficially similar MRIs in terms of regional volumes but very different network organization, and morphometric similarity networks can detect that difference.

The investigators then applied unsupervised machine learning — clustering algorithms that do not know which children have ADHD or which DSM subtype — to identify groupings within the data. The clustering produced three reproducible subgroups, each with distinct patterns of topological deviation from healthy controls.

The Three Biotypes Identified

The biotypes do not map cleanly onto the DSM-5-TR presentations of ADHD (predominantly inattentive, predominantly hyperactive-impulsive, combined). A child diagnosed as combined-presentation could fall into Biotype 1, 2, or 3 depending on underlying network organization. That mismatch is the most clinically interesting feature of the work: it suggests that symptom-based classification may be partially obscuring biologically meaningful subgroups.

The investigators also examined genetic risk profiles using polygenic risk scores and reported that the three biotypes had different patterns of genetic loading across psychiatric conditions, with some biotypes showing stronger correlation with autism spectrum and others with mood disorder genetic architecture. This is consistent with the broader genetic correlation findings between ADHD and other neurodevelopmental conditions, and it provides some external validation that the biotypes are tracking real biological variation rather than statistical noise.

The Translational Psychiatry 2025 Study: A Different Method, Convergent Result

The second important paper is: Distinct neuroimaging subtypes of ADHD among adolescents based on semi-supervised learning. Translational Psychiatry. 2025.

This study used data from 6,509 adolescents in the Adolescent Brain Cognitive Development (ABCD) Study — a long-running NIH-funded cohort that has scanned and followed thousands of children across the United States. The investigators applied a different machine learning approach, semi-supervised clustering on cortical thickness measurements, to identify subgroups defined by patterns of cortical structure relative to healthy controls.

They identified three subtypes:

• Under-developed subtype — adolescents with lower cortical thickness than controls across multiple regions, suggestive of a delayed or attenuated developmental trajectory
• Over-developed subtype — adolescents with higher cortical thickness than controls, which may reflect delayed pruning or atypical maturation
• Mixed subtype — adolescents with regional patterns combining features of both under- and over-development

Several features of this study deserve attention. First, the sample size — over 6,500 adolescents — is large by neuroimaging standards and allows real statistical separation of subgroups. Second, the ABCD cohort is followed prospectively, so the same children are scanned repeatedly across adolescence; this allows the investigators to ask whether the subtypes are stable over time or shift with development. Third, the cortical thickness approach captures a different aspect of brain structure than the network analysis in the JAMA Psychiatry paper, yet both studies converge on a three-subgroup solution.

That convergence is meaningful. When two independent methods applied to different cohorts produce a similar number of biologically distinguishable subgroups, the result is more likely to reflect real underlying biology than methodological artifact. The specific definitions of the three subgroups differ between studies — they are not identical biotypes — but the pattern of three-way heterogeneity is replicated.

What These Studies Cannot Do (Yet)

The most important thing to understand about both papers is what they did not attempt to do: produce a clinical test that classifies individual patients.

Group-level statistical findings — where 3,500 or 6,500 brains are analyzed together and statistical clusters emerge — do not translate directly into individual-level diagnostic tools. The challenge is well-known in medical imaging: a feature that distinguishes groups with high accuracy may have far lower accuracy when applied to a single new scan, because individual variability within any cluster is large.

Specifically, these studies do not establish that:

• A clinician can take a child's MRI and assign them to one of the three biotypes with diagnostic accuracy
• The biotypes predict medication response in a way that should change current prescribing practice
• Brain scans can substitute for clinical evaluation in ADHD diagnosis
• The biotypes will replace the DSM presentation categories in clinical use anytime soon
• Any commercial scan marketed to identify "your ADHD type" is using the methodology of these studies

What the studies do establish is a foundation for the next decade of work: identifying which biotypes predict which long-term outcomes, which respond best to which interventions, and whether prospective longitudinal data confirm the clinical relevance of biologically defined subgroups. This is research that will plausibly change clinical practice in the 2030s. It does not change clinical practice in 2026.

The Amen Clinics SPECT Problem

A separate question arises when patients ask about brain scans for ADHD: the commercial SPECT services marketed by clinics like Amen Clinics. These deserve a direct answer, because they are confused with the research described above in the public mind, and the conflation is consequential.

Single Photon Emission Computed Tomography (SPECT) is a nuclear medicine imaging technique that measures regional cerebral blood flow using injected radioactive tracers. It has legitimate medical uses, including evaluation of certain seizure disorders and dementias. It is not part of standard ADHD evaluation in any major academic medical center, professional society guideline, or evidence-based clinical pathway.

Amen Clinics has, over multiple decades, claimed to identify seven distinct "types" of ADHD on the basis of SPECT findings, with each type purportedly responding to a different treatment combination. They market diagnostic scans to patients out of pocket, typically at a cost of $3,000-$4,500 or more.

The evidence base for these claims has been criticized extensively. In 2005, the American Psychiatric Association Council on Children, Adolescents, and Their Families issued a white paper concluding: "At the present time, the available evidence does not support the use of brain imaging for clinical diagnosis or treatment of psychiatric disorders in children and adolescents." That position has not changed.

Specific scientific concerns include:

• Lack of independent replication. The seven-type ADHD framework has not been reproduced by research groups outside of Amen Clinics. Independent replication is a basic criterion of scientific validity, and its absence is not a minor issue — it is the central one.
• Insufficient peer-reviewed validation. The peer-reviewed work from Amen Clinics that does exist tends to compare ADHD patients to healthy controls on group-level perfusion patterns, which is a different claim than the diagnostic one used in their clinical marketing.
• Radiation exposure. SPECT imaging involves injection of a radioactive tracer with associated radiation dose. This is justified when the diagnostic yield warrants it. For ADHD evaluation, the diagnostic yield does not warrant it.
• Cost and opportunity cost. Out-of-pocket SPECT spending diverts resources from interventions with established efficacy — comprehensive clinical evaluation, evidence-based medication management, behavioral therapy, school-based accommodations, and treatment of comorbid conditions.

The clinical position is straightforward: if you are evaluating ADHD, do not pay for a SPECT scan. The money is better spent on a thorough evaluation with a clinician who is trained in ADHD and adjacent neurodevelopmental conditions, and on the treatment that follows.

How to Read Brain-Imaging Psychiatry News in 60 Seconds

Coverage of neuroscience findings in mainstream media follows predictable patterns. A rubric for reading these stories without being misled:

1. What was the sample size? A study with thousands of subjects is far more likely to produce stable findings than one with dozens. Both 2025-2026 ADHD biotype papers used thousands of subjects, which is part of why they merit attention.
2. Was the work done in a single clinic, or across multiple sites? Multi-site replication is harder, more expensive, and more credible. Single-clinic findings — especially from clinics that also sell the test — should be read skeptically.
3. Group-level finding or individual-level test? Most neuroimaging findings, including these biotype studies, are statistical patterns across groups. They do not directly produce individual diagnostic tools. Headlines that claim otherwise are usually overreaching.
4. Replicated by independent groups? If only one research group has produced a particular finding, treat the finding as preliminary regardless of how impressive the methods sound.
5. Peer-reviewed and where? A finding in JAMA Psychiatry, Nature Genetics, Translational Psychiatry, Molecular Psychiatry, or similar venues has cleared a substantial review process. A finding only described on a clinic website has not.
6. What does the editorial accompanying the paper say? Major journals typically publish an editorial alongside a high-profile paper, written by experts in the area. The editorial almost always notes limitations the press release omitted.

If you apply this rubric to the JAMA Psychiatry 2026 paper: large multi-site sample, peer-reviewed in a top journal, methodologically novel, single study, not yet independently replicated by other groups. That places the finding in the "scientifically credible, clinically immature" category. The Amen Clinics SPECT claims fall outside the rubric on most axes.

Why This Research Matters Even Though It Isn't a Clinical Tool Yet

It is fair to ask why patients and parents should care about findings that do not change what happens at their next appointment. Two reasons.

First, ADHD heterogeneity has clinical implications even before biotyping is operationalized. The clinical observation that two children with identical DSM diagnoses respond very differently to the same stimulant is real, and parents who have had this experience know it. The biotype research is the first credible attempt to put that heterogeneity on a biological footing rather than treating it as random clinical variation. Knowing that this work is underway should change how patients and parents understand the inherent uncertainty in initial medication choice: stimulant selection in 2026 is still partially empirical, not because clinicians lack skill, but because the biology has not yet been operationalized.

Second, the convergence between the 2025 and 2026 papers — both finding three biotypes with different methods — strengthens confidence that ADHD is not a single condition with a single brain signature. This is consistent with the pharmacological natural course of ADHD, which shows different developmental trajectories and treatment responses across patients, and with the genetic data showing that ADHD shares risk with multiple other neurodevelopmental and mood conditions.

The likely future is a more individualized approach to ADHD diagnosis and treatment that integrates clinical assessment, family history, comorbidity profile, and — eventually — biomarker data. We are not there yet. But the research direction is real, and patients who understand it can have more substantive conversations with their clinicians about treatment choices and what is still uncertain.

What to Actually Do Today If You Suspect ADHD

For parents of a child with possible ADHD, or adults wondering about their own attention and organization patterns, the clinically grounded next steps:

1. Get a structured clinical evaluation

The diagnostic standard in 2026 remains a clinical evaluation by a psychiatrist, developmental pediatrician, neuropsychologist, or experienced primary care provider with ADHD expertise. The evaluation should include:

• A thorough developmental history — symptom onset, course, and persistence across settings
• DSM-5-TR symptom assessment using validated rating scales (Vanderbilt, Conners, ASRS, or similar) completed by multiple informants where possible (parent, teacher, self for adults)
• Functional impairment assessment — concrete examples of how attention and executive function are affecting work, school, relationships, and safety
• Differential diagnosis and comorbidity screen — anxiety, depression, learning disorder, sleep disorder, substance use, thyroid dysfunction, sleep apnea, and other conditions that can mimic or co-occur with ADHD
• Family history — given the high heritability of ADHD, family patterns are diagnostically informative

2. Skip the brain scan, qEEG, and pharmacogenomic test

None of these are part of evidence-based ADHD evaluation. They add cost without adding diagnostic accuracy. If a clinic recommends them as part of an initial workup, that recommendation should prompt a second opinion.

3. Understand the consequences of leaving ADHD untreated

The cost of not getting evaluated and treated, when ADHD is present, is not abstract. Untreated ADHD is associated with substantial increases in risk for accidental injury, academic failure, substance use, suicide attempts, and shortened life expectancy. The decision to evaluate is not low-stakes.

4. If diagnosed, understand the treatment evidence base

The pharmacology of ADHD treatment rests on one of the strongest evidence bases in psychiatry. Stimulant medications are highly effective for most patients and have favorable safety profiles when monitored appropriately. Non-stimulant options exist for patients for whom stimulants are contraindicated or poorly tolerated. Behavioral therapy, coaching, and environmental accommodations have additive benefit, particularly in children and adolescents.

5. Engage with the heterogeneity productively

The biotype research described above is one expression of a clinical truth: ADHD is not one thing. If you or your child does not respond to the first medication tried, that is not a failure of diagnosis or a sign that ADHD is not the right framework. It is consistent with what every careful clinician already knows — that medication selection is partly empirical, and that finding the right agent and dose can take iteration. The clinical literature, the genetics literature, and now the biotype literature all converge on this point.

Frequently Asked Questions

Are there really three types of ADHD on brain scans?

In two recent rigorous research studies, yes — three statistically distinct subgroups were identified using brain imaging and machine learning. The subgroups are reproducible patterns in research data, not validated clinical categories. No individual child can yet be reliably assigned to one of these biotypes outside research settings.

Can a brain scan diagnose ADHD?

No. No brain imaging modality — MRI, fMRI, SPECT, PET, EEG, or qEEG — currently diagnoses ADHD in an individual patient with clinical-grade accuracy. The American Psychiatric Association explicitly addressed this in a 2005 white paper, and that position has not changed. ADHD diagnosis is clinical: developmental history, symptom criteria, evidence of impairment, and assessment for comorbid conditions.

Should I pay for a brain scan to confirm my ADHD diagnosis?

No. Commercial brain SPECT services marketed for ADHD diagnosis, typically costing $3,000-$4,500, do not have the evidence base to support their diagnostic claims. The research-grade biotype findings are statistical patterns across thousands of brains. They do not translate into a useful individual scan result. A thorough clinical evaluation is more accurate and orders of magnitude less expensive.

What did the JAMA Psychiatry 2026 biotypes study actually find?

Using brain network analysis of MRI data from over 3,500 children, the investigators identified three biotypes: one with prefrontal-pallidal emotional dysregulation circuits and severe combined symptoms; one with anterior cingulate impulse-control circuit disruption and prominent impulsivity; and one with superior frontal gyrus disruption and inattention-predominant symptoms. The biotypes had distinct genetic profiles and preliminary differences in stimulant response.

How is this different from the DSM ADHD presentations?

The DSM-5-TR presentations (inattentive, hyperactive-impulsive, combined) are defined by observed behavioral symptoms. The new biotypes are defined by neurobiology directly. The two systems overlap imperfectly — a child diagnosed with combined-presentation could fall into any of the three biotypes. This mismatch is the most clinically interesting feature of the work and the reason it may eventually inform a more biology-based diagnostic framework.

What should I do today if I think I or my child has ADHD?

See a qualified clinician for a structured evaluation. The evaluation should include developmental history, validated rating scales, screening for comorbid conditions, and assessment of functional impairment. Do not pay for brain scans, qEEG, or pharmacogenomic testing as part of initial diagnosis.

Primary References

Further Reading

• ADHD and Life Expectancy: What the UK Cohort and Barkley Data Actually Show — Companion analysis of the mortality and life expectancy findings, with treatment implications
• Adverse Outcomes of Untreated ADHD — Pillar resource: the full evidence on suicide, substance use, accidents, and academic failure in untreated ADHD
• ADHD Pharmacology and Natural Course — Pillar resource: medication evidence, including the JAMA Psychiatry 2025 Li SCCS study on functional outcomes
• Is ADHD Genetic? — Twin studies, GWAS, and what the heritability data mean for families
• ADHD Medications and Cardiovascular Risk — What the large-scale evidence shows about stimulant cardiac safety
• Complete ADHD Guide — Comprehensive resource on ADHD diagnosis, neurobiology, and treatment
• ADHD Psychiatrist NYC — Evaluation and medication management for adults and adolescents with ADHD