Functional MRI is one of the most misunderstood scans in medicine. It sounds as though it should be the definitive test for how a brain is working - depression, ADHD, autism, memory, the lot. In UK clinical practice in 2026, that is not what it does. It is a technically demanding scan used by a small number of neurosurgical teams to map the specific bits of a patient's brain that must be preserved during surgery. Everything else you have read about fMRI is almost certainly research, not clinical care.
This piece is what a functional MRI actually measures, what it is used for on the NHS and privately in the UK, what it is not used for, and what to expect if a consultant does request one for you.
One-line answer: what is a functional MRI?
A functional MRI is a brain scan that maps where activity is happening across the brain, in real time, by detecting tiny local changes in blood oxygenation. It does not measure neurons firing directly. It measures the blood-flow response that follows neural activity a few seconds later, and it uses that signal to build a map of which brain regions light up when a person performs a task or lies still. In UK clinical practice its one job is to help neurosurgeons operate around eloquent brain tissue without damaging it.
What fMRI actually measures: the BOLD signal
The signal that makes fMRI work is called the BOLD signal, short for Blood-Oxygen-Level Dependent. It is an indirect proxy for neural activity, not a direct recording of it. When neurons in a region become more active, that region demands more oxygen. The brain over-supplies fresh oxygenated blood to meet that demand, and the ratio of oxygenated to deoxygenated haemoglobin in that patch of tissue shifts. Deoxygenated haemoglobin has slightly different magnetic properties from the oxygenated form, and the MRI scanner is sensitive enough to detect the change.
The important honesty here is that fMRI does not show thoughts, memories or emotions. It shows regional changes in cerebral blood oxygenation, correlated with neural activity, averaged over several seconds and several cubic millimetres of tissue. It is a real biological signal, and a powerful one, but it is a haemodynamic proxy. Popular writing that treats fMRI as a mind-reader is not what the scan does.
The one clinical use: pre-surgical brain mapping
In UK hospitals in 2026, the overwhelming majority of clinical fMRI is performed for one reason: to map eloquent cortex before brain surgery. If a neurosurgeon is planning to remove a tumour or an epileptic focus that sits close to areas the patient cannot afford to lose - the primary motor cortex that moves a hand, Broca's or Wernicke's area for language, the visual cortex, verbal memory regions in the medial temporal lobe - the surgical team needs a patient-specific map. Textbook anatomy is not enough because tumours and long-standing epilepsy displace and reorganise function.
Pre-surgical fMRI shows the surgeon which side of the brain is dominant for language in this specific patient, where hand-motor cortex actually sits relative to the tumour, and how close the resection margin can safely go. It is planned alongside diffusion tensor imaging for the white-matter tracts and, in some cases, intraoperative awake mapping. It is a serious piece of surgical planning, not a screening scan.
| Clinical scenario | Is fMRI used? | What the scan answers |
|---|---|---|
| Brain tumour adjacent to motor or language cortex | Yes, routinely | Where is the eloquent cortex in this patient? |
| Drug-resistant epilepsy work-up for surgery | Yes, in specialist centres | Which hemisphere is language-dominant? Where is verbal memory? |
| Arteriovenous malformation near eloquent areas | Sometimes | Can the AVM be treated without functional deficit? |
| Depression, anxiety, ADHD, autism | No (research only) | Not a clinical diagnostic test in 2026 |
| General "brain health" or memory check | No | Not a validated clinical indication |
Task-based vs resting-state fMRI
There are two flavours of fMRI, and it matters which one your neurosurgical team requests.
Task-based fMRI is the traditional approach. The patient performs a specific paradigm in the scanner - finger-tapping to map motor cortex, silent verb generation to map language, picture naming, a memory encoding task. The scanner captures the BOLD signal during task and rest blocks, and the software subtracts one from the other to produce a map of active regions. It is highly reliable for motor and language mapping when the patient can cooperate.
Resting-state fMRI asks nothing of the patient beyond lying still with eyes open or closed. It looks for coherent, low-frequency fluctuations in the BOLD signal across networks of brain regions. In pre-surgical practice it is increasingly used for patients who cannot perform tasks reliably - young children, patients with aphasia from a tumour, sedated patients. It is a useful clinical tool in expert hands, but the analysis is more complex and the interpretation is done by a neuroradiologist trained in the technique.
Where fMRI is not a diagnostic test
This is the section most patients need. There is a large and growing consumer market for "brain health" MRI scans, and some of that marketing implies or states outright that fMRI can diagnose psychiatric or neurodevelopmental conditions. In 2026 UK clinical practice, that is not accurate.
There is no validated, regulator-approved fMRI test that diagnoses depression, anxiety, ADHD, autism, bipolar disorder, PTSD or any other psychiatric condition in an individual patient. The research literature contains many interesting group-level findings - averages across hundreds of patients that differ from averages across healthy controls - but those group differences do not translate into a diagnostic scan you can send one person into. The signal-to-noise, the individual variability and the overlap between clinical and non-clinical brains are all too large.
The same is true of dementia screening. Structural MRI, PET and cerebrospinal fluid biomarkers are the imaging tools with a clinical evidence base for cognitive decline. Functional MRI has a research role, not a diagnostic one, in Alzheimer's disease and related dementias in 2026.
The best defence a patient has is a simple question: is the scan you are selling me approved for this condition, and will the report be signed by a consultant neuroradiologist? For fMRI outside pre-surgical mapping, the honest answer today is usually no.
What patients experience in the scanner
A clinical fMRI appointment usually runs to around 60 minutes in the scanner, plus 20 to 30 minutes of set-up beforehand for consent, task briefing and practice runs on a laptop outside the room. You lie on your back on the scanner bed, your head goes into a padded coil that has a small mirror above your eyes, and you look at a screen at the end of the bore that shows the task instructions.
