A new study published in Nature Communications shows that when the healthy brain shifts from rest to demanding tasks, large-scale networks reconfigure flexibly; in individuals with psychiatric diagnoses this reconfiguration is dampened, or “flattened,” and the shifts may align more with a person’s symptom mix than with a formal label.
Researchers examined how the brain coordinates information by analysing functional brain networks—groups of regions across the cortex and deeper structures that work together. Using functional magnetic resonance imaging, or fMRI, they estimated functional connectivity by tracking how closely activity in different regions rises and falls together. The brain, they say, operates as a dynamic system that continually adjusts these connections to meet changing demands.
Previous work has shown that networks reorganise during tasks rather than staying fixed at rest. A 2016 study found distinct patterns for simple motor actions versus complex working memory tasks, while a 2021 study led by Carrisa Cocuzza suggested that task-related connectivity shifts, though subtle, improve predictions of brain activity across dozens of cognitive conditions.
Linking these network shifts to mental illness is tricky because psychiatric diagnoses group together patients with varied symptom profiles, and different disorders can share common difficulties such as low motivation or fatigue. Some researchers argue for looking at broad dimensions of functioning rather than rigid categories, a view echoed by earlier studies showing brain connectivity patterns align more closely with broad symptom types than with diagnostic labels.
In an earlier cross-age study, brain connectivity patterns were more consistent within broad symptom categories—internalising versus externalising—than between them across development. Time-varying connectivity during rest has also distinguished patients with schizophrenia, bipolar disorder, or ADHD from healthy peers more effectively than static measures. Building on this, the researchers set out to see whether whole-brain network reconfigurations across resting and task scans could map individualized symptom profiles across health and illness.
The team analysed data from 219 adults aged 18 to 70, drawn from the Transdiagnostic Connectome Project, with 134 participants spanning 11 primary diagnostic groups—including depression, anxiety, bipolar disorder, PTSD, schizophrenia or schizoaffective disorder, substance use disorder, ADHD, OCD, panic disorder and eating disorders—plus 85 individuals without psychiatric history. To guard against data leakage, they split the sample into a discovery set of 176 and a held-out validation set of 43, and both data and analysis code were shared publicly as an exploratory proof-of-concept.
Participants underwent six fMRI scans across three mental states: three resting-state runs, two Stroop tasks requiring cognitive control, and one Emotional Faces task with fearful expressions. For each scan, activity was measured across 434 brain regions organised into 21 functional networks, and a machine-learning approach extracted underlying network building blocks to track how strongly the most efficient, well-segregated block was expressed across the six scans.
Across the discovery set, 108 participants with psychiatric diagnoses showed less variation from scan to scan than 68 participants without a diagnosis. A similarity score, ranging from -1 to 1, found that non-diagnosed people averaged 0.91, while those with a diagnosis averaged 0.95—indicating statistically significant flattening when tested against thousands of shuffled models. Flattened dynamics appeared across conditions and were most pronounced in panic disorder, eating disorders, OCD and ADHD, though sample sizes were small in these groups.
The researchers reduced 110 clinical, behavioural and cognitive measures into four dimensions via clustering: internalising (depression, anxiety, neuroticism), externalising (irritability, extraversion), cognition (intelligence and cognitive performance) and social or reward functioning (relationship experiences and novelty seeking). Each participant received a binary fingerprint across these four domains, yielding 16 possible symptom fingerprints that cut across traditional diagnostic labels.
Participants also completed six fMRI scans, and the team measured correlations across 434 regions to form network-based profiles. A classifier assessed how well brain dynamics could predict three outcomes in the validation sample: case-versus-control status, primary diagnosis, and the 16 symptom fingerprints, performing above chance for all three—best for case status and notably stronger for symptom fingerprints than for diagnosis.
To probe the brain systems driving fingerprint classification, the researchers sequentially removed one functional network at a time and re-tested. Excluding networks involved in cognitive control—the frontoparietal control network and the thalamus—as well as some limbic, visual and somatomotor networks, caused a statistically significant drop in accuracy. Removing a subset of the default mode network slightly increased accuracy, though not significantly; the team suggests default mode and frontoparietal systems may work together when shifting between easy and demanding mental states.
Among seven scan combinations, the model using all six scans produced the lowest fingerprint accuracy, implying that simply adding more data does not linearly improve classification. The best-performing model combined resting-state and Stroop scans, with frontoparietal networks, the thalamus and the striatum acting as key hubs; the authors propose that how flexibly the brain moves from rest to cognitive control states could be central to transdiagnostic symptom profiles.
These findings align with earlier work linking altered time spent in certain brain states to psychiatric symptoms in adults with psychotic-like experiences and in children with autism, though those studies used different methods and populations. They also contrast with a prior resting-state study of antidepressant users, where dynamic network changes varied with medication rather than diagnosis—differences likely reflecting how dynamics were measured in the different studies.
Several caveats accompany the work. Because all data were collected in a single observational window, the study cannot determine whether flattened network dynamics cause psychiatric symptoms or arise from living with them. Symptoms fluctuate, underscoring the need for longitudinal studies that also track rapid trial-by-trial shifts within tasks.
Other limitations include that the validation sample was small and some diagnostic groups contained only a few participants. Medication use was not reported, leaving open whether observed flattening reflects illness, treatment, or both. Potential confounds such as head movement, breathing or alertness during scans, and scanner-site differences may also influence results despite careful corrections.
Nevertheless, the study offers a promising framework for examining how flexible brain coordination relates to mental health beyond traditional diagnostic boundaries, suggesting that patterns of network dynamics may better reflect individual symptom profiles than categorical diagnoses.
The study, Brain network dynamics reflect psychiatric illness status and transdiagnostic symptom profiles across health and disease, is by Carrisa V. Cocuzza, Sidhant Chopra, Ashlea Segal, Loïc Labache, Rowena Chin, Kaley Joss and Avram J. Holmes.
