Enrico Amico and Dimitri Van De Ville on the Brain Fingerprint That Survives Disease
Does a person’s brain fingerprint survive a disease that changes their brain? Enrico Amico, Dimitri Van De Ville and colleagues in Geneva and Lausanne tested this in Alzheimer’s disease. In two cohorts with 126 people in total, every individual could be identified from their own functional connectivity pattern, in healthy older adults, in people with mild cognitive impairment, and in people with Alzheimer’s dementia. The connections that made each person identifiable were different at each stage. The paper is Sara Stampacchia and colleagues, Fingerprints of brain disease, connectome identifiability in Alzheimer’s disease, Communications Biology 7, 1169, September 2024 (doi:10.1038/s42003-024-06829-8).
For a theory that ties a person to a pattern of brain dynamics, this is a direct test. The pattern stays individual while the brain that carries it degenerates, and the parts of the pattern that carry the individuality move.
Measuring how identifiable a brain is
A functional connectome is a matrix of correlations between the activity of brain regions, usually measured with fMRI. Emily Finn and colleagues showed in 2015 that a person’s connectome can identify them among more than a hundred others. Enrico Amico and Joaquín Goñi at Purdue then turned identification into a set of measures (Scientific Reports 2018, doi:10.1038/s41598-018-25089-1).
| Measure | What it compares |
|---|---|
| I-self | How similar a person’s connectome is to their own second scan |
| I-others | How similar a person’s connectome is to other people’s scans |
| I-diff | The difference between the two, how much more a person resembles themselves than others |
| Success rate | The share of people correctly matched to their own second scan |
The same 2018 paper showed that rebuilding each connectome from a limited number of group-level connectivity modes increases identifiability across rest and seven tasks.
When identity shows up in time
Dimitri Van De Ville, Younes Farouj, Maria Giulia Preti, Raphaël Liégeois and Enrico Amico asked at which time scale a fingerprint appears (Science Advances 2021, doi:10.1126/sciadv.abj0751). The best identification emerged at longer time windows. Short “bursts of identifiability” linked to neuronal activity were also present at shorter windows. Different parts of the fingerprint belonged to different time scales, with visual and somatomotor connections at short windows and frontoparietal and default mode connections at longer ones.
The fingerprint in Alzheimer’s disease
The 2024 study used two independent cohorts, 54 people from Geneva and 72 from the ADNI database. Each cohort had three groups, cognitively unimpaired people without amyloid, amyloid-positive people with mild cognitive impairment, and people with Alzheimer’s dementia. The team split each fMRI session into two halves and tested whether each person’s first half matched their own second half.
- The success rate of identification was 100% in both cohorts and all three groups. Every person resembled themselves more than anyone else.
- Between-person similarity was lower in the patient groups than in unimpaired people, so patients differed more from each other.
- The edges that most reliably identified a person had a different layout in each group, a reconfiguration of the fingerprint with disease stage.
- The authors argue that studies of Alzheimer’s disease should focus on individual variability as well as group differences.
A related clinical study by Natalie Voets and colleagues in Oxford used connectivity fingerprints in 44 patients with a tumor in the language-dominant hemisphere and 23 healthy volunteers (NeuroImage Clinical 2019, doi:10.1016/j.nicl.2019.101952). Patients showed highly variable fingerprints that distinguished each individual, shaped by tumor grade and location, which the authors propose as a way to measure plasticity in a single patient.
What this tells us about consciousness
Our consciousness fingerprint page asks whether a person’s identity can be found in their brain dynamics. These studies add two facts. A person stays identifiable in a brain that is losing tissue, and the parts of the network that carry that identity change as the disease advances. Identity, on this evidence, is a property of the whole pattern, and different connections can carry it at different times.
The fingerprint page also states the main limit. A connectome fingerprint marks the person and not the conscious state. Fingerprints are also present in sleep and sedation, so an unchanged fingerprint does not show that experience is unchanged. For questions about consciousness it has to be paired with a measure of the conscious state itself, such as the perturbational complexity index.
This is why the Neutral Core slow handover section names the fingerprint as a candidate check at each step of a handover, alongside a state measure. If tissue were replaced gradually, the fingerprint before, during and after each step should stay closer to the person’s own than to anyone else’s. That is a testable requirement, though no study has applied it to a planned replacement.
Comparison to The Consciousness AI
The project’s oscillator module, AKOrN, implements oscillatory binding, and the fingerprint page records that a pre-registered prediction about it was never confirmed. The project has not tested whether its models produce an individual fingerprint at all. The Alzheimer’s result gives a concrete standard for any synthetic system that claims to continue a person. Its activity should be matched to that person by the I-self and I-diff measures, at the time scales where human fingerprints appear. The Core is a simulation and has not been tested against a human fingerprint.
Limits of the brain fingerprint evidence
Enrico Amico, Dimitri Van De Ville and colleagues showed that brain fingerprints stay individual in Alzheimer’s disease while their spatial layout changes, and they built the measures that make such claims testable. The limits are specific. The Alzheimer’s study measured identifiability within a single session, by splitting each scan in half, because separate sessions were not available. It compared groups at different stages and did not follow the same people over time. The tumor study was cross-sectional too. Fingerprints are traits and say nothing direct about awareness. Within those limits, connectome fingerprinting is the best available tool for checking that a person is still the same person in a changing brain, and it belongs with the measurement debates in how scientists race to define and test AI consciousness in 2026.
Researchers covered here
- Enrico AmicoSchool of Mathematics and Centre for Human Brain Health, University of BirminghamThe differential identifiability framework for brain fingerprints, connectome fingerprinting in Alzheimer's disease, time scales of brain identifiability
- Dimitri Van De VilleNeuro-X Institute, EPFL, and Department of Radiology and Medical Informatics, University of GenevaDynamic brain fingerprints and the time scales of identifiability, brain signal processing for fMRI, connectome fingerprinting in Alzheimer's disease