SOMA
Clinical twins for Alzheimer's research.

Studying patient similarity and cognitive change over time, with Alzheimer's disease as the focus. SOMA combines longitudinal research inputs with federated matching and privacy-preserving methods.

Patent Pending (U.S.) · Priority 2026

ADNI
Primary research cohort
Longitudinal Alzheimer's data
1
Disease focus
Alzheimer's disease
5
Research data types
Cognition, MRI, genetics, CSF & PET
Research
Current stage
Retrospective cohort evaluation

SOMA is an Alzheimer's research initiative by Curadai. The work uses retrospective research cohorts to investigate patient representations, cognitive trajectories, and privacy. It is a research foundation, not a clinical diagnostic service. The work is conducted independently of the institutions that contributed the research data.

The Problem

Understanding Alzheimer's progression takes more than one visit.

Alzheimer's researchers need to compare cognitive change across participants and time. Fragmented records, privacy requirements, and incomplete measurements make it difficult to identify participants with similar trajectories.

Data Fragmentation

Patient records siloed across institutions with incompatible EHR systems and clinical vocabularies.

Privacy Regulation

Research permissions and privacy requirements shape how sensitive participant data can be accessed and shared.

Static Snapshots

Existing approaches treat patients as frozen in time, losing critical trajectory and progression information.

Missing Modalities

Alzheimer's research visits may have cognitive scores without every scan, genetic measure, or biomarker.

Alzheimer's Research

What the Alzheimer's work investigates.

Seven research capabilities connect the participant's measurements, cognitive history, and potential peers. The scope is retrospective Alzheimer's cohort research; clinical usefulness requires further independent evaluation.

Patient Representation Research

The Alzheimer's work combines cognitive assessments, brain imaging, genetics, and biomarkers into patient representations. Evaluation examines whether nearby representations reflect meaningful differences between cognitively normal participants, mild cognitive impairment, and Alzheimer's dementia.

Privacy-Preserving Similarity Research

SOMA studies patient similarity across research sites while keeping source records local. The Alzheimer's privacy experiments examine protected population summaries and the effect of privacy noise on matching quality.

Longitudinal Matching Research

Repeated observations help distinguish participants with similar cognitive trajectories from those who merely look similar at one visit. The Alzheimer's work investigates progression from mild cognitive impairment and the value of temporal information for finding research peers.

Combining Research Inputs Research

Cognitive assessments, MRI, PET, genetics, and cerebrospinal fluid biomarkers provide different views of Alzheimer's disease. Learned fusion is evaluated against simpler combinations to study which inputs support useful patient similarity.

Incomplete Research Records Research

Alzheimer's research visits do not always include every assessment or scan. Missing-input experiments examine how patient representations and matching change when some of the available research data types are absent.

Cohort Comparison Research

ADNI provides the primary longitudinal setting. Comparisons with ROSMAP examine representation transfer within Alzheimer's and aging research, where measurements and participant populations differ. Research-cohort transfer is distinct from prospective clinical validation.

Continual Learning Ongoing

The Alzheimer's work studies whether incorporating observations from another site changes previously learned patient relationships. Retaining earlier structure and monitoring shifts in the research population remain ongoing evaluation questions.

What We Learned

What Alzheimer's research asks of the system.

Longitudinal cognitive change, incomplete visits, and differences between research cohorts shape how we evaluate patient similarity.

A useful representation comes first.

Patient matching depends on how well the representation captures Alzheimer's-related clinical structure. Privacy methods, temporal matching, and input fusion must be assessed against that foundation.

Cognitive change needs longitudinal context.

A single assessment cannot describe an entire trajectory. Repeated cognitive measurements and visit timing help frame the question of which research participants are progressing in similar ways.

Different inputs answer different questions.

Cognitive tests, imaging, genetics, and biomarkers should not be treated as interchangeable signals. Their contribution to Alzheimer's patient similarity needs to be measured for the specific research task.

Missing measurements need explicit evaluation.

Incomplete visits are part of longitudinal Alzheimer's research. Evaluate matching with the inputs actually available, and distinguish measured observations from any estimated values.

Generalization stays within the evidence.

Results depend on the cohort, available measurements, and evaluation setup. A promising retrospective Alzheimer's result is a reason for further study, not a substitute for independent or prospective evaluation.

Architecture

What happens to patient data at each stage.

The Alzheimer's research pipeline maps cognitive assessments, imaging, genetics, and biomarkers into representations for similarity analysis. Source records and the representations used for matching have separate privacy requirements.

Patient Records
N × D raw
Modality
Encoders
Per-type
Bottleneck
Fusion
N × 64
Unit
Hypersphere
S63
Similarity
Matching
cosine

In the Alzheimer's research configuration, cognitive scores, MRI, PET, genetic inputs, and cerebrospinal fluid biomarkers are encoded into a shared 64-dimensional representation. Similarity and temporal information support research matching; privacy controls govern what a participating site may share.

Network Topology

In the research architecture, site nodes prepare representations locally and the Soma Core compares Alzheimer's trajectories. The diagram and simulation illustrate the design; they do not show a live hospital network.

SOMA Central Brain Learned Fusion · Velocity Encoding · Trajectory Matching Protected representations · Source records stay local Protected representations · Source records stay local Blind Handshake (onboarding) Hospital A Edge Node Hospital B Edge Node Hospital C Edge Node Research Lab Edge Node Genomics Center Edge Node New Institution Joining... Active connection Onboarding Hospital Research

Dendrite Nodes

Research-site nodes prepare Alzheimer's patient representations from locally held assessments, imaging, genetics, and biomarkers. The architecture separates source records from the representations used for matching.

Soma Core

The core studies similarity and cognitive trajectories within Alzheimer's research. It brings together representations and temporal information to identify comparable research participants.

Blind Handshake

The onboarding protocol is designed to calibrate representations from another research site before matching. Access, compatibility, and privacy controls remain part of the evaluation.

Privacy Architecture

Alzheimer's experiments examine protected population summaries, privacy budgets, and their effect on similarity. Representations require privacy safeguards even when source records remain local.

Research Data

Grounded in Alzheimer's research cohorts.

ADNI provides the primary longitudinal research setting. ROSMAP provides a comparison within Alzheimer's and aging research. These cohorts support research evaluation, not a claim of a deployed clinical network.

ADNI · Primary Cohort

Alzheimer's Disease Neuroimaging Initiative data supports longitudinal research using cognitive assessments, brain MRI, genetics, cerebrospinal fluid biomarkers, and PET imaging.

ROSMAP · Research Comparison

ROSMAP provides an Alzheimer's and aging research comparison setting, including brain tissue gene-expression measurements. Different measurements and populations require a separate interpretation of transfer results.

Patent pending (U.S.) · Priority date 2026. The federated embedding architecture, Blind Handshake onboarding protocol, trajectory velocity encoding, and privacy-preserving aggregation mechanism are covered by pending patent claims. Institutions considering engagement should contact us for licensing terms.

What's Next

Deepening the Alzheimer's work.

The next steps stay focused on Alzheimer's disease: stronger longitudinal evaluation, independent research partnerships, and a clearer understanding of performance over time.

Deeper Alzheimer's Evaluation

Expand the assessment of cognitive trajectories, participant subgroups, and incomplete visits within the Alzheimer’s research scope.

Alzheimer's Research Partnerships

Explore independent evaluation with Alzheimer’s researchers, including site-specific data permissions, cohort compatibility, and prospective study design.

Longitudinal Monitoring

Study how changes in visit schedules, measurements, and research populations affect the stability of Alzheimer’s patient matching over time.

Alzheimer's research by Curadai

SOMA is the Alzheimer's research initiative of Curadai — a company building AI infrastructure for non-profit and healthcare institutions. Contact us to discuss Alzheimer's research collaboration, longitudinal data, or evaluation of federated patient similarity.