← All modules
AURORA-HYDRO/Deep dive/Private pilot

AURORA-HYDRO is an open, federated software stack for the multidisciplinary team caring for a patient with hydrocephalus.

Hydrocephalus is the most common neurosurgical condition worldwide. AURORA-HYDRO models the cerebrospinal compartment as a coupled mechanical system across formation, decision-making and lifetime monitoring.

What it is
A software substrate. 3 subsystems on one federated runtime, installed inside the hospital — not a cloud service, not a black box.
What it does
AURORA-HYDRO models the cerebrospinal compartment as one continuous problem from birth to old age. Three subsystems cover formation physics, the shunt-vs-ETV decision, and lifetime compartment surveillance.
Who uses it
The paediatric · adult neuro · icu · bioengineering MDT at pilot partner institutions today. Open to any hospital under MIT at public alpha (Q4 2026).
Why it exists
So a hydrocephalus case at any hospital benefits from the substrate the top centres already have. Same code, same model cards, same audit story everywhere.
Subsystems
3
all spec-frozen
Pilot sites
2 paediatric units · 1 ICU group
scoping → live
Working papers
3
drafts in flight
Latency
target <80ms
median per case
License
MIT
@ public alpha
Stage
Pilot
public alpha Q4 2026
I · The clinical surface

What clinicians actually see.

A non-functional preview of the AURORA-HYDRO surface. Composable panels, logged overrides, explainable end-to-end. Pilot sites tune this to their own workflow.

FidelityPixel-accurate mock — no live data.
SubstrateReact 18 · CSS custom properties · no framework lock-in.
ScopeVocabulary for all AURORA-* surfaces.

MRN-880412 · Aqueductal stenosis · 11mo

live2 paediatric units override · 2
Stage
Compartment model in alpha
Subsystems
3/3 ready
Latency
target <80ms
Audit log
on
Primary modality · overlayupdated 2s ago
10 mm
σ = 1.4 mm·ROI 24.7 cm³
overlay field
plan draft
Subsystem signalsupdated 2s ago
HYDRO-COMPFORM   Compartment Formation84%
HYDRO-COMPDECIDE   Compartment Decision71%
HYDRO-COMPLIFE   Compartment Lifetime62%
ETV success calibration
Brier −0.04
24-mo revision risk
21%
II · The clinical case

Why an open module for hydrocephalus.

AURORA-HYDRO models the cerebrospinal compartment as one continuous problem from birth to old age. Three subsystems cover formation physics, the shunt-vs-ETV decision, and lifetime compartment surveillance.

Hydrocephalus is the most common condition treated by neurosurgeons worldwide. The compartment is governed by a small set of physical laws that most clinical systems do not model coherently. AURORA-HYDRO treats CSF dynamics as one continuous problem across the lifespan — congenital, post-haemorrhagic, post-infectious, and normal-pressure.

The shunt-versus-ETV question is well-studied — the ETVSS captures much of the easy signal — but personalised priors remain rare. AURORA-HYDRO carries a compartment model into the decision so success probability is grounded in physics, not just population statistics.

Lifetime compartment surveillance is the most under-served piece. Most patients with a shunt will need at least one revision. AURORA-HYDRO's lifetime-watch subsystem reduces the chance that a slow trajectory becomes a late emergency.

We are not building a product around Hydrocephalus. We are seeding an infrastructure — so that any hospital with the will to use it can.
HYDRO · CSF COMPARTMENTS

One compartment problem, lifetime-long.

AURORA-HYDRO treats the cerebrospinal compartment as one continuous problem from birth to old age. The shunt-vs-ETV branch shares a calibrated prior — ETVSS-aligned, with patient-specific flow signals layered on top.

VENTRICULAR · 1.2 mL/minDECIDEOPTION AShuntdurable · revision-proneOPTION BETVavoids hardware · ETVSS-priorsP(success)0.41P(success)0.59
HYDRO-COMPFORM · the physics, fed by patient-specific flow signals.
HYDRO-COMPDECIDE · the branch, with ETVSS-aligned priors and Brier-calibrated probabilities.
HYDRO-COMPLIFE · the revision-risk forecast, decades out.
III · The stack

3 subsystems. Each one independently useful.

Every AURORA-HYDRO subsystem can be adopted alone or as part of the bundle. Each ships with its own model cards, eval results and a one-command install.

HYDRO · 01 live

Compartment Formation

HYDRO-COMPFORM

CSF production, absorption and pulsatile transmantle pressure model.

CSF production / absorption / pulsatile pressure model. Patient-tuned during onboarding.

HYDRO · 02 live

Compartment Decision

HYDRO-COMPDECIDE

Shunt vs. ETV decision support with personalised flow priors.

Shunt vs. ETV decision support with personalised flow priors and ETVSS-aligned baselines.

HYDRO · 03 beta

Compartment Lifetime

HYDRO-COMPLIFE

Lifetime shunt surveillance, infection and revision forecasting.

Lifetime shunt surveillance — infection, malfunction and revision forecasting from telemetry-light signals.

IV · The product, in detail

What AURORA-HYDRO produces, end-to-end.

Inputs the module reads from your existing systems, outputs it returns to them, the protocols it speaks, and the lifecycle of one case as it moves through AURORA-HYDRO.

01Inputs4 types
Imaging
Modality set defined by the module's RFC.
DICOM 3.0
Records
Hospital records and protocol metadata, scoped by consent cohort policy.
FHIR R4 / HL7v2
Clinical
Operative notes, outcome forms, longitudinal events as available.
FHIR R4
Audit seed
Consent metadata, prior override log if migrating from another system.
AURORA audit (NDJSON)
02Outputs4 artefacts
Integrated assessment
Module-specific structured report combining 3 subsystem signals.
Signed PDF + JSON
Imaging derivatives
Per-voxel maps, ROI segmentations and uncertainty masks as relevant.
DICOM SEG · NIfTI
Plan draft
Treatment plan or follow-up draft for clinician review and signature.
Module-specific structured doc
Audit envelope
Every output ships with a hash chain back to inputs, weights and config.
AURORA audit (NDJSON)
03Case lifecycle6 steps · median target <80ms
  1. 01
    Bind
    Imaging or record arrives at PACS/EHR. AURORA binds the case by SOP/MRN, checks consent policy.
  2. 02
    Foundation pass
    Module foundation model produces a first-pass representation. Median target <80ms on partner-site hardware.
  3. 03
    Subsystem fan-out
    3 subsystems run in parallel; each writes its signal + uncertainty to the case.
  4. 04
    Synthesise
    Subsystem outputs are fused into the integrated assessment; the case is queued for the MDT surface.
  5. 05
    Clinician review
    MDT reviews on the AURORA surface. Every override carries a reason; the audit log writes durably before the next render.
  6. 06
    Sign + emit
    Signed report + structured artefacts flow back to PACS / FHIR / radiotherapy planning systems. Hash chain closes.
04Integrations the module speaks6 endpoints
PACS
DICOM C-STORE inbound + C-FIND outbound · STOW-RS for derived series.
EHR
FHIR R4 — DiagnosticReport, Condition, Observation, ServiceRequest. OAuth2 client credentials.
Lab feeds
HL7v2 (ORU^R01) for legacy molecular and laboratory results.
Auth
OIDC — Microsoft AD FS, Okta, Keycloak. Per-clinician identity in every override log entry.
Provenance
Sigstore-signed weights + container digests. Per-site federation key in Vault or KMS.
Telemetry
Off by default. When enabled, opt-in per-site; payload schema is in the module RFC.
05Per-subsystem, in detail3 units · independently installable
HYDRO-COMPFORM
Compartment Formation
live

CSF production / absorption / pulsatile pressure model. Patient-tuned during onboarding.

HYDRO-COMPDECIDE
Compartment Decision
live

Shunt vs. ETV decision support with personalised flow priors and ETVSS-aligned baselines.

HYDRO-COMPLIFE
Compartment Lifetime
beta

Lifetime shunt surveillance — infection, malfunction and revision forecasting from telemetry-light signals.

Reversibility
Nothing the module installs is destructive. Uninstall removes containers + audit pointer; underlying records are untouched.
Determinism
Pinned commit + container digest + weights hash + dataset hash. Every AURORA-HYDRO run is re-runnable and re-attributable.
Air-gap parity
Air-gapped deployments produce byte-identical outputs to networked deployments on identical inputs. No silent telemetry, ever.
Override SLO
Every clinician override is logged ≤ 200ms after the action; the log is durable before the next prediction renders.
V · Working papers

Our own writing on AURORA-HYDRO.

3 drafts by the module team and pilot collaborators. Each documents one slice of the substrate — methods, evaluation protocol, lessons. Slot PDFs in as they're ready; the entries point to placeholders until then.

⚠ DRAFTS · DOI-STAMPED AT PUBLIC ALPHA · NO EXTERNAL "EVIDENCE LIBRARY" SHIPPED ON THIS PAGE
VI · Endpoints we will track

The metrics that matter over a lifetime.

These are the endpoints AURORA-HYDRO is built to measure across pilot deployments. The targets below are pilot goals, not retrospective results.

−31%
target
Shunt revisions @ 24mo

On networks running lifetime watch (target).

+9%
target
ETV success calibration

Brier-score improvement vs ETVSS baseline.

61ms
target
Avg model latency

End-to-end across compartment models.

−40%
target
Time-to-emergency-revision

Earlier triage via lifetime watch.

100%
target
Auditable override rate

Every override logged with reason.

+12%
target
iNPH responder identification

Better selection of shunt-responsive iNPH candidates.

⚠ PILOT TARGETS · NOT RETROSPECTIVE RESULTS · TO BE VALIDATED AT PUBLIC ALPHA
VII · Install AURORA-HYDRO

From pip install to a hospital deploy in one afternoon.

Available to pilot partners today on private registries. At public alpha (Q4 2026) the same images, weights and signatures ship under MIT on public registries. For the full per-product download grid, see the unified download page.

step 1 of 3
pip install aurora-hydro
step 2 of 3
aurora doctor hydro
# → AURORA-HYDRO  ✓ python  ✓ compartment model
# → three subsystems loaded
# → ready in target ≤80ms
step 3 of 3
from aurora.hydro import decide
plan = decide(case_id="bench/hydro/case-0014")
print(plan.recommend)         # → "ETV"
print(plan.success_prob)      # → 0.71 ± 0.06
Hardware
CPU · NVIDIA · AMD · Apple silicon
OS
Linux · macOS · Windows · WSL
Deploys
Cloud · On-prem · Air-gapped · Edge
Telemetry
Opt-in · Off by default

Hardware footprint, in practice

AURORA-HYDRO is designed to run on hardware that already exists inside hospital networks. CPU-only inference is supported for the lighter subsystems; the heavier ones benefit from a recent GPU but do not require one. A single modern workstation handles routine cases under the latency target.

Networking and consent

Federated deploys do not require open inbound ports. The runtime opens an outbound mTLS connection to the federation control plane; model updates are signed, audited and pulled. Patient data never traverses the federation. Consent metadata is a first-class object — subsystems with an explicit consent dependency refuse to run on records that lack the appropriate cohort policy.

Migration paths

Most pilot sites land on AURORA-HYDRO with an existing system in place. The recommended path is: install alongside, compare outputs on a held-out cohort for one quarter, then move read-only surfaces to AURORA, and decide on the rest. Reversibility is a design goal — nothing in the install creates lock-in.

Download grid

For the full per-product, per-platform download buttons (macOS / Windows / Linux installers + pip / docker / helm / rust commands), see the unified download page. Each subsystem of AURORA-HYDRO is independently downloadable there.

VIII · Questions

The honest questions on AURORA-HYDRO.

Eight diseases is the beginning.

AURORA is in private pilot today and opens to the world at public alpha in Q4 2026 under MIT. If you carry the weight of these diseases — as a clinician, scientist, builder, patient or advocate — there is a seat at the table.

Changelogv0.6.0