← All modules
AURORA-GLIO/Module 01 of 08/Deep dive/Private pilot

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

It reads the imaging, molecular and surgical inputs a real case already produces, runs them through eleven subsystems on a shared substrate, and returns four artefacts: an integrated diagnosis aligned to WHO 2021, an invasive-margin field with calibrated uncertainty, a draft radiotherapy plan, and a list of eligible trials. Every output is explainable and every clinician override is logged.

What it is
A software substrate. 11 subsystems on one federated runtime, installed inside the hospital — not a cloud service, not a black box.
What it does
Glioblastoma and lower-grade gliomas remain the deadliest primary brain cancers.
Who uses it
The neuro-oncology · radiology · radiation oncology · computational biology MDT at pilot partner institutions today. Open to any hospital under MIT at public alpha (Q4 2026).
Why it exists
So a glioma case at any hospital benefits from the substrate the top centres already have. Same code, same model cards, same audit story everywhere.
Subsystems
11
all spec-frozen
Pilot sites
3 pilot sites scoping
scoping → live
Working papers
11
drafts in flight
Latency
target <150ms
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-GLIO 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.

Case MRN-487219 · GBM IDH-WT · MGMT methylated

livepilot · OXON override · 2
Stage
Pre-op MDT
Subsystems
11/11 ready
Latency
target <150ms
Audit log
on
T1-Gd · axial · slice 24/41 · margin overlayupdated 2s ago
10 mm
σ = 1.4 mm·contrast-enhancing GTV 24.7 cm³
infiltration field
contrast-enhancing tumour
resection draft
Subsystem signalsupdated 2s ago
GLIO-AI   Foundation Models92%
GLIO-DIAG   Integrated Diagnosis88%
GLIO-INFIL   Infiltration Field79%
GLIO-RAD   Radiotherapy Plan71%
GLIO-CLIN   Clinical Trials64%
Molecular profile · GLIO-DIAGsigned report
IDH
wildtype
conf 0.99
MGMT
methylated
conf 0.83
TERT
C228T mutant
conf 0.97
EGFR
amplified
conf 0.94
Chr 7/10
+7 / −10
conf 0.91
WHO 2021
GBM, IDH-WT
conf 0.96
Margin field σ over time
σ = 1.4 mm
Cohort-relative recurrence risk
68th percentile
II · Clinical context

The standard of care, and where AURORA-GLIO fits in.

A short, plain account of glioma today — the parts of the pathway that are settled, the parts that are still institution-specific, and the seam AURORA-GLIO is built to fill.

Glioma is the most common malignant primary brain tumour in adults. The pathway has a well-established shape: imaging, biopsy or resection, an integrated histo-molecular diagnosis, then radiotherapy and temozolomide for glioblastoma — the Stupp protocol from 2005. Lower-grade gliomas follow a parallel pathway with longer surveillance. None of this is in dispute, and AURORA-GLIO does not try to rewrite it.

The central problem is heterogeneity — between patients, and inside a single tumor. Single-cell work over the last decade has shown that a glioblastoma is not one tumor cell-type behaving badly; it is a network of cellular states that drift, interconvert and re-form after treatment. The radiomic signature does not capture this. The surgical microscope does not capture this. The trial-arm allocation does not capture this. AURORA-GLIO is an attempt to put the cell-state map, the imaging signature, the surgical reality and the longitudinal outcome into one substrate that any hospital — not just a top-tier academic centre — can run.

Why we wrote our own substrate

Standard-of-care references — Stupp, MGMT, the WHO 2021 framework, the TCGA-era genomic landscape, Sanai on extent-of-resection — are the floor; they are not in dispute. What does not exist is a shared substrate that lets a hospital combine all of that with single-cell state maps, radiomic foundation models, and methylation classifiers without rebuilding the plumbing from scratch.

AURORA-GLIO is our working answer. Eleven subsystems sharing one federated runtime. The molecular layer talks to the imaging layer. The imaging layer talks to the surgical layer. The surgical layer talks to the trial-routing layer. Every output explains itself. Every override is logged. Patient data never leaves the institution. The methods, design choices and evaluation protocols are written up in our own working papers — drafted by the module team and pilot collaborators, listed further down this page.

We are not trying to "solve" glioma. We are trying to make sure that the next ten years of glioma work is additive — across institutions, across modalities, and across the decade.

Why this module is the flagship

AURORA-GLIO is the deepest pilot module not because gliomas are the easiest disease to address — they are arguably the hardest — but because the field's infrastructure debt is the most expensive. A federated, open, audit-first substrate has to prove itself on the disease that has historically refused to yield to closed, single-institution AI. If AURORA works for gliomas, the rest of the modules become straightforward port-and-tune work.

GLIO · EXTENT-OF-RESECTION × OUTCOME

Marginal resection is the variable.

AURORA-GLIO models the Sanai-curve relationship between extent-of-resection and outcome, with the margin-field uncertainty propagated through. The current case sits on the curve as a band, never a point.

EXTENT OF RESECTION →SURVIVAL % →CASE MRN-487219 · 84% EOR · σ ±4.2%
GLIO-INFIL · the curve's x-axis carries margin-field σ as a band.
GLIO-RAD · the y-axis is dose-aware, not just resection-aware.
GLIO-CAUS · the band, not the point, is what reaches the tumour board.
III · The stack

Eleven subsystems. Each one independently useful.

Every AURORA-GLIO subsystem can be adopted alone or as part of the bundle. The numbering reflects roughly the order they fire on a real case — from systems-atlas priors through to the organoid digital twin.

GLIO · 01 live

Systems Atlas

GLIO-SYS

Multi-omic systems map of glial lineages, niches and signalling.

Built on TCGA-era genomic landscape work and single-cell state maps (Patel 2014, Neftel 2019); the priors most other GLIO subsystems consume.

GLIO · 02 live

Macro-Micro Bridge

GLIO-MMB

Connects scanner-scale imaging with cell-scale histology features.

Macro-to-micro bridge: connects MRI radiomic signatures to histopathology features and spatial transcriptomics at the case level.

GLIO · 03 live

Tumor Microenvironment

GLIO-TME

Immune-stromal-vascular crosstalk simulator with spatial priors.

Immune-stromal-vascular crosstalk simulator with spatial priors from public neuro-oncology atlases.

GLIO · 04 live

Foundation Models

GLIO-AI

Pre-trained vision-language-omics transformer for glioma.

Per-disease vision-language-omics foundation model. Distilled to edge sizes so it runs on clinic hardware.

GLIO · 05 live

Infiltration Field

GLIO-INFIL

Diffusion-reaction PDE solver for invasive margins beyond contrast.

Reaction-diffusion PDE solver for invasive margins beyond the contrast-enhancing region. Differentiable inverse problem on patient data.

GLIO · 06 beta

Causal Trajectories

GLIO-CAUS

Counterfactual graphs over treatment, resection and progression.

Counterfactual graphs over treatment, resection extent and progression — explicit about which comparisons are confounded.

GLIO · 07 live

Integrated Diagnosis

GLIO-DIAG

WHO-aligned molecular + histo + radio diagnostic synthesizer.

WHO-2021-aligned diagnostic synthesiser combining histology, molecular and imaging into one signed report.

GLIO · 08 beta

Radiotherapy Plan

GLIO-RAD

Auto-contoured GTV/CTV/PTV with dose-response priors.

Auto-contoured GTV/CTV/PTV with dose-response priors; consumes the infiltration field directly.

GLIO · 09 beta

Clinical Trials

GLIO-CLIN

Eligibility matcher and adaptive arm allocator for active studies.

Trial eligibility matcher and adaptive-arm allocator for active glioma studies.

GLIO · 10 live

Ethics & Governance

GLIO-ETHIC

Bias, consent and equity audit layer with provenance reports.

Bias, consent and equity audit layer with provenance reports. Runs inside the training loop, not after release.

GLIO · 11 soon

Organoid Twin

GLIO-ORG

Patient-derived organoid digital twin synced with bench data.

Optional patient-derived organoid digital twin, synced with bench data for consenting patients.

IV · The product, in detail

What AURORA-GLIO 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-GLIO.

01Inputs6 types
Imaging
Multi-parametric MRI — T1, T1c, T2, FLAIR are required. DWI/ADC and DSC perfusion improve the foundation pass but are not gating.
DICOM 3.0 · 1.5–3T · isotropic preferred
Molecular
WHO-2021 panel: IDH1/2, MGMT methylation, 1p/19q, TERT, EGFR, ATRX. Methylation classifier optional and wired in when present.
HL7v2 ORU · FHIR Observation · Idylla / NGS pipelines
Histology
Whole-slide scans for tumour board cases; AURORA-GLIO consumes them through GLIO-MMB to ground radiomic signals against cellular state maps.
OpenSlide-readable WSI (SVS, NDPI, MRXS, BIF)
Surgical
Operative notes, EOR estimates from immediate-post-op MRI, intraop video if recorded. None are gating; all improve longitudinal performance.
FHIR Procedure + free-text notes
Trial registry
Active glioma studies and their molecular/imaging eligibility criteria, refreshed nightly from ClinicalTrials.gov + EU CTR + ANZCTR.
FHIR ResearchStudy (mirrored)
Audit seed
Cohort consent metadata, prior override log if migrating from another platform. AURORA refuses to render predictions on records lacking consent.
AURORA audit (NDJSON)
02Outputs6 artefacts
Integrated diagnosis
WHO-2021-aligned histo-molecular-radio synthesis. Includes confidence intervals per axis and the audit chain back to source assays.
Signed PDF + JSON · attached to PACS as SR
Margin field
Per-voxel infiltration probability with calibrated σ, beyond the contrast-enhancing volume. Solver state is checkpointed for re-runs.
DICOM SEG + NIfTI + JSON sidecar
Surgical-corridor map
Candidate corridors annotated with eloquence priors and per-corridor risk envelopes. Surgeon edits live on the AURORA surface.
DICOM RTSTRUCT · JSON graph
Radiotherapy draft
Auto-contoured GTV/CTV/PTV with dose-response priors; OAR contours flagged. Always reviewed before submission to the RT planning system.
DICOM-RT struct set · RTPLAN draft
Trial routing
Ranked list of eligible active trials with per-arm fit scores, contact data, and the molecular axes that gated the rank.
FHIR ServiceRequest · JSON
Audit envelope
Hash chain: inputs → weights → config → outputs. Every run is re-runnable and the chain is signed by the site's federation key.
AURORA audit (NDJSON + Sigstore)
03Case lifecycle9 steps · median target <150ms
  1. 01
    Bind (≤1s)
    Imaging study lands on PACS. AURORA-GLIO binds the case by SOP UID, checks the cohort consent policy, fetches molecular/histology if available.
  2. 02
    Foundation pass (8–12s)
    GLIO-AI runs the vision-language-omics transformer on the available modalities. The output is a dense case representation, not a label.
  3. 03
    Macro–micro grounding (3–6s)
    GLIO-MMB aligns the radiomic signal against histology and spatial transcriptomics where present; reports a discrepancy if scales disagree.
  4. 04
    Margin solver (4–7s)
    GLIO-INFIL solves the reaction-diffusion PDE on the patient's anatomy. Differentiable inverse problem; σ is part of the solution.
  5. 05
    Integrated diagnosis (1s)
    GLIO-DIAG fuses histology + molecular + radio into the WHO-2021 axes. Discrepancies surface as overrides for the pathologist to resolve.
  6. 06
    Plan drafts (5–9s)
    GLIO-RAD drafts radiotherapy contours constrained by σ; GLIO-CLIN ranks active trials by per-arm eligibility. Both are decision-support, not orders.
  7. 07
    Equity gate
    GLIO-ETHIC checks stratified performance on this site's cohort. If any stratum is below threshold, the case is flagged before MDT review.
  8. 08
    MDT review
    Tumour board reviews on the AURORA-GLIO surface. Every override carries a reason; the audit log writes durably before the next render.
  9. 09
    Sign + emit
    Signed report + structured artefacts flow back to PACS as SR, to FHIR as DiagnosticReport, to the RT planning system as DICOM-RT.
04Integrations the module speaks9 endpoints
PACS
DICOM C-STORE inbound + C-FIND outbound · STOW-RS for derived series · SR attachment on signed reports.
EHR / FHIR
FHIR R4 — DiagnosticReport, Condition, Observation, Procedure, ServiceRequest. OAuth2 client credentials. SMART scopes accepted.
Molecular feeds
HL7v2 ORU^R01 for legacy NGS pipelines; FHIR Observation for newer LIS. AURORA-GLIO ingests both shapes without site-side mapping.
Histology
Direct connector to Aperio AT2, Hamamatsu NanoZoomer, 3DHISTECH; DICOM-WSI accepted natively where the scanner supports it.
RT planning
DICOM-RT struct + draft RTPLAN to Eclipse, RayStation, Pinnacle. Round-tripped without lossy resampling.
Auth
OIDC — Microsoft AD FS, Okta, Keycloak. Per-clinician identity in every override log entry; role mapping per RFC.
Provenance
Sigstore-signed weights and container digests. Per-site federation key in Vault, KMS or HSM. Audit log is signed before the next render.
Telemetry
Off by default. When enabled, opt-in per-site; payload schema is in the GLIO-ETHIC RFC, reviewed by the patient-advocate council.
Air-gap
OCI tarball + offline Helm chart. Identical outputs to networked deployments on identical inputs; bytewise reproducible.
05Per-subsystem, in detail11 units · independently installable
GLIO-SYS
Systems Atlas
live

Multi-omic systems atlas. Combines TCGA/CGGA cellular-state maps with single-cell drift trajectories from the Patel-Neftel lineage. Other GLIO subsystems consume it as a prior; it is not user-facing, but it is the reason their outputs are coherent across cases.

Producesprior tensors (.aurora.prior)
GLIO-MMB
Macro-Micro Bridge
live

Macro–micro bridge. Takes the MRI signal at scanner scale and aligns it against histology features and spatial transcriptomics at the cellular scale. Surfaces a discrepancy when the macroscopic radiomic signal and the underlying micro-anatomy disagree — useful for catching atypical presentations.

Producesalignment map (NIfTI) + discrepancy report (JSON)
GLIO-TME
Tumor Microenvironment
live

Tumour microenvironment simulator. Models immune-stromal-vascular crosstalk under treatment-induced shifts. Outputs a TME state estimate that conditions both GLIO-DIAG (subtype) and GLIO-CLIN (immunotherapy eligibility).

ProducesTME state vector + uncertainty (JSON)
GLIO-AI
Foundation Models
live

Foundation model. Vision-language-omics transformer distilled to 7–13B parameters with two-pass quantisation; a single RTX-class GPU handles a case under the latency target. Produces a dense case representation that downstream subsystems consume — it does not produce a label.

Producescase representation (.aurora.case) + attention maps (NIfTI)
GLIO-INFIL
Infiltration Field
live

Reaction-diffusion PDE solver. Differentiable inverse problem on patient anatomy. Outputs per-voxel infiltration probability beyond the contrast-enhancing volume, with calibrated σ that GLIO-RAD's contours respect.

Producesmargin field (DICOM SEG + NIfTI + σ sidecar)
GLIO-CAUS
Causal Trajectories
beta

Counterfactual graph engine. Treatment, resection extent, progression and outcome as nodes; the engine is explicit about which comparisons are confounded. Used by tumour boards to interrogate retrospective decisions, not to make new ones.

Producescounterfactual graph (JSON-LD) + retrospective annotations
GLIO-DIAG
Integrated Diagnosis
live

Integrated-diagnosis synthesiser. Fuses histology, molecular and radio into a single WHO-2021-aligned report with confidence per axis. Discrepancies surface as overrides; the pathologist is the resolver.

Producesintegrated diagnosis (signed PDF + JSON SR)
GLIO-RAD
Radiotherapy Plan
beta

Radiotherapy planner. Auto-contoured GTV/CTV/PTV with dose-response priors that consume GLIO-INFIL's margin field directly — contours respect σ rather than treating it as a footnote. Plans round-trip into Eclipse / RayStation / Pinnacle without lossy resampling.

ProducesRTSTRUCT + RTPLAN draft + dose constraint table
GLIO-CLIN
Clinical Trials
beta

Trial routing. Eligibility matcher and adaptive-arm allocator for active glioma studies. Refreshes nightly; per-arm fit scores explain the rank. Surgeons can pin a trial for re-review at a future tumour board.

Producestrial routing list (FHIR ServiceRequest + JSON)
GLIO-ETHIC
Ethics & Governance
live

Equity-as-loss audit. Stratified performance across ancestry, sex, age band, socioeconomic proxy and access proxy. Runs inside the training loop; release-gates a module if any stratum is below threshold. Threshold revisable only by RFC.

Producesper-case equity audit (JSON) + cohort report (NDJSON)
GLIO-ORG
Organoid Twin
soon

Optional organoid digital twin. For consenting patients with a tissue bank arrangement, synchronises bench-derived organoid responses with the in-silico case. Off by default; opt-in per site, per cohort, per patient.

Producesorganoid sync record (JSON) + bench-twin index
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-GLIO 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-GLIO.

Drafts by the module team and pilot collaborators. Each is a working paper that documents one slice of the substrate — methods, evaluation protocol, lessons. They are slot-marked here; full PDFs attach as you provide them.

WP-01
AURORA-GLIO Systems Atlas: a cross-scale prior over glioma cellular states and patient signatures
Module team · GLIO-SYS
Working paper · Drafting · GLIO-SYS
Systems atlas
WP-02
From MRI signature to spatial transcriptomics: an open macro-to-micro bridge for glioma
Module team · GLIO-MMB
Working paper · Drafting · GLIO-MMB
Macro–Micro
WP-03
Modelling immune-stromal-vascular crosstalk in glioma under treatment-induced microenvironment shifts
Module team · GLIO-TME
Working paper · Drafting · GLIO-TME
Microenvironment
WP-04
An open foundation model for glioma: vision, language and omics in one substrate
Module team · GLIO-AI
Preprint · In review · GLIO-AI
Foundation model
WP-05
Reaction-diffusion priors for the invasive margin: a differentiable PDE solver for glioma
Module team · GLIO-INFIL
Preprint · In review · GLIO-INFIL
Infiltration PDE
WP-06
Counterfactual graphs over treatment, resection extent and progression in glioma
Module team · GLIO-CAUS
Working paper · Drafting · GLIO-CAUS
Causal
WP-07
An auditable WHO-2021-aligned integrated diagnosis pipeline
Module team · GLIO-DIAG
Protocol RFC · Ratified · GLIO-DIAG
Integrated dx
WP-08
Auto-contoured GTV/CTV/PTV with dose-response priors: an open planning pipeline
Module team · GLIO-RAD
Working paper · Drafting · GLIO-RAD
RT planning
WP-09
Adaptive-arm trial routing on a federated substrate: design and ethics
Module team · GLIO-CLIN
Position paper · Drafting · GLIO-CLIN
Trials
WP-10
Equity-as-loss in glioma: stratifications, thresholds and release gates
Module team · GLIO-ETHIC
Position paper · Drafting · GLIO-ETHIC
Ethics & equity
WP-11
A patient-derived organoid digital twin for consenting glioma patients
Module team · GLIO-ORG
Concept note · Sketch · GLIO-ORG
Organoid twin
⚠ DRAFTS · DOI-STAMPED AT PUBLIC ALPHA · NO EXTERNAL CITATIONS ARE SHIPPED AS "EVIDENCE" ON THIS PAGE
VI · Endpoints we will track

The metrics that matter over a lifetime.

Surgical wins are not the goal — quality-adjusted years are. These are the endpoints AURORA-GLIO is built to measure across pilot deployments. The targets below are pilot goals, not retrospective results.

+18.4%
target
Gross total resection rate

Target lift on partner-hospital cohorts versus pre-AURORA baseline, measured against the Sanai 2011 threshold.

−40%
target
Time to integrated diagnosis

From sample arrival to WHO-2021-aligned report, tumor-board ready. Baseline is two-to-three weeks in most centres.

94.2%
target
Margin field accuracy

Validated against post-resection histology of invasive margins on multi-site held-out cohorts.

+12%
target
Trial enrolment rate

GLIO-CLIN matches eligible patients to active trials earlier in their pathway.

0.91AUC
target
Recurrence vs pseudoprogression

Distinguishing true recurrence from treatment-effect on longitudinal scans.

100%
target
Auditable override rate

Every clinician override of a model recommendation has a logged reason and explanation chain.

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

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. Pick your stack — same code, same model cards, same audit story.

step 1 of 4
# 1 · install
pip install aurora-glio              # just AURORA-GLIO
# or
pip install aurora-neuro[glio]       # full stack with GLIO included

# pilot sites use a private credential; public alpha (Q4 2026) opens the public registry.
step 2 of 4
# 2 · verify
aurora doctor glio
# → AURORA-GLIO  ✓ python ≥3.11   ✓ torch ≥2.4   ✓ cuda 12.4
# → models       ✓ glio-foundation-2026-Q1   ✓ glio-infil-pde-v3
# → datasets     ✓ glio-bench (open splits)
# → license      MIT @ public alpha
# → ready in target <150ms (median, RTX 4090)
step 3 of 4
# 3 · try it on a public case
from aurora.glio import load_case, diagnose

case = load_case("bench/glio/case-0007")
rep  = diagnose(case)

print(rep.subtype)        # → "GBM, IDH-wildtype"
print(rep.mgmt)           # → "methylated, conf 0.83"
print(rep.margin.sigma_mm) # → 1.4
step 4 of 4
# 4 · attach to your PACS (pilot sites)
from aurora.glio.pacs import attach

attach(
    pacs_endpoint="dicom.hospital.local:11112",
    aurora_node="aurora.hospital.local",
    consent_policy="cohort/neuro-onco-2026",
)
# → bound 412 patients, 9,318 studies
# → federation node: aurora-glio.hospital.local:8443
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-GLIO is designed to run on hardware that already exists inside hospital networks. The foundation model is distilled to a 7B–13B parameter band with two-pass quantisation; a single recent RTX-class GPU handles a case in well under the latency target. CPU-only inference is supported for the lighter subsystems (DIAG, CLIN, ETHIC) so a workstation deploy without a GPU is realistic for low-volume sites.

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 treated as a first-class object: subsystems with an explicit consent dependency (e.g. GLIO-ORG, GLIO-ETHIC equity stratification) refuse to run on records that lack the appropriate cohort policy.

Migration paths

Most pilot sites land on AURORA-GLIO with an existing imaging-AI vendor in place. AURORA does not replace that vendor in a single step. The recommended path is: (1) install AURORA alongside, (2) compare outputs on a held-out cohort for one quarter, (3) move read-only surfaces (margin field, integrated diagnosis) to the AURORA surface, (4) decide on the rest. Reversibility is a design goal — nothing in AURORA's install creates lock-in.

VIII · Questions

The honest questions.

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