Patient Graph
Longitudinal context across visits, documents, languages, and episodes.
Platform
The patient-context layer.
MedLineage starts with one painful workflow — preparing complex records before specialist review. Each reviewed packet compounds into reusable patient context the same tenant can read back: source-cited timeline, provenance trail, missing-record state, and exportable structure for downstream workflows.
From packet to patient-context layer
The packet is the wedge. The source-cited patient-context layer is the platform.
MedLineage starts with one painful workflow: preparing complex records before specialist review. Each reviewed packet can compound into reusable patient context — source-cited timeline, provenance trail, missing-record state, and exportable structure for downstream workflows.
Provenance Ledger
Every finding remains tied to source document, page, language, and extraction path.
Workflow Layer
Packets can support second opinions, referrals, tumor boards, trial screening, payer review, and clinician-approved agents.
—Clinical State EngineLive
What is known about this patient — and what is no longer known
Most systems return a snapshot. This one answers state(patient, t): the knowledge state at a point in time, recomputed deterministically from the record with no model in the loop.
Three clocks: when the document is dated, when the fact was measured, and when we learned it. Replaying a past date returns what was known then — not today's answer redrawn.
Knowledge age: each fact is current, aging, out of date, or permanent, against curated per-metric windows. This describes the record's age — it never says an exam should be repeated.
A record-health meter that shows its own formula: coverage, freshness, verified ratio and conflict burden — with 'not measurable' reported as exactly that, never as zero.
Explicit unknowns: a missing measurement is reported as a gap in the record, not silently treated as a negative result.
STATE(PATIENT, T)DETERMINISTIC · NO MODEL IN THE LOOP
Inside the packet
What the buyer actually gets.
Case snapshot
Active issues, header summary, and readiness verdict at the top — the specialist sees the open questions before scrolling.
Clinical timeline
Every document on one chronological spine, with source-language flags and typed event badges. Each entry traces back to its original PDF.
Source-cited findings
Key findings grouped by category, ranked deterministically. Every claim cites document, page, and source language.
Missing-record checklist
Specialty-aware, three-tier — critical, recommended, contextual. The factual gap the receiving specialist will ask about.
Readiness / completeness
A coverage score and verdict over the expected record categories for the case specialty — surfaced before the specialist starts reading.
Clinician sign-off
Reviewable artifact with checkboxes, write-in records-requested area, date and specialist signature line. Verdicts feed the Provenance Ledger.
Full packet sections are visible in the sample PDF.
How it works
Four steps from raw records to a reviewable packet.
- 01
Upload fragmented records
PDFs, reports, labs, discharge summaries, HL7/FHIR-style data, and CSV exports.
- 02
Structure the case
Documents are classified, key events are placed on a timeline, and findings are extracted with source references.
- 03
Surface gaps
MedLineage checks for missing or stale records before the specialist review.
- 04
Generate the packet
The clinician receives a reviewable packet with cited findings, timeline, completeness checks, and sign-off area.
04Why now
Clinical AI has reached the easy part. Trust is now the bottleneck.
LLMs can now read more of the messy record layer: PDFs, scanned reports, multilingual notes, lab tables, discharge summaries, HL7 messages, FHIR feeds, and CSV exports.
But complex care cannot run on black-box summaries.
Before a specialist, tumor board, trial-screening workflow, payer review, or clinical agent can safely act, patient context has to be structured, cited, and reviewable.
MedLineage sits at that transition: after raw records, before clinical decisions.
Records are more digital, but not more usable.
Hospitals have EHRs, PDFs, portals, HL7, FHIR, and data exports — but the patient story is still scattered across systems, languages, and formats.
LLMs make extraction possible. Provenance makes it usable.
The breakthrough is not "summarize this PDF." The breakthrough is turning messy records into claim-level, source-cited clinical context a human can verify.
Interoperability increases the need for context.
More exchangeable health data does not automatically create clinician-ready understanding. Raw records still need to become structured, cited, and reviewable.
Why this becomes huge
Start with one packet. Compound into every clinical-AI workflow.
The same source-cited patient context is the prerequisite for every workflow above the packet. We start narrow on complex second opinions, then compound into referrals, tumor-board prep, clinical-trial screening, payer review, medico-legal review and human-approved clinical AI workflows.
- 1Today
Complex-care second opinions and specialist referrals — multilingual packet, missing-record checklist, provenance ledger.
- 2Next
Tumor-board briefings, surveillance signals, clinical-trial eligibility screening — same patient graph, new lenses.
- 3Later
Payer review, medico-legal review, EHR write-back, agentic clinical workflows — all gated on the same source-cited substrate humans already signed.
06On the roadmap
One day, your history follows you — not the building.
We are building toward a patient-controlled, consent-gated view of the same verified graph: a portable, source-cited history a patient could carry to a new hospital or specialist. This is where MedLineage is heading; it is not available yet.
- Patient-held access (planned)
- Consent-gated sharing (planned)
- Source-cited, portable history (planned)
- Every claim cites its document, page and source language
- Tenant-isolated · best-effort de-identification · dry-run by default
- Multi-format ingest & FHIR-like exports — no EHR migration, no lock-in
Connect your Fascicolo Sanitario Elettronico
LiveLog in with SPID or CIE on the real regional portal and your referti become a longitudinal, source-cited case. Nothing to install.
How it works, and why it's safeNot a medical device. Not certified for FHIR, GDPR or EHDS conformance. No diagnostic capability or anonymization guarantee is claimed.
THE PORTABLE CLINICAL VAULTVision — not yet available
Talk to us about a pilot.
Early customer conversations only — no paid pilots claimed, no hospital deployments to point at yet. If your team is preparing complex multilingual cases for specialist review, we’d like to hear from you.