Turning clinical proformas into a reviewable AML register.
A focused application for deterministic extraction, evidence review and governed cohort exploration of Cambodian AML proformas.
- Stage
- Prototype
- Documents
- Tumor Board and CanReg PDFs
- Core method
- Deterministic and source-linked
Useful clinical data is often present, but trapped inside documents and inconsistent workflows.
The Cambodia application is designed around the forms already used by the clinical team. It extracts a governed set of AML fields from digitally generated PDFs and keeps each value connected to its source evidence for review.
Approved records then form a shared register that can be searched and explored through typed, allow-listed cohort queries without giving free-form text direct access to the database.
Documents become data without becoming a black box.
- 01
Import the existing proformas
Users can upload supported PDFs individually or stage larger PDF and ZIP imports. Unsupported or ambiguous documents fail closed for review.
- 02
Review source-linked extraction
Each governed value is shown with its source evidence. Corrections preserve the original extracted value and are recorded in the audit trail.
- 03
Approve the shared record
Reviewed cases enter a role-controlled clinical workspace with retained source documents, activity logging and recoverable deletion.
- 04
Explore the cohort
A visual builder and compact typed language compile through the same allow-listed field catalogue to parameterised queries.
- Deterministic extraction from supported digital PDFs
- Evidence-linked review and correction history
- Bulk-import controls and explicit conflict handling
- Governed cohort queries with reusable templates
- Synthetic documents and records are used in development
- Unsupported scans fail closed rather than being guessed
- The optional language assistant never receives patient records and cannot execute queries
- Go-live requires local governance, security, storage and recovery controls