How we plan to build it, in plain language.
Lucentaa is an early concept. This page covers what we intend the AI to do, what it should work with, how we plan to handle patient data, and what we expect to run on.
Assistive AI, with a radiologist deciding.
This is where things stand today: an early concept, with no clinical testing done and no finished product.
Early concept stage
No model has been clinically validated, and nothing has been tested with healthcare providers yet. Lucentaa is not cleared for diagnostic or clinical use.
Assistive, not autonomous
The intent is for the AI to flag candidate findings and help order the worklist, not to diagnose anything. Radiologists would confirm, edit, or dismiss any suggested finding.
Planned validation
We plan to work directly with radiologists during pilot testing to validate the platform's outputs, and to test models on annotated imaging datasets before they go near a pilot.
DICOM compatibility.
The aim is to work with the formats and systems radiology departments already have.
DICOM support (planned)
The plan is for Lucentaa to read standard DICOM series directly, so a team would not need to change how images are acquired or stored.
PACS-compatible intake (planned)
We intend to support sending studies from an existing PACS, so Lucentaa would sit next to current imaging systems instead of replacing them.
Multi-modality parsing (planned)
X-ray, CT, and MRI studies would be converted to a consistent format at intake, so the analysis step could treat every modality the same way.
How we plan to handle patient data.
This is the part of the design that needs the most care.
Data protection
Data protection practices are being designed in line with applicable regulations, including Nigeria's NDPR, as the platform moves toward pilot testing.
Security standards
Infrastructure and data handling will be built to meet healthcare-grade security standards. That is a plan, and none of it has been built yet.
No real patient data today
This site uses made-up sample data only. Lucentaa does not currently collect or process any real patient data.
Infrastructure.
What we expect to build, described at a high level.
Hosting still to be decided
We haven't settled on hosting or architecture yet. Infrastructure and data handling will be built to meet healthcare-grade security standards, and we'll describe the specifics once they exist.
Traceability (planned)
We intend to record each study's path through intake, analysis, and review, so any result could be traced back.
Room to iterate
We expect to evaluate and update models often once pilots begin, and we'll plan the infrastructure with that in mind.