Health AI project support
From initial question to prototype: scoping, architecture, experimentation and documentation.
Explore project support →Artificial intelligence and health
Design health AI systems. Verify the robustness of those that already exist.
QSTOM-IT supports organisations that build, fund or assess artificial intelligence systems for health. We work from initial project design through to independent review of data, methodology and reported performance.
Run the robustness diagnostic →QSTOM-IT
Build or assess: QSTOM-IT works on both sides of the lifecycle.
Useful health AI should be designed together with its evaluation protocol. We can build the project with you or act as an independent third party on an existing system.
Build
Are you developing a health AI project?
Problem scoping, feasibility, data, architecture, prototyping, experimentation and documentation. Evaluation is designed alongside the model.
Design a health AI projectAssessDo you need to review an existing model?
Technical audit, reproducibility, bias, data leakage, metrics, generalisation and failure points before funding, partnership or deployment.
Assess a health AI systemFramework in preparation
PRISM structures how we assess robustness.
Performance, Robustness, Integrity, Safety & Methodology: five dimensions used to organise questions, evidence and tests without confusing technical assessment with regulatory certification.
PPerformance
Metrics, thresholds, calibration.
RRobustness
Generalisation and failure points.
IIntegrity
Cohorts, leakage and reproducibility.
SSafety
Critical errors and oversight.
MMethodology
Protocol and scope of claims.
Free diagnostic - 2 minutes
Does your protocol show methodological weaknesses?
Five questions on data, cohorts and metrics identify several theoretical risks. No code or data is uploaded.
Additional services
Beyond assessment
QSTOM-IT also supports project delivery and team capability building.
Health AI training
Robustness, bias, metrics, data leakage, generative AI and critical review for technical or business teams.
Explore training →Resources
Publications and work in progress
We progressively publish our methods and analysis so that our approach can be examined and challenged.
EU AI Act and health
A practical reading of requirements related to data, robustness, documentation and human oversight, with particular attention to the interplay with MDR and IVDR.
View the dossier →Health AI assessment framework
The PRISM framework is being formalised and is intended for publication with a first application case.
Explore PRISM →Lille - France
A Lille-based company focused on health AI
QSTOM-IT is based in Lille and works with companies, investors, research teams and health organisations in France and remotely across Europe.
About QSTOM-ITQSTOM-IT
Quantitative & Scientific Testing of Operational Models - Intelligent Technologies
Measure, experiment, document and test artificial intelligence systems intended for real-world use.
QSTOM-IT
A project to build, a model to assess or a team to train?
Tell us the context, the decision at stake and what is already available. We will propose a proportionate scope of work.