AI testing, benchmarking and validation
Ensuring AI solutions are safe,
accurate and unbiased.
- Robust
- Reliable
- Reproducible
- Standardised testing
We can leverage our unique medical imaging dataset to benchmark AI performance to the largest cohort of specialised human readers using challenging, standardised test sets.*
Currently, we have the framework and capacity to deliver this product for breast imaging. However, please note that our solution for lung imaging is still under development.
* UK patent pending, 2312219.5
Enhancing Clinical AI Implementation
External validation of AI is the required standard before implementation into clinical practice.
PERFORMS AI provides rapid and reliable external quality assurance using the same framework used for humans.
AI algorithms are uploaded to a secure virtual environment and are then evaluated on the same standardised and challenging test sets used for humans.
PERFORMS AI benchmarks algorithm performance against the largest cohort of specialist screening reporters; putting performance into clinical context.
The platform enables the rapid assessment of different AI models or versions, alerting stakeholders urgently to performance drift.
Cases used in the test sets are collected from multiple imaging vendors and centres ensuring generalisability of AI results.
PERFORMS AI
key advantages
Expert radiologists curate and annotate validation data.
Diverse multivendor, multi-country datasets reveal AI strengths and weaknesses.
Rigorous clinical evaluation with statistical analyses for certification.
International radiologist performance data allows effective benchmarking.
Confidential, controlled environment ensures secure evaluations.
