Medimetrics is a web-based tool that helps researchers calculate various metrics commonly used to evaluate machine learning (ML) models. As ML popularity rises, the application of ML to high-risk fields, such as medicine, is becoming more and more common. Applying ML models to the real-world is very different from a typical experimental setting, where what works during testing may fail in a clinical setting. Therefore, it is essential to thoroughly evaluate all models using the appropriate metrics to fully understand how it may perform in different environments and scenarios. Medimetrics aims to make the calculation and understanding of these metrics more accessible to a non-technical audience.
This tool was launched in conjunction with a study on how ML models may be inaccurately evaluated when only using a subset of metrics to measure performance.
Please contact steven@simula.no, inga@simula.no, michael@simula.no, or paalh@simula.no for any questions regarding the tool or study.