Unicancer 2020 award, 6th edition
Aim of the project:
Our project is to develop an application that will enable the evaluation, in a multicenter, of radiomic or AI models proposed for the management of lung cancer patients. Multicenter validation of models remains indeed essential to consider their clinical use.
Methods:
This application will actually be a new module added to LIFEx, a software we are developing, to easily describe a radiomic model. As for any module of this application, the libraries used will guarantee that the application remains free and autonomous (i.e. without installation of third party software). Interfacing with the ConSoRe platform, which is a Unicancer's Big Data platform, will later make it possible to combine features extracted from medical images and text included in the medical reports.
Impact:
The main benefit will be to give medical teams a tool to evaluate radiomic and AI models on their own data, using models that have already been published or developed by colleagues. This large-scale, multicenter validation step provides information on the reliability and robustness of the models and is absolutely essential for clinical translation.
Problem to be solved:
If the field of radiomics and AI in radiology and nuclear medicine is growing exponentially, the current major bottleneck is the difficulty to validate the many published models in order to establish their performance and clinical added value. Our project will make this evaluation possible in an independent way, via a public software guaranteeing the transparency of the results.
Target users:
Our project will serve both the current users of LIFEx (10 Unicancer centers already use LIFEx, the 3600 users worldwide, including more than 700 in France) and future LIFEx users, given that we have about 100 new users every month. These are mainly radiologists, nuclear physicians, researchers and engineers involved in medical image analysis.
The Best of the AACR journals,
The most-cited research articles:
