Related Experiment Videos
A framework for validating AI in precision medicine: considerations from the European ITFoC consortium
Rosy Tsopra1,2,3,4, Xose Fernandez5, Claudio Luchinat6
1Centre de Recherche Des Cordeliers, Inserm, Université de Paris, Sorbonne Université, 75006, Paris, France. rosy.tsopra@nhs.net.
BMC Medical Informatics and Decision Making
|October 3, 2021
Summary
Artificial intelligence (AI) can improve healthcare, but needs validation. A new framework from the ITFoC consortium ensures AI
Area of Science:
- Oncology
- Artificial Intelligence
- Bioinformatics
Background:
- Artificial intelligence (AI) holds significant potential to revolutionize healthcare systems, particularly in clinical decision-making.
- Current implementation of AI in healthcare is hindered by a lack of robust validation procedures for machine learning models.
- Developing reliable assessment frameworks is crucial for the clinical validation of AI technologies in medical applications.
Purpose of the Study:
- To present an approach for assessing AI's ability to predict treatment response in triple-negative breast cancer (TNBC).
- To utilize real-world clinical data and molecular -omics data for AI model evaluation.
- To establish a foundation for a validation platform within the ITFoC Challenge.
Main Methods:
- The European ITFoC (Information Technology for the Future Of Cancer) consortium developed a framework for clinical AI validation in oncology.
- The framework comprises seven key steps: intended use, target population, evaluation timing, datasets, data safety, performance metrics, and explainability.
- This framework underpins a validation platform for the ITFoC Challenge, a community competition to assess AI algorithms using external real-world datasets.
Main Results:
- The ITFoC framework provides a structured approach to the clinical validation of AI for predicting treatment response in TNBC.
- The framework details essential components for evaluation, including data safety and AI explainability.
- The ITFoC Challenge will enable community-wide assessment and comparison of AI algorithms against external real-world datasets.
Conclusions:
- Robust, unbiased, and transparent assessment of AI predictive performance and safety is essential before clinical implementation.
- The ITFoC consortium's framework is expected to facilitate the safe integration of AI into clinical settings.
- This work supports the advancement of AI in precision oncology and personalized patient care.