Explainable Artificial Intelligence for Clinical Text Classification: A Scoping Review of Methods and Applications

Le résumé fourni par la source

Machine learning-based artificial intelligence systems often operate as "black boxes," offering high performance but limited interpretability, which raises concerns in clinical settings where transparency is essential. This study presents a scoping review of explainable artificial intelligence (XAI) approaches used in clinical text classification of unstructured electronic health record data. A total of 22 studies were analyzed, covering binary, multi-class, and multi-label classification tasks, with transformer-based models showing the highest performance. Model-specific methods, particularly attention mechanisms, were predominantly used in deep learning models, while model-agnostic approaches such as SHAP and LIME provided flexible interpretation across different architectures. Overall, XAI methods enhance the transparency and trustworthiness of AI systems in healthcare, although further research is needed to ensure reliable and clinically meaningful explanations.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.