Explainable Artificial Intelligence: Transitioning DL Model Decisions to User-Understandable Features in Healthcare
Вантажиться...
Дата
2024-10-05
Назва журналу
Номер ISSN
Назва тому
Видавець
CEUR-WS.org
Анотація
Modern artificial intelligence (AI) solutions often face problems of the “black box” nature of deep learning (DL) models, which limits their transparency and trustworthiness in critical medical applications. In this study, we propose and evaluate a scalable approach for enhancing the interpretability of DL models in medical signal and image processing by translating complex model decisions into features that are understandable to healthcare professionals. The proposed approach was tested on two medical datasets: ECG signals for arrhythmia detection and MRI scans for heart disease classification. The performance of the DL models was compared with expert annotations, using Cohen’s Kappa coefficient as the primary metric to assess agreement. The study found strong agreement between DL model predictions and expert annotations, with Cohen’s Kappa coefficients of 0.89 for the ECG dataset and 0.80 for the MRI dataset, demonstrating the usefulness of the approach in providing reliable and interpretable results. In sum, the proposed scalable approach significantly improves the interpretability of DL models in medical applications.
Опис
Ключові слова
explainable artificial intelligence, deep learning, medical signal processing, medical image analysis, model interpretability, Cohen’s Kappa
Бібліографічний опис
Explainable artificial intelligence: Transitioning DL model decisions to user-understandable features in healthcare / P. Radiuk et al. 4th International Workshop of IT-professionals on Artificial Intelligence 2024 (ProfIT AI 2024) : CEUR-Workshop Proceedings, Cambridge, MA, USA, 25–27 September 2024 / ed. by D. Chumachenko et al. Vol. 3777. CEUR-WS.org, Aachen, 2024. P. 185–199. URL: https://ceur-ws.org/Vol-3777/paper12.pdf