Манзюк, Едуард АндрійовичСкрипник, Тетяна КазимирівнаЛукманов, Титур КаюмовичКириченко, Олександр Миколайович2025-01-102025-01-102024Manziuk E., Skrypnyk T., Lukmanov T.,Kyrychenko O. Ensuring confidence in neural network decisions in medical diagnostics based on visual data // X International Conference “Ukrainian-Polish Scientific Dialogues. Actual problems of modern science 2024” Bydgoszcz – Khmelnytskyi, 2024. p. 521-525https://elar.khmnu.edu.ua/handle/123456789/17594The paper presents a novel method for medical image analysis that combines the high accuracy of deep learning models and the interpretability of logical models. The proposed approach involves training a convolutional neural network (CNN) for accurate image classification, applying a spatial attention mechanism to localize important features, and constructing an interpretable Decision Rule Network (DRN) based on these features. The DRN is a set of logical rules linking feature values to diagnoses, allowing for transparent decision-making. The method was evaluated on brain MRI scans, achieving high accuracy with the CNN (>95%) and interpretability with the DRN. The authors emphasize the importance of achieving consistency between the CNN and DRN decisions for specific clinical cases, ensuring trust and compliance with ethical and regulatory requirements in medical AI applications.ukmedical image analysisconvolutional neural networksexplainable AIDecision Rule NetworkinterpretabilityТези доповідей