Artificial Neural Network, Attention Mechanism and Fuzzy Logic-Based Approaches for Medical Diagnostic Support: A Systematic Review

Abstract

Accurate medical diagnosis is essential for informed decision making and the delivery of effective treatment. Traditionally, this process relies on clinical judgment, integrating data and medical expertise to inform decision making. In recent years, artificial neural networks (ANNs) have proven to be valuable tools for diagnostic support. Attention mechanisms have enhanced ANNs performance, while fuzzy logic has contributed to managing uncertainty inherent in clinical data. This systematic review analyzes how the integration of these three approaches enhances computational models for medical diagnostic support. Following PRISMA 2020 guidelines, a comprehensive search was conducted across five scientific databases (IEEE Xplore, ScienceDirect, Web of Science, SpringerLink, and ACM Digital Library) for studies published between 2020 and 2025 that implemented the combined use of ANNs, attention mechanisms, and fuzzy logic for medical diagnostic support. Inclusion and exclusion criteria were applied, along with a quality assessment. Data extraction and synthesis were conducted independently by two reviewers and verified by a third. Out of 269 initially identified articles, 32 met the inclusion criteria. The findings consistently indicate that the integration of ANNs, attention mechanisms, and fuzzy logic significantly improves the performance of diagnostic models. ANNs effectively capture complex data patterns, attention mechanisms prioritize the most relevant features, and fuzzy logic provides robust handling of ambiguity and imprecise information through continuous degrees of membership. This integration leads to more accurate and interpretable diagnostic models. Future research should focus on leveraging multimodal data, enhancing model generalization, reducing computational complexity, and exploring novel fuzzy logic techniques and training paradigms to improve adaptability in real-world clinical settings.

Ähnliche Arbeiten