A comprehensive review of explainable artificial intelligence in healthcare methods, evaluation, and clinical integration

Explainable artificial intelligence (XAI) is essential for healthcare trust, yet a substantial gap persists between XAI techniques and actual clinical adoption. This review addresses this gap by framing clinical integration through three complementary lenses. First, we introduce a three-dimensional XAI classification framework-property, dependency, and scope-that moves beyond descriptive cataloging and serves as a practical guide for matching XAI approaches to specific clinical tasks. Second, we propose an integrated evaluation system that balances technical robustness, including fidelity, with measures of clinical utility such as workflow alignment and clinician confidence. Third, we analyze the divergent and often competing needs of key stakeholder groups to produce a role-characteristic mapping that clarifies what constitutes meaningful explainability in different clinical contexts. By positioning clinical integration as the center, this review outlines a pathway for translating XAI from methodological innovation to a dependable component of clinical decision support.