Objective:

To map responsible-AI evidence in medical imaging disease detection, identify methodological gaps, and synthesize clinical implications across heterogeneous modalities, diseases, and AI architectures using a PRISMA-informed systematic review.

Approach:

• Systematic Review: This PRISMA-informed systematic review synthesized 24 studies published between 2020 and 2025 that used AI for disease detection in various imaging modalities, including X-ray, CT, MRI, mammography, ultrasound, dermoscopy, retinal fundus imaging, and optical coherence tomography.

Key Findings:

• Explainability methods dominated the evidence base, while fairness, privacy, and uncertainty were less represented. Several studies reported accuracy or sensitivity above 90%, but these should be interpreted cautiously due to reliance on internal validation and curated datasets.
• Responsible AI in medical imaging requires evaluation through multidimensional evidence, including external validation and privacy risk analysis.

Interpretation:

Responsible AI in medical imaging must consider generalizability, privacy, fairness, explainability, uncertainty, safety, and physician trust.

Limitations:

• Included studies varied significantly in disease area, imaging modality, dataset size, model architecture, validation type, and performance metrics. No quantitative meta-analysis was conducted due to the heterogeneity of included studies, which limited the ability to draw generalized conclusions.

Conclusion:

Responsible medical-imaging AI should be evaluated through multidimensional evidence, including external validation and privacy risk analysis, rather than solely on diagnostic accuracy.