Artificial Intelligence (AI) in radiology promises enhanced diagnostic accuracy, efficiency, and consistency, yet introduces significant ethical and accountability challenges related to transparency and interpretability. Traditional transparency frameworks, process-based (focusing on technical details), outcome-based (centred on performance metrics), and value-based (highlighting embedded ethical considerations), alone do not sufficiently empower clinicians and patients to challenge AI-driven diagnoses effectively. To bridge this critical gap, this article proposes a novel contestability-centred framework operationalised through structured mechanisms, such as human-in-the-loop decision-making, intuitive override functionalities, continuous auditing, and systematic error reporting. Empirical evidence from real-world deployments demonstrates that integrating these contestability mechanisms significantly improves diagnostic reliability, clinician trust, and professional accountability. Furthermore, structured human oversight helps mitigate risks associated with AI errors, biases, or unforeseen system behaviours, directly addressing concerns highlighted by regulatory guidelines. By embedding actionable transparency and meaningful oversight directly into clinical workflows, the proposed approach fulfils emerging ethical and regulatory demands and ensures that AI remains a supportive tool that enhances, rather than replaces, human expertise, thus aligning technological advancement with ethically responsible medical practice.