Radiology reporting in the age of artificial intelligence: implications for patient safety

Radiological reporting is a central component of clinical decision-making and a patient safety-critical system.Radiology reports inform diagnosis, guide therapeutic decisions and shape interdisciplinary communication; deficiencies in how imaging findings are documented, structured or communicated therefore have direct downstream consequences for patient care. 1 2 Despite this central role, radiological reporting remains vulnerable to welldocumented safety risks.Variability in report structure, inconsistent terminology, omissions of relevant findings and ambiguous language can impair interpretation and increase the likelihood of diagnostic misunderstanding or delayed action.Unstructured or inconsistently structured reports have been shown to contribute to communication failures and complicate clinical decision-making in high-stakes scenarios.[1][2][3] These vulnerabilities long predate the introduction of artificial intelligence (AI).Reporting variability and incompleteness were recognised as persistent threats to patient safety and clinical communication, prompting early efforts towards standardised reporting.1 2 Framing radiological reporting as a patient safety system emphasises its role as the interface through which imaging expertise is translated into clinical action.When this interface is fragile or ambiguous, even accurate interpretations may fail to result in appropriate patient management.This fragility is particularly consequential in the context of AI, as algorithmic outputs will flow through the report and may amplify existing weaknesses rather than correct them.[1][2][3] Protected by copyright, including for uses related to text and data mining, AI training, and similar technologies..