2024 Annual Meeting Abstracts – General Poster Discussion

Purpose:The application of artificial intelligence (AI) for an efficient IMRT quality assurance (QA) -which uses AI to identify IMRT plans for selective measurement -has demonstrated promising results in properly triaging plans for QA and potentially reducing the workload for measurement-based QA.Yet its clinical implementation requires a comprehensive risk assessment to ensure patient safety.This work evaluated our AI-assisted hybrid quality assurance (AIHQA) workflow using failure mode and effect analysis (FMEA).Methods: The process maps and associated risks of the measurement based IMRT QA (MBQA) and AIHQA workflows were analyzed and compared.We used our institution's IMRTQA-related event reports and related expertise as a guide to identify and score each workflow's failure mode.The TG-100 recommendation is followed in ranking the failure modes' severity, occurrence, and detectability on a 1-10 scale.More analysis was done on the high-risk failure modes to create AIHQA mitigation strategies.Results: The average standard deviation and maximum risk priority numbers (RPN) for MBQA and AIHQA were 54.1374.84and 315 and 381.4436.56 and 900, respectively.The AIHQA exhibited five unique failure modes, which were attributed to automation bias, AI-model prediction accuracy, and operator errors.Four of five failure modes exceeded the action threshold of RPN=125, necessitating the implementation of mitigation strategies.Before the clinical integration of AIHQA, comprehensive onboarding and continuous training materials should be developed.These materials should cover the appropriate utilization and interpretation of AI prediction results, the risks associated with excessive reliance on predictions, and the limitations of the AI model.Conclusion: FMEA was performed on AIHQA to assess the potential risks associated with this new AI-assisted workflow and develop strategies to mitigate them.This proactive analysis provides a foundation for the safe introduction of the AIHQA, followed by continuous evaluation and development of mitigation solutions after its clinical implementation.