Clinical efficacy of AI in lung SABR planning: A comparative retrospective analysis.

Abstract

This study evaluated the effectiveness of an integrated Artificial Intelligence (AI) planning tool in a lung stereotactic ablative body radiotherapy (SABR) planning workflow. The aim was to determine whether the AI planning tool would facilitate the generation of consistent high-quality plans while simultaneously improving treatment plan efficiency. The study compares clinically treated planner derived lung SABR plans with AI-generated. Nineteen cases planned with traditional planner derived techniques which make up the control cohort human, were re-planned using AI to determine the efficiency and quality of AI generated plans. The study derived a set of AI criteria to create the AI cohort of plans, and further refinement with an additional optimization created AI + human cohort. Each plan was assessed using departmental criteria, including time efficiency, to determine plan quality. The best plans, chosen after a blind review by the treating RO, were documented and analyzed to demonstrate the effectiveness of AI assistance in Lung SABR planning. Ethics approval was given for this study at a local health district level. Across 19 patients, the human cohort showed a total of 3.3% criteria unmet, which dropped to 2.6% for AI assisted plans in the AI cohort. The percentage of unmet goals was further reduced to 1.84% after the addition of manual planner input in AI + human cohort. All plans selected by the RO in the blind review were produced using AI + human input, and the average time taken to produce AI assisted plans was 1.08 hours. The study demonstrates that AI, in conjunction with human expertise, significantly enhances the efficiency and quality of lung SABR plans for patients, with quality confirmed through blinded evaluation.