Artificial Intelligence Assistive Tool for Radiotherapy Plan Evaluation Based on Analysis of Integral Dose

ABSTRACT The aim of the present study is to propose an in-house developed artificial intelligence (AI) tool based on Python programming for the purpose of integral doses (IDs) calculation useful in plan evaluation in modern radiotherapy techniques. Retrospectively, curative radiotherapy plans of cancer head and neck planned with intensity-modulated radiation therapy techniques employing seven and nine photon beams of 6 MV, were included in this study. The derived dose-volume histogram data were analyzed for the calculation of ID for each of the contoured structures including high-risk planning target volume (HR-PTV) and surrounding normal structures using an in-house developed Python program. In this study, variation of ID between nine-beam and seven-beam plans was calculated. It was found that the ID for HR-PTV volume was almost equal in both nine and seven beam plans with the percentage variation range 0.4%–1.4%, however, significant variation up to 14.4% in the ID of organ at risk was found. Furthermore, we utilized the standard deviation (SD) as a metric to assess the variability of the ID within the PTV and the surrounding normal tissues. The HR-PTV exhibited a low SD of 0.71, suggesting consistent ID patterns. In contrast, the organs at risk (OAR) exhibited noteworthy variations in SD values, with some reaching as high as 16.75. The SD was relatively elevated in the OAR in comparison to the HR-PTV. These elevated SD values within the OAR indicate significant dose variability across different patients. It is found that ID increases as the number of beams increases. The Python program used in this study for the calculation of ID, as an AI assistive tool for plan evaluation, can be run on the TPS or on a side-by-side computer which may be helpful in finalizing radiotherapy plans.