Qualitative and dosimetric evaluation of AI-generated contours for conventional and under-reported organs at risk in prostate cancer radiotherapy.

PURPOSE: Accurate delineation of organs at risk (OARs) is essential in radiotherapy planning, influencing both dosimetry and potential short-and long-term side effects. The objective of this study was to evaluate the performance of MediQ RT, an autocontouring (autosegmentation) software, in prostate cancer radiotherapy, focusing on both traditional OARs as well as under-reported organs. MATERIAL AND METHODS: Treatment plans with volumetric modulated arc therapy (VMAT) for a cohort of 40 prostate cancer patients were created based on both manual and autocontouring methods. The two contouring methods were then quantitatively assessed using the Dice Similarity Coefficient (DSC), Hausdorff distance (HD95), Jaccard Index (JAC), Contouring Efficiency Index (CEI), Confidence Interval (CI), precision, and sensitivity metrics. Additionally, the dosimetric parameters and the time required for OARs delineation were analyzed to compare efficiency. RESULTS: The most accurate contouring results were observed for the bladder (DSC = 0.92, HD95 %=5.66 mm, CI:4.77-6.54 mm) and bowels (DSC = 0.91, HD95 %=7.30 mm, CI = 6.36-8.23 mm). The penile bulb had the least optimum values (DSC = 0.60, HD95 %= 12.09 mm, CI = 10.69-13.49 mm), attributable to its small structures that are challenging to identify automatically. The CEI demonstrated high efficiency for the femoral heads (≈10.00 (1/min)). Furthermore, the autocontouring, subsequent to minimal editing, resulted in a reduction of contouring time by approximately 90 %. The most relevant dosimetric impact was observed in under-reported OARs, which exceeded the dose constrains, therefore could be associated with clinical symptoms. CONCLUSION: Autosegmentation improves contouring efficiency and may support clinical workflow. However, disparities between manual and autocontouring can occur in certain organs, underscoring the necessity for expert review and correction prior to treatment planning.