Training and Evaluation of a Knowledge-based Planning System for Treatment Planning of Malignant Pleural Mesothelioma to the Intact Lungs

Abstract

Radiotherapy treatment planning of malignant pleural mesothelioma (MPM) with intact lungs is challenging not only due to the size and the complexity of the target volume but also because of the need to spare critical organs within and in close proximity. Our purpose was to develop and evaluate a knowledge-based model (KBM) for this purpose. Since June 2010, 62 patients have been treated at our institution with radiation for MPM after pleurectomy/decortication (P/D) and thus having both lungs intact. The treatment planning and delivery technique used was Volumetric Modulated Arc Therapy (VMAT) with 2 coplanar partial arcs and 6 MV photons. Prior clinically accepted treatment plans of these patients served as standard for plan quality. We chose 50 patients to build a dose volume histogram (DVH) estimation model based on the range of prescription doses, PTV volumes, volumes of organs at risk (OARs), essentially the ipsilateral lung, total lung, contralateral lung and the heart, as well as a range of their overlapping volumes with the PTV. Dose constraints for the total lung, contralateral lung and heart were similar to those for planning non-small cell lung cancer (NSCLC) cases while constraints for the liver, kidneys, stomach, esophagus and bowel were similar to those for planning abdominal tumors using conventional fractionation. Coverage constraints were PTV D95 and V95 = 94%. Relative priorities given for OAR sparing and coverage were same across all plans chosen for training. The DVH estimation model was trained and tested using knowledge-based planning software. Treatment plans for 12 patients generated clinically as well as using KBM were compared for dosimetric parameters using the Wilcoxon Rank-sum test. The prescription dose for these plans ranged from 30.6 Gy to 50.4 Gy (median dose 45 Gy). For 9 out of 12 patients (75%), the KBM plans were able to meet acceptable dosimetric criteria at least as well as the original clinical plans. No significant differences (p>0.05) were seen in the mean total lung dose (constraint of ≤ 20 Gy), V20 Gy to the total lung (constraint ≤ 40%), mean heart dose (constraint ≤ 30 Gy) as well as V30 Gy to the heart (constraint ≤ 50%) and the maximum achievable prescription dose (capped at 50.4 Gy). The remaining dosimetric parameters met the clinically acceptable guidelines while using KBM, however the original plans showed a trend towards improved sparing, although not statistically significant. The use of a knowledge-based planning system is capable of achieving clinically acceptable results for treatment planning of MPM with intact lungs having the potential of reducing treatment-planning time. Future work however is needed for testing more new cases and further refinement of the model to improve the current DVH prediction especially with respect to organs other than the lungs and heart.