Feasibility of automatic IMRT planning for cervical cancer based on a database of previously-treated patients

Objective To develop and evaluate an automatic intensity-modulated radiation therapy (IMRT) program for cervical cancer based on a database of overlap volume histogram (OVH) and high-quality cervical IMRT plans for previously-treated patients. Methods A database consisting of high-quality IMRT plans and OVHs from 200 cervical cancer patients was established. OVHs of another 26 cervical cancer patients were converted into gray level images to calculate the image similarity compared with those from the database. The planning optimization function of the patients from the database with the highest image similarity was selected and inherent Pinnacle3 scripts were utilized to automatically generate IMRT plan. Finally, the dosimetric parameters, plan quality and design time were statistically compared between the automatic and manual plans. Results The target coverage, conformity index and homogeneity index did not significantly differ between two plans (all P>0.05). The V40, V45 and mean dose for the rectum in the automatic plans were significantly decreased by 6.1%, 1.3% and 50.7 cGy than those in the manual plans (all P 0.05). The plan design time was shortened by 71% in the automatic plans. Conclusions The automatic IMRT plans based on a database of OVH and high-quality IMRT plans can not only significantly shorten the plan design time, but also reduce the irradiated dose of normal tissues without compromising the target coverage and conformity index. Key words: Cervical neoplasm/intensity-modulated radiation therapy; Overlap volume histogram; Automatic plan