Initial evaluation of automated treatment planning software.

Area of Science:

• Radiation Oncology
• Medical Physics
• Cancer Treatment

Background:

• Radiation treatment plan optimization is time-consuming and variable.
• Automated approaches are being developed to improve efficiency and consistency.
• A commercial autoplanning (AP) technique mimics human operators using automated regions of interest and dose-volume objectives.

Purpose of the Study:

• To evaluate the initial performance of a commercial autoplanning (AP) algorithm.
• To compare AP-generated plans with human-driven plans (HDP) for head and neck cancer.
• To assess dosimetric differences, including organ-at-risk (OAR) sparing and target dose homogeneity.

Main Methods:

• Ten challenging cases of locoregionally advanced head and neck cancer were analyzed.
• Patients were treated with VMAT (Volumetric Modulated Arc Therapy).
• An objective scoring system with dosimetric metrics and score functions was used for comparison. Five AP techniques were tested against HDPs.

Main Results:

• The best AP technique achieved an 8% lower composite score than HDPs, with statistically significant differences (p=0.027).
• HDPs demonstrated superior dose homogeneity, while APs provided significantly better OAR sparing (e.g., parotid glands, spinal cord, brainstem).
• Nine APs met RTOG 1016 protocol criteria with lower OAR doses than HDPs.

Conclusions:

• Automated planning is a promising clinical tool for head and neck cancer radiation therapy.
• AP offers improved OAR sparing but may require adjustments for optimal target dose homogeneity.
• Further refinement is needed to balance target coverage, homogeneity, and OAR sparing effectively.