Framework for Evaluation of Automated Knowledge-Based Planning Systems Using Multiple Publicly Available Prostate...

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Biology

Background:

• Knowledge-based planning (KBP) offers personalized treatment but requires significant effort to develop.
• Evaluating and comparing existing KBP routines is crucial for widespread adoption.
• Standardized frameworks are needed to assess KBP system performance.

Purpose of the Study:

• Establish an evaluation framework for KBP routines.
• Empower clinicians to select KBP systems aligned with clinical priorities.
• Demonstrate framework utility using four public prostate cancer KBP routines.

Main Methods:

• Applied four public prostate KBP routines (CCMB, Miami, UCSD, WUSTL) to 25 patients.
• Analyzed dose-volume histograms (DVHs) for planning target volume (PTV) and organs at risk.
• Compared dosimetric parameters using statistical tests across different PTV margin schemas.

Main Results:

• Calculated plans generally matched routine DVH estimations, except for CCMB organ-at-risk Dmax.
• Significant differences observed in PTV DMAX, PTV D99%, Rectum V40, and Femur DMAX among routines.
• UCSD and Miami routines yielded lower rectal doses; CCMB and WUSTL showed higher PTV homogeneity.

Conclusions:

• Publicly available KBP routines reduce development effort for clinicians.
• The proposed framework enables selection of prostate KBP routines based on clinical priorities.
• This methodology can be extended to compare KBP routines for other treatment sites.