Click here to


Are you sure ?

Yes, do it No, cancel

BEST IN PHYSICS (THERAPY): Manual Plan Refinement Offers Little to No Improvement Over Knowledge-Based Automated Plans Across Multiple Disease Sites

R Kaderka*, R Mundt , N Li , V Bry , M Cornell , K Moore , UC San Diego, La Jolla, CA


(Wednesday, 8/1/2018) 10:15 AM - 12:15 PM

Room: Davidson Ballroom A

Purpose: To assess any treatment plan quality gains after manual refinement of knowledge-based planning (KBP) generated plans. Plans were compared in multiple disease sites across 498 patients treated with a KBP-based workflow.

Methods: Disease sites investigated were left lung (n=46), right lung (n=64), prostate (n=244), prostatic fossa (n=55) and head-and-neck (n=99). Since the introduction of KBP, our clinical workflow initiates a patient’s treatment planning process with a deliverable KBP (plan A) and subsequent manual refinement (plan B) performed at human planner discretion. The effect of manual refinement was quantified by site-specific dosimetric parameters. Plans were normalized to our institutional standard PTV D95%=100%, excepting cases where human planners altered this. Dose values are given relative to prescription. A/B samples were compared using two-sided paired t-tests, significance threshold of p<0.01.

Results: No statistically-significant differences were found across 36 dosimetric parameters after manual refinement of left lung plans. In right lung, manual refinement changed parameters at the p<0.01 level in only 6/36 metrics. Most significant changes occurred in PTV D1% (averaged A=126.2%,B=121.4%,p<0.001), offset by KBP-superior rib D1cc (A=67.0%,B=70.0%,p<0.001). In prostate, only 5/20 metrics were altered. Largest systematic change observed was PTV D1% (A=105.0%,B=105.3%,p<0.001), offset by increases to penile bulb Dmean (A=24.3%,B=26.1%,p=0.004). Similarly in prostatic fossa, only 6/20 metrics were altered; largest changes in PTV D1% (A=106.6%,B=105.8%,p<0.001) and rectum Dmax (A=105.9%,B=105.2%,p<0.001), set against KBP-superiority in penile bulb Dmean (A=39.5%,B=42.0%,p=0.002).Head-and-neck exhibited the greatest proportion of statistically-changed parameters (18/39) with improvements in PTV D1% (A=108.1%,B=106.3%,p<0.001), cord Dmax (A=62.9%,B=59.8%,p<0.001), right submandibular gland Dmean (A=77.1%,B=75.7%,p<0.001) and oral cavity V50% (A=71.6%,B=69.7%,p<0.001), offset by right cochlea Dmax (A=16.4%,B=18.2%,p=0.002).

Conclusion: Manually refined clinical plans showed little to no differences to unmodified KBP plans across nearly 500 patients and 5 disease sites. Where differences were observed, manual refinement resulted in largely clinically insubstantial gains offset by deteriorations in other parameters.

Funding Support, Disclosures, and Conflict of Interest: Dr. Moore reports grants and personal fees from Varian Medical Systems, during the conduct of the study; In addition, Dr. Moore has a patent Developing Predictive Dose-Volume Relationships for a Radiotherapy Treatment licensed to Varian Medical Systems, and a patent Knowledge-based prediction of three-dimensional dose distributions pending.


Treatment Planning


Not Applicable / None Entered.

Contact Email