Click here to


Are you sure ?

Yes, do it No, cancel

Successful Turing Test of a Physician-Specific Knowledge Based Planning Model for Prostate Radiotherapy

J Lamb1*, J Ginn2, C King1, N Agazaryan1, M Steinberg1, A Kishan1, S Tsai1, (1) University of California, Los Angeles, Los Angeles, CA, (2) Washington University, Saint Louis, MO


(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

In recent years substantial effort has been dedicated to the development of automated intensity modulation radiotherapy (IMRT) treatment planning using knowledge based planning (KBP). Several groups have reported that KBP automated planning produces plans that are clinically acceptable based on quantitative dose-volume metrics. However, very few reports describe blinded physician evaluation of KBP automated plans. Here we report such a study using a commercial implementation of KBP, helping to address a key hurdle to clinical acceptance of KBP automated planning.

A KBP model was created using a set of 100 prostate-only +/- seminal vesicle cases planned by a single dosimetrist and approved by a single radiation oncologist at our institution. Automated planning was tested on 20 retrospective cases not used for model building. The set of 20 automated plans and corresponding 20 manual plans were evaluated in a blinded fashion by a radiation oncologist based on qualitative evaluation of dose distributions and quantitative evaluation of DVH metrics.

For the plan grading under randomized presentation, the mean clinical plan score was 4.5 (min:4, max: 5, mode: 4.5, rms: 0.28). The mean automated plan score was 4.7 (min: 4.5, max: 5, mode: 4.5, rms: 0.26). In the blinded head-to-head comparison 5 automated plans were graded as better than the clinical plans in a clinically significant way, 5 automated plans were graded as better than the clinical plans but in a clinically non-significant way, and 10 plans were graded as no difference.

Knowledge-based automated planning of prostate-only radiotherapy using a physician-specific knowledge base was found to be practically indistinguishable from manual planning in a blinded physician plan review of 20 cases, suggesting that its routine clinical use is warranted.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email