Click here to

Session: AI/ML Autoplanning, Autosegmentation, and Image Processing I [Return to Session]

Automated Rectal Cancer Radiotherapy Planning

K Huang1*, P Das2, A Olanrewaju3, C Cardenas4, D Fuentes5, L Zhang6, D Hancock7, H Simonds8, D Rhee9, S Beddar10, T Briere11, L Court12, (1) ,Houston, TX, (2) U.T. M.D. Anderson Cancer Ctr, Houston, TX, (3) Md Anderson Cancer Center, ,,(4) The University of Alabama at Birmingham, Birmingham, AL, (5) MD Anderson, Houston, TX, (6) MD Anderson Cancer Center, Houston, TX, (7) Md Anderson, ,,(8) Stellenbosch University, Stellenbosch, ,ZA, (9) MD Anderson Cancer Center, Houston, TX, (10) UT MD Anderson Cancer Center, Houston, TX, (11) MD Anderson Cancer Ctr., Houston, TX, (12) UT MD Anderson Cancer Center, Houston, TX


SU-E-BRB-3 (Sunday, 7/10/2022) 1:00 PM - 2:00 PM [Eastern Time (GMT-4)]

Ballroom B

Purpose: To develop an automated solution for rectal cancer 3DCRT treatments that combines deep-learning (DL) aperture predictions and forward planning algorithms.

Methods: We designed an algorithm to automate the clinical workflow for 3DCRT planning with field apertures creations and field-in-field planning. DL models(DeeplabV3+ architecture) were trained, validated, and tested on a total of 555 patients to automatically generate aperture shapes for primary (PA and Lat) and boost fields. Network inputs were DRRs, GTV, and diseased nodal GTV. A physician scored each aperture for 20 patients on a 5-point scale(>3 is acceptable). A planning algorithm was then developed to give a homogeneous dose using a combination of wedges and sub-fields. The algorithm iteratively identifies a hotspot volume, creates a subfield, calculates dose, and optimizes beam weight all without user intervention. The algorithm was tested on 20 patients using clinical apertures with different settings, and the resulting plans (4 plans/patient) were scored by a physician. The end-to-end solution was tested and scored by a physician on 39 patients using DL-generated apertures and planning algorithms.

Results: The predicted apertures had a Dice score of 0.95, 0.94, and 0.9 for PA, Lat, and boosts, respectively. 100%, 95%, and 87.5% of the PA, Lat, and boosts apertures were scored as clinically acceptable, respectively. A clinically acceptable plan was generated for all patients. Wedged and non-wedged plans were clinically acceptable for 85% and 50% of patients. The final plan hotspot dose percentage was reduced from 121±14% to 109±5% of prescription dose. The integrated end-to-end solution of automatically generated apertures and optimized field-in-field gave clinically acceptable plans for 38/39(97%) of patients. Feedback from different institutions indicated different clinical practices/criteria would require customization of the tool.

Conclusion: We have automated the clinical workflow for generating radiotherapy plans for rectal cancer for our institution.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email