Click here to

Session: Clinical Implementation of AI-based Auto-Segmentation [Return to Session]

Clinical Implementation of AI-Based Auto-Segmentation

J Yang1*, G Sharp2*, S Elguindi3*, (1) MD Anderson Cancer Center, Houston, TX, (2) Massachusetts General Hospital, Beverly, MA, (3) Memorial Sloan Kettering Cancer Center, Brooklyn, NY


WE-B-BRA-0 (Wednesday, 7/13/2022) 8:30 AM - 9:30 AM [Eastern Time (GMT-4)]

Ballroom A

Auto-segmentation has received renewed popularity in recent years thanks to the improved performance offered by artificial intelligence (AI). Many clinics are implementing auto-segmentation to take advantage of its potential improvements over manual contouring, including reducing inter-observer variability and staff workload. Currently, there is still relatively little guidance for how to best implement AI auto-segmentation into the clinical workflow. Whether acquiring commercial auto-segmentation software or implementing a model developed in-house, these tools need to be carefully evaluated and commissioned properly prior to releasing for clinical use.

This session will provide an overview of AI-based auto-segmentation, discuss the issues medical physicist should consider when clinically implementing these tools, and give recommendations for how to commission such systems. Furthermore, we will discuss how to evaluate the results of auto-segmentation and its expected performance. Finally, we will review methods for safely deploying in-house AI models, how to efficiently monitor their performance once clinically released, and give example strategies for improving performance with new data.

AI Auto-Segmentation: Clinical Implementation Concerns and Commissioning Guidelines – Speaker: Jinzhong Yang
Evaluation Methods for Auto-Segmentation and Expected Results – Speaker: Gregory Sharp
End-to-End Pipeline for Clinical AI Auto-Segmentation – Speaker: Sharif Elguindi

Learning Objectives:
1. Understand how to safely implement AI-based auto-segmentation in the clinic
2. Understand how to assess AI-based auto-segmentation and its expected results
3. Understand the full process, from design to deployment to retraining for in-house AI-based auto-segmentation



Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email