MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

An Automated Contouring Workflow for Increased Standardization and Efficiency

D Hoffman1*, J Meyers2, R Manger1, D Hoopes1, I Dragojevic1, (1) UC San Diego, La Jolla, CA, (2) MIM Software Inc., Cleveland, OH

Presentations

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

Room: AAPM ePoster Library

Purpose: To develop, implement, and evaluate a new automated contouring workflow for external beam radiation therapy, that allows for improved contouring standardization and reduces dedicated contouring time.

Methods: At the initial 3D-simulation, a DICOM tag is assigned to all scans, describing the scan as one of seven pre-determined anatomical treatment sites. MIM Assistant (MIM Software, Inc.) then uses this tag to generate a list of appropriate contours for planning, with standard order, colors, and names. Contour names are informed by AAPM TG-263 and previous naming practices. These contour templates are used by a central server, where a subset of the contours is then automatically segmented and sent back to the local server for further planning and approval. Finally, during the plan check, the final approved contours are returned to the server to allow for Dice scoring and future contour auto-segmentation improvement and machine learning.

Results: The new workflow managed 511 simulation scans, constituting every simulation since implementation. The workflow generated 10,624 contours, of which 3,137 (~30%) were automatically segmented. Comparing automatically segmented contours with the final approved contours shows 36% have Dice similarity > 0.9, and 23% scored > 0.99, indicating the portion of generated contours that were useful. The dosimetrists further planning with these contour sets self-reported an average time savings of 24 minutes per plan due to the workflow.

Conclusion: The principle benefit of this automated workflow is the standardization. The immediate clinical benefit is improved plan review and peer review from consistent naming and coloring, and the longer-term benefit will be easier retrospective study of plans with uniform structure naming. This workflow allows the clinically approved contour sets to inform future auto-segmentation through machine learning. An additional value is the subjective time saved, which based on the self-reporting and number of simulations, represents ~200 hours.

Keywords

Not Applicable / None Entered.

Taxonomy

IM/TH- image Segmentation: CT

Contact Email