Click here to


Are you sure ?

Yes, do it No, cancel

AI-Guided Automated Target Delineation and Planning Optimization by Cloud-Based Treatment Planning System

H Zhang1*, R Yang2 , (1) LinkingMed Inc. Beijing, Beijing(2) Peking University Third Hospital, Beijing, Beijing


(Sunday, 7/29/2018) 4:30 PM - 5:00 PM

Room: Exhibit Hall | Forum 6

Purpose: Accurate target delineation is the basic of precise radiation therapy. Cloud-Based Treatment Planning System (CTPS) developed a solution for intensity-modulated radiotherapy, containing AI-guided automated target delineation and treatment planning optimization. Subsequently some comparisons with manual target delineation are made to evaluate the performances of the solution.

Methods: The CTPS is running on a public cloud which contains image viewing, contouring (automated mode and traditional manual mode), and treatment planning. In this study, ten nasopharyngeal carcinoma (NPC) cases are used for proof-of-concept. AI-guided and manual contours are separately delineated for the treatment planning optimization. FMO method is used for optimization, and their DVHs are calculated. Finally, a multidimensional model is used to evaluate the DVH performance and contour similarity. For the target delineations, six indexes (Dice Similarity Coefficient, Volume variation coefficient, Center of Mass Deviation, Sensitivity index, Inclusiveness index, Jaccard coefficient) are used for consistency assessment.

Results: The calculation results show that it has agreed well between the AI-guided delineations and manual delineations by physicians. On the other hand, for the DVH performance, the optimized DVH curves shows that structures of the two delineation methods reached the dose objectives and achieved similar dose levels.

Conclusion: The preliminary results indicate that the AI-guided target delineations have showed similarity with those of manual target delineations by physicist. And the DVHs from planning optimizations based on the two kinds’ delineations also have good consistency. On the basis, it concludes that the solution, combining AI-guided automated target delineation with planning optimization, achieve the ability that has equivalent level with experienced physicist.


Segmentation, Optimization


IM/TH- image segmentation: General (Most aspects)

Contact Email