Click here to


Are you sure ?

Yes, do it No, cancel

Evaluation of Two Different Systems for the Deformable Segmentation of a Lung Tumor in the Treatment of NSCLC Using SBRT

Y Cao1*, D Ghosh2, J Gomez3,A Singh4, H Malhotra5, (1) University at Buffalo, Buffalo, NY, (2) Williamsville North High School, Buffalo, NY, (3) Roswell Park Comprehensive Cancer Center, Buffalo, NY, (4)Roswell Park Comprehensive Cancer Center, Buffalo, NY, (5)Roswell Park Comprehensive Cancer Center, Buffalo, NY


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

Room: AAPM ePoster Library

Purpose: evaluation of MIM and Eclipse systems for the 4DCT automatic propagating contour of lung tumors in the Treatment of non-small cell lung cancer (NSCLC) using SBRT and compare the result with physician’s segmented tumors.

Methods: patients of NSCLC with various lobe location were tested. Two methods called brute deformation and sequential deformation were used to generate the internal target volume (ITV) by propagating the manual contouring of Gross Tumor Volume of 50% phase (GTV50), i.e. end exhalation to other phases. In brute method, the GTV50 was directly propagated to phase of 0, 10, 20, 30…90. In sequential method, the contour was sequentially deformed and then propagated to adjacent phases with each subsequent phase then becoming the primary phase for deformation. The ITV were defined by combining the GTV on all 10 phases of the respiratory cycle by using each of the above options available in the software. This resulted in 4 datasets to be compared against physician’s drawn ITV by using Jaccard Similarity Coefficient (JSC). A dependent-samples t-test was conducted to test the hypothesis that the two method and two system were associated with statistically significantly different JSC.

Results: brute deformation method provided slightly better JSC in both system and the MIM system provide slightly better JSC than the Eclipse system. However, Both the methods failed to achieve a perfect JSC of 1. The result of dependent-samples t-test showed the two method and two system was associated with a statistically significantly different JSC.

Conclusion: ITV generated using physician’s drawn 10 phases of a 4DCT dataset remains the gold standard for planning of NSCLC tumors using SBRT. Our results have supported this observation. However, automated 4DCT contour propagation tools still offer a useful tool in a busy clinic so long as the resultant volumes are vetted by the physician.


Deformation, Contour Extraction, Lung


IM/TH- Image Segmentation Techniques: General (Most aspects)

Contact Email