Program Information

Tracking Tumor's Boundary in MV Image Sequences for Image-Guided Radiation Therapy

X Zhang

X Zhang¹*, N Homma¹ , Y Narita² , Y Takai² , K Ichiji¹ , M Abe¹ , N Sugita¹ , M Yoshizawa¹ , (1) Tohoku University, Sendai, JAPAN. (2) Hirosaki University Graduate School of Medicine, Hirosaki, Aomori, JAPAN.

Presentations

TH-E-17A-11 Thursday 1:00PM - 2:50PM Room: 17A

Propose:
To develop a level set method (LSM)-based algorithm to track the tumor's boundary in MV image sequences for image-guided radiation therapy (IGRT).

Method:
Four MV image sequences, each of which consists of 100 frames at frame rate of 7.5 Hz, are acquired by using the electronic portal imaging device (EPID) (Varian Medical Systems, Pal Alto, USA) during the treatment of lung cancer. In the first frame of each sequence, we roughly delineate an initial contour of the target tumor by hand. Using a LSM-based algorithm, the initial contour can automatically shape itself to fit the tumor, and eventually detect the tumor's boundary. We then employ the tumor's boundary obtained from the previous frame as the initial contour in the subsequent frame, so that the LSM-based method can drive this initial contour to the tumor's boundary quickly and fulfill a tracking task.

Results:
The proposed method has been evaluated on four MV image sequences. The mean tracking errors were 0.23, 0.29, 0.37, and 1.18 mm, and their corresponding standard deviations were 0.97, 0.75, 1.2 and 1.48 mm, respectively.

Conclusion:
Compared with conventional tumor tracking techniques, the proposed system is capable not only of tracking the tumor's position, but also of detecting the tumor's boundary varying with the respiration during the treatment. Considering current radiation therapy technique, for example, dynamical multi-leaf collimator (DMLC) has been widely applied in clinical treatment, this study indicates the potential for significant accuracy improvement in radiation therapy.


Funding Support, Disclosures, and Conflict of Interest: This work was partially supported by a research grant from Varian Medical Systems (Palo Alto, California).


Contact Email: