Unencrypted login | home

Program Information

Error of Template Matching for Tracking Tumor Motion Is No Larger Than Inter-Operator Variability


X Shi

X Shi*, T Diwanji, K Mooney, W D'Souza, N Mistry, University of Maryland School of Medicine, Baltimore, MD

TH-C-141-10 Thursday 10:30AM - 12:30PM Room: 141

Purpose: Recently, template matching has been proposed to track tumor motion using cine-MRI images. However, the technique has been tested primarily on blood vessels in healthy subjects. In this work, we demonstrate the utility of automated template matching in patients with thoracic tumors. We also compare the variability in predicting tumor position using template matching and 3 human operators.

Methods: Five patients with non-small cell lung cancer (NSCLC) were recruited in a prospective study. Cine-MRI imaging was performed while the patients were instructed to breathe normally. True FISP (fast imaging in steady state) cine-images were obtained in the sagittal plane. The centroids of the tumors were determined manually by 3 operators and also by the automatic template matching technique. The differences both in the AP and SI directions between the automatic and manual methods and the inter-operator variability were calculated respectively and compared.

Results: For all patients the variability between automatic and human operators in the SI direction was less than the inter-operator variability. For 4 out of 5 patients the variability between automatic and human operators in the AP direction was less than the inter-operator variability. However, the variability in 1 patient was higher by 14%. This is attributed to out-of-plane motion that currently is not captured using the cine-MRI imaging.

Conclusion: The effectiveness of template matching to estimate tumor position during respiratory motion was tested in patients with lung cancer. Results show that, the difference between the automatic and manual methods was comparable to the inter-operator variability. Future efforts to integrate on-board MRI imaging with linear accelerators can benefit from automated tumor tracking with template matching. Furthermore, the technique can also be utilized to replace the laborious manual tracking process necessary to establish the ground-truth when evaluating external surrogates of tumor motion.


Funding Support, Disclosures, and Conflict of Interest: NIH/NCI CA 122403

Contact Email: