MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Human-Level Comparable Control Volumes Mapping with An Unsupervised-Learning Model for CT-Guided Radiotherapy

X Liang*, M Bassenne, D Hristov, T Islam, W Zhao, M jia, Z Zhang, C Huang, L Xing, Stanford University Cancer Center, Palo Alto, CA

Presentations

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

Room: AAPM ePoster Library

Purpose: develop an unsupervised deep learning model with auto-mapped control volume (CV) from planning computed tomography (pCT) to daily patient positioning CT for highly accurate and efficient patient positioning.


Methods: unsupervised learning framework is proposed to automatically generate the couch shifts (translations and rotations) for mapping CV from pCT to dCT (dCT). Inputs to the network are the dCT, the pCT, and the CVs’ locations within the pCT. The outputs are the transformational parameters of the dCT for head-and-neck (HN) patient positioning. We train the network to maximize image similarity between the CV in the pCT and dCT using normalized cross-correlation. Network training was performed with 470 dCT scans from 146 patients. These scans were acquired during different treatment fractions. The trained network was tested with a large number of daily scans. For each test case, couch shifts are obtained by averaging translational and rotational parameters derived with different CVs. These means are then compared to ground-truth reference shifts obtained by the alignment of bony landmarks identified by an experienced radiation oncologist.


Results: positioning errors between the model prediction and the reference are smaller than 0.49/1.17 mm and 0.13/0.28° in translations and rotations, respectively. Pearson’s correlation coefficient between model predictions and reference values exceeded 0.97. The proportion of the head and neck patient cases within a clinically accepted tolerance is improved from 63.89% to 86.11% compared the proposed method with the conventional method. The runtime for model prediction is less than 0.1 seconds per fraction.


Conclusion: novel unsupervised learning technique was established to map CVs from pCT to dCT for HN patient positioning. Our results show that fast and highly accurate HN patient positioning is achievable by leveraging state-of-the-art deep learning strategies.

Funding Support, Disclosures, and Conflict of Interest: This work was partially supported by NIH (1R01 CA176553 and R01CA227713), a Faculty Research Award from Google Inc.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email