Encrypted login | home

Program Information

An Efficient Tool for Structure Label Harmonization

T Xie

T Xie1*, Y Ge2 , J Kirkpatrick1 , S Yoo1 , F Yin1 , C Mayo3 , Q Wu1 , (1) Duke University Medical Center, Durham, NC, (2) UNC Charlotte, Charlotte, NC, (3) University of Michigan, Ann Arbor, MI


SU-K-601-2 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: 601

Purpose: To develop an efficient solution to addressing the inconsistency of structure labels in existing treatment plans. This tool allows us to identify the structure labels and conform them to the proposed standards, which is an important capability for integrating large datasets of Radiation Therapy data to support efforts of rapid learning and precision medicine.

Methods: We propose a rule-based method for structure label harmonization. Regular expression, a special language that defines character patterns, is used to identify and encode similar patterns in the labels of a structure. In the preprocessing, each structure will be given one or multiple well-formed structure labels. By using an algorithm similar to human perception, the system finds similar patterns among the labels from the same structure. It then generates a regular expression that forms all the rules. The system will also perform a crosscheck in the generated regular expressions to prevent mismatch between different labels having similar patterns. Then, it creates a dictionary specifying for each structure, the standard structure name proposed by AAPM TG263 and its corresponding regular expressions.

Results: We used 402 prostate cases and 102 brain-SRS cases for training and validation. The regular expressions in the dictionary covered all OARs (Organ-at-risk) and non-planning structures. We randomly selected 50 prostate cases and 30 brain-SRS cases to train the system, then validated the system using the remaining cases. We ran 3 independent tests on prostate cases, and 2 independent tests on brain-SRS cases, 96.5% of the prostate cases has all the structure labels correctly identified and relabeled, while the brain cases have a rate of 96%.

Conclusion: This study presents a rule-based approach to harmonize structure labels in existing clinical treatment plans. Preliminary tests demonstrate the feasibility of this system to efficiently and automatically harmonize structure labels across large datasets.

Funding Support, Disclosures, and Conflict of Interest: Supported by NIH under grant #R01CA201212 and a master research grant by Varian Medical Systems.

Contact Email: