Purpose: To achieve the nomenclature standardization based on the AAPM-TG263 and verify the method in multi-center for H&N tumor.
Methods: The H&N radiotherapy plans are randomly selected from three centers: SYSUCC (98 cases), PLA (85 cases), and GYFW (45 cases). Firstly, the CT images and the RT structure files are analyzed via an in-house developed software by Matlab. Secondly, through statistical structure naming and its color definition (RGB), the differences of each center are analyzed, and a multi-center standard naming library and color library are developed to complete a Semanteme-Based Standardizing Nomenclatures (SBSN). Thirdly, the geometric features of OAR, the texture features of the first-order gray histogram and the gray level co-occurrence matrix (GLCM) are used to construct a Content-Based Standardized Nomenclatures (CBSN) library and the discriminant library. Finally, 85% of the cases are used to build a multi-center CBSN-knowledge base, and the remaining 15% of the cases are used as a test set for the model validity evaluation respectively.
Results: We developed a standard naming library and color library for polycentric structures, unifying the normalized naming and coloring of a total of 93 structures across the three centers. The CBSN knowledge library of various centers was established, the test cases of each center were tested, and 20 errors/mismatches were found. For example, SYSUCC has the abnormal position and missing structure; PLA has empty structure and missing structure, and GZFW mainly has missing structure.
Conclusion: Based on AAPM-TG263, this study standardized the semantics, color, and the content of H&N radiotherapy structures and applied them in multi-center. Standardizing different structure names and colors is conducive to data sharing and communication between different institutions; at the same time, the knowledge bases established by different centers are compared and independently tested, and the model can detect the mismatches between semantic names and labels.