MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All Sessions
  4. All SPS Undergraduate Research & Outreach Posters
  5. Validation of Production Standardizing Radiation Therapy Structures Names by the Content-Based Standardizing Nomenclatures (CBSN) in Radiation Oncology

Validation of Production Standardizing Radiation Therapy Structures Names by the Content-Based Standardizing Nomenclatures (CBSN) in Radiation Oncology

X MAI1,2*, S HUANG1,2 , S Huang1 , Y XIA1 , X HUANG1 , X Yang1 , (1) Sun Yat-sen University Cancer Center; State Key Laboratory of Oncology in South China; Collaborative Innovation Center for Cancer Medicine; Guangdong Key Laboratory of Nasopharyngeal Carcinoma Diagnosis and Therapy, Guangzhou, Guangdong, 510060, China. (2) Xinhua College of Sun Yat-sen University, Guangzhou, Guangdong, 510520,China.

Presentations

(Sunday, 7/14/2019) 3:00 PM - 3:30 PM

Room: Exhibit Hall | Forum 4

Purpose: A Semanteme-Based Standardizing Nomenclatures (SBSN) method in radiotherapy (RT) was developed in previous work. SBSN could automatically correct the typo or transfer the name as a standardization one. However, it can NOT detect human labeling errors with a correct semanteme name. The Content-Based Standardizing Nomenclatures (CBSN) method is proposed for implementation in RT upon SBSN, which tries to solve the problems such as data logging repetition and semantic illegibility in TPS, and enhances the capability of knowledge share and reuse.

Methods: We had developed SBSN in Matlab according to AAPM TG-263 and transplanted it into Eclipse Script v13.5. Furthermore, the SBSN had a multiple centers validation. 1013 H&N cases were randomly selected for CBSN. All structures were delineated by a radiation oncologist and verified independently by another one. After SBSN execution, 80% patients were randomly chosen as training set, and the remaining 20% cases were regarded as test set for CBSN performance evaluation. We explore five categories features, including line color, texture feature, appearance, profile and spatial relations. And there are 28 sub-kinds of different feature values. Then a knowledge-based library is generated based on those statistical information. The structures similarity in testing data set is calculated based on Manhattan Distance, Product-moment Correlation and Gower Similarity Coefficient in CBSN.

Results: There were 66±11 delineation structures in total (N=1012). And the CBSN could output one patient structure list within 2.2±1.5seconds with a sensitivity of 100% and find out 13 errors in all (N=202), such as naming the "Left Eye" as a "Right one. It can works well in those ones volume larger than 10cc and the Pituitary has the worst performance.

Conclusion: The CBSN method proposed in this work can help to avoid confusion from inconsistency and inadequacy of nomenclatures, errors; enable and improve the quality and safety in RT.

Keywords

Radiation Therapy, DICOM-RT, Software

Taxonomy

TH- Dataset analysis/biomathematics: Machine learning techniques

Contact Email

Additional Links