초록접수 현황
| 20F-062 | 구연 발표 |
Validation study of Computer Aided Analysis of Risk Yield (CANARY) for prediction of ground-glass nodule invasiveness by chest computed tomography in Korean patients
Juyoung Lee¹, Kang Houn Lee¹, Jin Sung Kim¹, Seong Yong Park²
¹Department of Radiation Oncology, Yonsei Cancer Center, Yonsei University College of Medicine, Seoul, Republic of Korea
²Department of Thoracic and Cardiovascular Surgery, Yonsei University College of Medicine, Seoul, Republic of Korea
Purpose : It is widely known that classifying ground-glass nodules (GGN) into subspecies is clinically important in terms of the extent of lung resection before surgery. Several institutions have sought their own solutions, and we evaluated Computer-Aided Analysis of Risk Yield (CANARY), which is developed in Mayo Clinic, to find out the universality by applying Korean patients’ dataset in this study.
Methods : A total of 186 chest computed tomography (CT) sets are used and all patients underwent lung resection surgery between 2018 and 2019. 204 GGN cases from these patients were identified as 160 of invasive adenocarcinomas or 44 were ‘indolent’ (AAH, AIS, or MIA), based on the pathology report in our institution. With lesion area confirmation of thoracic surgeon, the Score Indicative of Lung cancer Aggression (SILA) from CANARY was calculated and evaluated to classify GGN as invasive or indolent.
Results : Our preliminary SILA to distinguish between invasive and indolent tumors were represented by the AUC of 0.729 with our dataset by using CANARY. It includes the f1 score of 0.351 and the highest 0.894 according to the threshold of 0.338 and 0.04 for each.
Conclusion : AUC was slightly lower than the research SILA has proposed. However, since CANARY has been validated in a few institutions in the US, there is room for improvement to apply it to Korean patients, such as adjusting the thresholds of the SILA and other parameters. Further study is necessary to figure out the reasons for the difference in current CANARY applications between the two organizations and countries.
Methods : A total of 186 chest computed tomography (CT) sets are used and all patients underwent lung resection surgery between 2018 and 2019. 204 GGN cases from these patients were identified as 160 of invasive adenocarcinomas or 44 were ‘indolent’ (AAH, AIS, or MIA), based on the pathology report in our institution. With lesion area confirmation of thoracic surgeon, the Score Indicative of Lung cancer Aggression (SILA) from CANARY was calculated and evaluated to classify GGN as invasive or indolent.
Results : Our preliminary SILA to distinguish between invasive and indolent tumors were represented by the AUC of 0.729 with our dataset by using CANARY. It includes the f1 score of 0.351 and the highest 0.894 according to the threshold of 0.338 and 0.04 for each.
Conclusion : AUC was slightly lower than the research SILA has proposed. However, since CANARY has been validated in a few institutions in the US, there is room for improvement to apply it to Korean patients, such as adjusting the thresholds of the SILA and other parameters. Further study is necessary to figure out the reasons for the difference in current CANARY applications between the two organizations and countries.
책임저자: Seong Yong Park
Department of Thoracic and Cardiovascular Surgery, Yonsei University College of Medicine, Seoul, Republic of Korea
발표자: Juyoung Lee, E-mail : juyounglee@yuhs.ac