초록접수 현황
| 21F-113 | 구연 발표 |
Deep learning-based prediction of invasive adenocarcinoma among pure ground glass nodules of the lung
Duk Hwan Moon, Bong Jun Kim, Wongi Woo, Sungsoo Lee
Department of Thoracic and Cardiovascular Surgery, Gangnam Severance Hopsital, Yonsei University College of Medicine, Seoul, Republic of Korea
Purpose : Invasive adenocarcinoma (IA) presenting as a pure ground-glass nodule (GGN) is rare and remains to be controversial among thoracic surgeons. Herein, we aimed to investigate the radiologic and histopathologic characteristics of pure GGNs and to predict IA using deep learning algorithm.
Methods : Patients with pure GGNs who underwent surgical resection between April 2013 and April 2021 were included in this study. Based on the pathology reports, GGNs were divided into two groups: the non-IA group, including AAH, AIS and MIA, and the IA group. The radiologic and demographic data were reviewed using deep learning algorithm, and invented a prediction model for IA. The longest diameter, shortest diameter, mean diameter, and total volume of lesions were investigated to find the potential predictor
Results : A total of 165 patients – 195 GGNs – were analyzed in this study. The mean age was 54.1 ± 12.3 years, and the mean pathologic diameter of GGNs was 9.1 ± 4.3 mm. Pathologic classifications were AAH (n=15), AIS (n=69), MIA (n=73), IA (n=38). The median follow-up period was 4 years, with no mortality or recurrence during this period. Based on deep learning algorithm, longest diameter was the most suitable parameter in predicting IA (AUC: 0.708, the cut-off value:12.5mm).
Conclusion : The proportion of IA among pure GGNs was relatively high and its prognosis was excellent. Based on our findings, long diameter may be the most reliable parameter in predicting IA among GGNs. However, future studies are warranted to confirm this finding.
Methods : Patients with pure GGNs who underwent surgical resection between April 2013 and April 2021 were included in this study. Based on the pathology reports, GGNs were divided into two groups: the non-IA group, including AAH, AIS and MIA, and the IA group. The radiologic and demographic data were reviewed using deep learning algorithm, and invented a prediction model for IA. The longest diameter, shortest diameter, mean diameter, and total volume of lesions were investigated to find the potential predictor
Results : A total of 165 patients – 195 GGNs – were analyzed in this study. The mean age was 54.1 ± 12.3 years, and the mean pathologic diameter of GGNs was 9.1 ± 4.3 mm. Pathologic classifications were AAH (n=15), AIS (n=69), MIA (n=73), IA (n=38). The median follow-up period was 4 years, with no mortality or recurrence during this period. Based on deep learning algorithm, longest diameter was the most suitable parameter in predicting IA (AUC: 0.708, the cut-off value:12.5mm).
Conclusion : The proportion of IA among pure GGNs was relatively high and its prognosis was excellent. Based on our findings, long diameter may be the most reliable parameter in predicting IA among GGNs. However, future studies are warranted to confirm this finding.
첨부파일 : pure GGN figure.pptx
책임저자: Sungsoo Lee
Department of Thoracic and Cardiovascular Surgery, Gangnam Severance Hopsital, Yonsei University College of Medicine, Seoul, Republic of Korea
발표자: Duk Hwan Moon, E-mail : pupupuck@yuhs.ac