|
KMID : 1114620200170020077
|
|
Journal of the Korean Society for Breast Screening
2020 Volume.17 No. 2 p.77 ~ p.84
|
|
|
Retrospective Analysis for Interval Cancer in Screening Mammography Applied for Artificial Intelligence Based Computer Aided Diagnosis
|
|
Kim Yeon-Soo
Chang Jung-Min
|
|
|
Abstract
|
|
|
Purpose: The purpose of this article is to retrospectively evaluate the diagnostic performance of artificial intelligence-computer aided diagnosis (AI-CAD) for interval cancers in screening mammography.
Materials and Methods: From January 2016 to December 2019, thirty-three breasts of 33 women (age range 40-76, mean age 54.2 years) who were newly diagnosed with breast cancer and underwent preoperative mammography, whose screening mammography was negative within one year, were included. Two dedicated breast radiologists reassessed the previous screening mammography in consensus, and thirty-three interval cancers were classified into missed and true interval cancers. The reassessment results were compared with those of AI-CAD.
Results: In reassessment, thirty-three interval cancers were classified into 12 missed cancers (36.4%) determined to be actionable for recall and 21 true interval cancers (63.6%) determined to be insufficient for recall. Of the 36 AI-CAD marks, thirteen (36.1%) were marked at the true positive cancer. Of the 33 interval cancer, AI-CAD had correctly depicted 9 interval cancers (27.3%, [9/33]) including 7 missed cancers (58.3%. [7/12]).
Conclusion: AI-CAD had correctly marked a substantial number of missed cancer determined to be actionable. Application of AI-CAD for screening mammography is expected to reduce missed cancers caused by the interpretation error of radiologists.
|
|
|
KEYWORD
|
|
|
Breast, Mammography, Missed cancer, Artificial intelligence
|
|
|
FullTexts / Linksout information
|
|
|
|
|
|
Listed journal information
|
|
|
|