Improved unsupervised 3D lung lesion detection and localization by fusing global and local features: Validation in 3D low-dose computed tomography

Download PDF
Open Webpage

Ju Hwan Lee, Seong Je Oh, Kyungsu Kim, Chae Yeon Lim, Seung Hong Choi, Myung Jin Chung

Published: 01 Jan 2025, Last Modified: 07 Nov 2025Medical Image Anal. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•L-UADL boosts sensitivity by 17.5% via novel local feature use in chest LDCT.•Fusion-UADL fusing local/global features boosts sensitivity/specificity by 17.6/33.5%.•Fusion-UADL improves lesion localization by 17.7%+ on public lung cancer data.