Na Natio ional I l Instit itutes o of H Healt lth Clinical Ce Cl Center Image Translation by Latent Union of Subspaces for Cross-Domain Plaque Detection Yingying Zhu 1 , Daniel C. Delton 1 , Sungwon Lee 1 , Perry J. Pickhardt 2 , Ronald M. Summers 1 1 Imaging Biomarkers and Computer-Aided Diagnosis Laboratory, Radiology and Imaging Sciences, National Institutes of Health Clinical Center, Bethesda, MD 20892, USA 2 School of Medicine and Public Health, University of Wisconsin, Madison, WI 53706, US
Challenges • Generalization of calcified plaque detection model on different domains Contrast Enhanced CT Co CT (CE CECT CT) Non-Co No Contrast CT CT (NCCT CCT) Ha Hand La Labelled Pl Plaque ue Calcified plaque Calcified plaque Can we translate image across domains? Na Nati tional I Insti titu tute tes o of H f Health th Clinic Clin ical l Ce Center
Current Cross-Domain Image Translation Method Contrast-En Co Enhanced d CT Synt Sy nthet hetic No Non-Contrast CT (Li Liu et al. 2018) Calcified plaque Calcified plaque Calci Ca cified plaques are not ot preserve ved after im image transla latio ion Na Nati tional I Insti titu tute tes o of H f Health th Clinic Clin ical l Ce Center
Extract Small Patches to Union of Subspaces Calcified plaque Different image patch lies in different clusters Di Nati Na tional I Insti titu tute tes o of H f Health th Clin Clinic ical l Ce Center
Cross-Domain Image Translation by Shared Union of Subspaces Synt Sy nthet hetic Non0Contrast(CT((our(model) Contrast(Enhanced(CT Calci Ca cified plaques are preserve ved much ch better a be after i image ge t translation Na Nati tional I Insti titu tute tes o of H f Health th Clinic Clin ical l Ce Center
Results • Contrast Enhanced CT (CECT) calcified plaque detection & segmentation by Mask-RCNN He,%Kaiming et%al.%“Mask%R4CNN.” 2017%IEEE%International%Conference%on%Computer%Vision%(ICCV) (2017):%298042988. Training Data Tr No Non-Con Contrast CT CT Non-Con No Contrast CT CT No Non-Con Contrast CT CT (NCCT (N CCT) (NCCT (N CCT) (NCCT (N CCT) Testing data Te Synthetic NCCT Synthetic NCCT Synthetic NCCT by Cycle GANS by UNIT Liu et. al by our model Zhu et. al Precision 60.5±2.87% 63.2±2.64% 77.5 77.5±2.58 2.58% Recall 65.7±3.21% 69.5±3.05% 78.6 78.6±2.87 2.87% Dice 0.534±0.236 0.566±0.198 0.676 0.676±0.176 0.176 Na Nati tional I Insti titu tute tes o of H f Health th Clin Clinic ical l Ce Center
Na Natio ional I l Instit itutes o of H Healt lth Cl Clinical Ce Center Thanks Email:'yingying.zhu@nih.gov Nati Na tional I Insti titu tute tes o of H f Health th Clin Clinic ical l Ce Center
Recommend
More recommend