Performance of cardiac PET/CT with and without phase analysis for detection of scar in cardiac sarcoidosis: Comparison to cardiac magnetic resonance imaging

Abstract

The presence of myocardial scar in CS patients results in poor prognosis and worse outcomes. 18F-fluorodeoxyglucose (18F-FDG) PET/CT excels at visualizing inflammation but is suboptimal at detecting scar. We evaluated PET/CT sensitivity to detect scar and investigated the incremental diagnostic value of automated PET-derived data. 176 patients who underwent cardiac magnetic resonance (CMR) and N-13 ammonia/18F-FDG cardiac PET/CT for suspected CS within 3 months were enrolled. Scar was defined as late gadolinium enhancement (LGE) on CMR without concordant 18F-FDG uptake on 18F-FDG PET/CT. Accuracy of cardiac PET/CT at detecting scar (perfusion defect without concordant 18F-FDG uptake) was assessed before and after addition of automated PET-derived data. Sensitivity of PET/CT for scar detection was 45.3% (specificity 88.9%). Addition of PET-derived LV volumes and function in a logistic regression model improved sensitivity to 57.0% (specificity: 80.0%, AUC 0.72). Addition of phase analysis maximum segmental onset of myocardial contraction >\u200961 improved AUC to 0.75, correctly relabeling 16.3% of patients as scar (net reclassification index 8.2%). Sensitivity of gated PET MPI alone for scar detection in CS is suboptimal. Adding PET-derived volumes/function and phase analysis data results in improved detection and characterization of scar.