Quantification of Left Atrium Fibrosis From LGE MRI in Atrial Fibrillation

Late gadolinium enhanced magnetic resonance imaging (LGE-MRI) is a recent technique used for fibrosis distribution assessment in the left atrium (LA) wall. Unfortunately, LGE-MRI analysis does not rely on a standardized image processing protocol. The aim of the study was to compare different methods to quantify fibrosis in the LA in the 3D domain. LGE-MRI from 60 AF patients were analyzed applying five different approaches for fibrosis segmentation: an histogram-based (H-6SD), the image intensity ratio (IIR), the blood pool normalization (BP), the Chan-Vese (CV) and the graph-cut (GC) method. For each method we quantified the percentage of fibrosis with respect to the entire LA wall mass. Results showed these five approaches can be divided in two groups; the group composed by the H-6SD, CV and GC is characterized similar results (mean coefficient of variation =0.3), while the results obtained applying BP and IIR strongly depend on the quality of the acquisition (mean coefficient of variation=0. 62). Utah stage classification resulted in a wide disagreement (22/60 patients, 37%) among BP and IIR. The two approaches which best matched the Utah classification were the CV and GC (49/60 patients, 82%). This study confirmed the evaluation of the quantification method appears critical and further research is needed to define a satisfactory standard for atrial fibrosis segmentation.