D-PO01-144 - Side-by-side Comparison Of Two Methods For Characterizing Atrial Fibrosis Assessed By Late Gadolinium Enhancement Magnetic Resonance Imaging: What Is The Optimal Iir Threshold? (ID 938)
Background: There are two methods for characterizing atrial fibrosis in late gadolinium enhancement (LGE)-MRI. The Utah method involves threshold selection by expert operators from a histogram of wall intensity values; the image intensity ratio (IIR) method normalizes wall intensity to mean blood pool intensity, then classifies voxels exceeding an IIR threshold (IIRthr) as fibrotic. Inter-method differences have not been assessed. The lack of a standard IIRthr has prompted widespread confusion.
Objective: We compared Utah and IIR fibrosis patterns extracted from the same MRI scans to identify the IIRthr that leads to maximal agreement.
Methods: Pre-ablation LGE-MRI was acquired for 16 AF patients (Fig A). The Utah method was carried out by a third-party image processing service (Merisight, Marrek Inc). For IIR, we used ADAS 3D (Galgo Medical SL). Five distinct IIRthr values spanning the published range were used. LA volume was compared to confirm the methods concurred. IIR models were deformed by non-rigid registration to match Utah geometry, facilitating analysis of fibrosis overlap (Fig B) and Dice coefficient calculation.
Results: LA volume from the two methods agreed (Fig C). IIRthr values of 1.03 and 1.08 produced models with fibrosis burdens that were not statistically different from those in Utah models (Fig D). Published IIRthr values of 0.97 and 1.2 led to over- and underestimation of fibrosis burden, respectively. Spatial overlap between Utah/IIR fibrosis patterns was highest for IIRthr ≤1.03 (Fig E).
Conclusion: Maximal agreement between Utah and IIR methods is seen for IIRthr = 1.03, taking into account both global burden and spatial distribution of fibrosis.