1349436 - Novel Threshold-free Fibrosis Signature Index (FSI) for Quantification of Left Atrial Fibrosis and Scar Burden from 3D LGE

Background: 3D left atrial (LA) LGE CMR has emerged as a promising technique for fibrosis and scar quantification in atrial fibrillation (AF) pre- and post-ablation. However, existing methods for LGE fibrosis/scar quantification rely on different thresholds for defining fibrosis burden. Inherent inconsistency in threshold selection hampers reproducibility and clinical translation. Recently, a novel threshold-free fibrosis signature index (FSI) method was introduced for standardized quantification of fibrosis burden ¹. FSI encodes multi-million intensity disparities over the entire LA voxels of 3D LGE to derive the unique fibrosis signature. This study aimed to assess the threshold-free FSI method in terms of: 1) feasibility for LA fibrosis and scar burden; 2) repeatability in post-ablation follow-ups; 3) comparison to two widely-used threshold-based methods.

Methods: This study analyzed 3D LA LGE of N=59 patients (age 67±10 yrs; 48% female) from publicly available database of CARMA Center ². We evaluated repeatability of scar analysis in a sub-cohort of the same database (N=23; age 66±10yrs; 57% female) with two 3D LGE follow-ups at 3-4 and 6-8 months post-ablation. Fig. 1 summarizes the three stages for computing the threshold-free FSI index of fibrosis burden from ~105±46 million pairwise intensity co-disparities including self-calibration for data uncertainty. We compared the threshold-free FSI method to two threshold-based fibrosis quantification methods: threshold >μ+3.3σ method (μ and σ as LA blood pools’ mean & stdev) ³, and the Image Intensity Ratio (IIR) method with threshold IIR >1.2 ⁴.

Results: Fig. 2a shows high correlation of FSI index of fibrosis burden and the two threshold-based methods’ fibrosis percentage in pre-ablation patients (Spearman ρ >0.74, p< 0.001) and scar percentages in post-ablation (ρ >0.92, p< 0.001). Fig. 2b shows examples of high and low FSI. The repeatability results of scar burden between the 1st and 2nd follow-up post-ablation scans (Fig. 3 a, b) showed that FSI method had higher agreement with 1.8 times lower variation along with higher correlation (Coefficient of Variation (CoV_FSI) = 17.7%, Spearman ρ_FSI=0.66) than the ρ_{μ+ 3.3 σ} method (CoV_{μ+ 3.3 σ} = 32.1%, ρ_{μ+ 3.3 σ} =0.39) and 1.3 times lower variation than the IIR method (CoV_{IIR 1.2} =23.2%, ρ_{IIR 1.2} =0.52). Fig. 3c shows an example result for follow-up scans.

Conclusion: This study demonstrates the feasibility of a novel threshold-free FSI method for standardized quantification of LA fibrosis and scar burden in pre- and post-ablation 3D LGE. FSI is a continuous index of fibrosis burden with higher values indicating higher burden. The threshold-free FSI method involves 1) automated self-calibration to scan-specific uncertainties; 2) inherent normalization to the LA size. Higher reproducibility of FSI in comparison with threshold-based methods was confirmed in the analysis of serial post-ablation scans (up to ~2 times lower variability). Thus, FSI may facilitate 3D LA LGE clinical translation.