AI-enabled longitudinal echocardiographic tracking of domain-specific multi-parametric cardiac diastolic parameters after enzyme replacement therapy in rare lysosomal storage fabry cardiomyopathy
K T Sung, D M Niu, M E Liu, P L Lin, Y H Lai, H W Yang, T C Hung, C L Hung, P S ChenAbstract
Background
Cardiac remodeling in Fabry cardiomyopathy evolves across structural and functional domains that do not change synchronously during longitudinal follow-up. Left ventricular (LV) hypertrophy often progresses or regresses gradually, whereas diastolic indices and left atrial (LA) mechanics may vary independently of ventricular geometry. Such domain-specific heterogeneity complicates treatment response assessment when relying on a single echocardiographic parameter. We applied artificial intelligence (AI)–enabled automated echocardiographic quantification to characterize multidomain longitudinal remodeling after enzyme replacement therapy (ERT).
Methods
In this multicenter retrospective study, 169 transthoracic echocardiograms from approximately 100 patients with Fabry disease were analyzed using an FDA-cleared AI-based platform. Sixty-nine patients with paired baseline and follow-up examinations were included (35 untreated; 34 ERT-treated). Longitudinal change (Δ = follow-up − baseline) was assessed for LV structural indices, diastolic parameters, and LA reservoir strain.
Results
Follow-up duration was comparable between groups (5.35 ± 3.79 vs. 4.16 ± 2.50 years; p = 0.128).
Untreated patients demonstrated progressive myocardial hypertrophy, whereas ERT-treated patients showed regression of LV mass (Δ LV mass index +22.65 ± 6.74 vs. −13.53 ± 6.45 g/m²; p < 0.001). Diastolic relaxation diverged between groups, declining in untreated patients but improving with ERT (Δ lateral e’ −1.01 ± 0.42 vs. +0.53 ± 0.39 cm/s; p = 0.010).
The clearest separation was observed in LA reservoir strain, which decreased in untreated patients but increased substantially in the ERT group (Δ −1.63 ± 1.97% vs. +5.08 ± 1.69%; p = 0.012). In contrast, longitudinal changes in LV end-diastolic volume index, LV end-systolic volume index, LV ejection fraction, and lateral E/e’ were not significantly different between groups.
Conclusions
ERT was associated with regression of myocardial hypertrophy and improvement in diastolic and atrial mechanics despite stable ventricular volumes. Within an AI-enabled longitudinal framework, LA reservoir strain demonstrated clear discrimination between treated and untreated remodeling trajectories, highlighting its utility as a sensitive marker for tracking treatment-related cardiac remodeling in Fabry cardiomyopathy.Figure