Deep learning-enhanced 3D imaging unveils semaglutide impact on cardiac fibrosis

Myocardial fibrosis (MF), a hallmark of structural cardiac remodelling, drives disease progression across most forms of heart failure and plays a central role in heart failure with preserved ejection fraction (HFpEF). Despite its clinical relevance, effective treatments remain scarce. In preclinical models, current methods for quantifying MF fail to capture its regional heterogeneity, limiting reliable assessment of novel anti-fibrotic compounds. This study aimed to develop a whole-heart imaging and deep learning (DL)-based quantification pipeline for MF, and to validate its utility by evaluating the efficacy of a glucagon-like peptide-1 receptor (GLP-1R) agonist in a mouse model of HFpEF.

By utilising fluorescent collagen-labelling dye, tissue clearing and three-dimensional light sheet fluorescence microscopy (3D LSFM), we developed a high-throughput imaging platform for MF. We established a DL framework to quantify interstitial, perivascular and replacement fibrosis, as well as hypertrophy, in 17 left ventricular (LV) segments. The antifibrotic efficacy of the GLP-1R agonist semaglutide was evaluated in the db/db UNx-ReninAAV mouse model, which exhibits diabetes, kidney failure, obesity and hypertension.

Whole-heart 3D LSFM, combined with DL, enabled micrometre-resolution mapping of MF in rodents. Using this approach, we observed that interstitial collagen content increases proportionally with cardiac hypertrophy. Chronic treatment with semaglutide reduced LV hypertrophy and perivascular fibrosis but did not affect the extent of replacement fibrosis.

The established 3D imaging and quantification approach provides a powerful tool for evaluating the therapeutic efficacy of antifibrotic compounds and studying the pathological mechanisms underlying cardiovascular diseases.