Rapid Fire Abstracts
Negar Firoozeh, MD
Postdoctoral Scholar
University of Washington
Negar Firoozeh, MD
Postdoctoral Scholar
University of Washington
Anna V. Naumova, PhD
Research Associate Professor
University of Washington
Peter Muzi, BSc
Research scientist
University of Washington
Avanti V. Gulhane, MD, FSCMR
Instructor (acting)
University of Washington - Seattle, WA
Francis Kim, MD
Professor
University of Washington
Karen Ordovas, MD, FSCMR
Professor of Radiology
University of Washington
SGLT2 inhibitors show significant cardiovascular benefits in type 2 diabetes (T2D) patients. Experimental models suggest these inhibitors enhance glucose and fatty acid oxidation, increase ketones, and provide alternative energy to failing myocardium (1). However, the exact mechanisms, such as a potential reduction in cardiac fibrosis, remain unclear (2). Previous studies found no significant changes in myocardial T1, ECV, strain, cardiac volumes, function, LV mass, or LGE in T2D patients (2). This project further analyzed CMR using radiomics, which assesses pixel signal intensity beyond visible features. We hypothesize that radiomics analysis of LGE-CMR could reveal imaging features distinguishing T2D patients treated with dapagliflozin from controls, potentially identifying new biomarkers for early cardiac dysfunction to improve treatment planning and outcomes.
Methods:
In this double-blind, randomized trial, adults with T2D were assigned to receive either placebo or 10 mg of dapagliflozin daily for 12 months. CMR exams were performed at baseline and one-year follow-up on a 3T Ingenia scanner (Phillips). Standard CMR sequences assessed myocardial tissue and function. LGE phase-sensitive inversion recovery images were acquired 10 min after Gadavist injection in the short axis plane. Myocardial segmentation was semi-automatically performed using MIM encore (v7.3.2) following the 17-segment AHA model. Radiomics analysis was conducted using Pyradiomics (v3.0.1). Preprocessing included resampling to 1x1x1 mm and discretization with a fixed bin width of 25 units. Statistical analysis employed Stata with a multivariate LASSO regression model, adaptive lasso method, and BIC for variable selection. Univariate analysis (two-sample t-test and paired t-test) was performed on selected variables.
Results:
The mean age of participants was 62 years, 17% female, and 88% overweight (BMI ≥ 25). 61% had hypertension, 60% hyperlipidemia, and 40% a family history of CAD. No significant baseline differences were seen between dapagliflozin and placebo groups. Radiomics features extracted included 14 shape, 18 first order, 24 GLCM, 14 GLDM, 16 GLRLM, 16 GLSZM, and 5 NGTDM features. Multivariate LASSO regression identified key features including first-order shape elongation, skewness, and second-order GLCM Cluster Prominence, GLCM Inverse Variance, and GLSZM Variance (Table 1). While no significant radiomics differences were observed at baseline or one year between groups, feature changes differed over time. First-order features changed in the dapagliflozin group, while second-order features altered in the placebo group, indicating a potential treatment effect and their utility as biomarkers.
Conclusion:
Detailed LGE-CMR analysis detects changes in radiomics features after treatment with SGLT2 inhibitors that are not seen in controls. Monitoring subtle radiomics changes with CMR could help detect early treatment effects and aid in patient selection for long-term therapy.
