Skip to main content
SCMR 29th Annual Scientific Sessions Main banner
Final Program


Program-At-A-Glance

Rapid Fire Session

Session: Rapid Fire Session 4: Machine Learning and Artificial Intelligence, Water-Fat Imaging (Monitor 10)

Higher fat fraction in the epicardial adipose tissue is associated with insulin resistance and obesity: a water-fat Dixon MRI study in type 2 diabetes

Friday, February 6, 2026
4:20 PM - 4:27 PM  
Location: Galapagos / Aruba / Ilhabela (2nd Lower Level)
Background: In addition to the epicardial adipose tissue (EAT) volume, the qualitative characteristics of EAT are increasingly recognized as risk factors for cardiovascular disease [1]. In this study we investigate the relationship between the EAT fat fraction and measures of metabolic disease severity as well as functional left ventricular characteristics in type 2 diabetes (T2DM).

Methods: We recruited 221 participants with T2DM (65% male, mean age 63.7 ± 7.9 years) from five primary care centers in Sweden. Participants underwent cardiac magnetic resonance imaging at 1.5 T. Images from a water-fat separated (Dixon) sequence were automatically segmented using a deep learning model [2] to obtain epicardial adipose tissue volume (EATVol) and mean epicardial fat fraction (EATFF). Within the epicardial adipose tissue, EATFF was calculated as: Fat signal / (Fat signal + Water signal). From a 2D cine sequence, feature tracking was used on short-axis images to assess radial and circumferential left ventricular systolic strain. HOMA-IR (calculated as fasting P-glucose x P-insulin / 22.5, in patients without insulin treatment) was used as a measure of insulin resistance.

Results: Mean BMI was 28.9 ± 4.3 kg/m2 and 38.9% of participants were obese. The mean time since diabetes diagnosis was 8.9 ± 7.0 years. Treatment was non-pharmacological in 21 subjects (10%), oral in 148 subjects (67%) and insulin in 52 subjects (23%). The mean EATVol was 53.1 ± 20.8 mL/m2 and the mean EATFF was 76.4 ± 2.5%. There was a strong positive correlation between EATVol and EATFF (R=0.75, p< 0.001).

EATFF increased across quartiles of both BMI (p< 0.001) and HOMA-IR (p< 0.001, non-insulin users), but did not differ based on type of diabetes treatment (p=0.52).

Mean global circumferential strain (GCS) was -20.9 ± 3.5 % and mean global radial strain (GRS) was +52.3 ± 12.6 %. EATVol showed no significant correlations to neither global circumferential nor radial strain. EATFF showed a weak correlation with GRS (R=-0.16, p=0.027), but not with GCS.

Conclusion: In this cohort of well-characterized patients with T2DM, we found that higher fat fraction in EAT was associated with measures of metabolic disease severity. Moreover, EATFF displayed an attenuated but significant association to LV strain, while EATVol did not.

Epicardial adipose tissue (EAT) automatic segmentation (shown in green) in two female participants with below and above median EAT volume: individual 1 (age 67, BMI 23 kg/m2, EAT volume 37 mL/m2, EAT fat fraction 76.6%) and individual 2 (age 72, BMI 35 kg/m2, EAT volume 88 mL/m2, EAT fat fraction 80.2%). (A) Three-dimensional (3D) MR images are shown from the left and right view of the heart. (B) Two-dimensional (2D) views of axial slices of the heart from cranial, mid, and caudal level.
Epicardial adipose tissue (EAT) fat fraction across quartiles of HOMA-IR (A) and BMI (B), p<0.001 for ANOVA for both. ** denotes Bonferroni corrected p<0.001 for pairwise comparison with Student’s T-test.