Association of increased epicardial adipose tissue derived from cardiac magnetic resonance imaging with myocardial fibrosis in Duchenne muscular dystrophy: a clinical prediction model development and validation study in 283 participants

Figures

• 

  Figure 1. The flowchart of enrolled patients in the derivation and validation sets of the prediction model. DMD, Duchenne muscular dystrophy; CMR, cardiac magnetic resonance; LGE, late gadolinium enhancement.

• 

  Figure 2. The CMR images in two 13-year-old boys with DMD. (A-D) An LGE-negative patient and (E-H) an LGE-positive patient. (A,E) Images from the four-chamber long-axis view in end diastole showing the EAT in the left AVG (red arrows), right AVG (yellow arrows), anterior IVG (blue arrows), paracardial adipose tissue (gray arrows), and subcutaneous adipose tissue (black arrows). (B,F) The images of the short-axis view in end diastole showing EAT in the superior IVG (orange arrows), inferior IVG (green arrows), and RVFW (purple arrows). (C,G) Reconstructions of the EAT volumes (brown areas). The red reticulate sections represent the epicardium. (D) An LGE image suggesting no myocardial fibrosis in the left ventricular wall. (H) An LGE image suggesting myocardial fibrosis in the left ventricular free wall (white arrows). EAT, epicardial adipose tissue; LGE, late gadolinium enhancement; DMD, Duchenne muscular dystrophy; AVG, atrioventricular groove; IVG, interventricular groove; RVFW, right ventricular free wall; MR, cardiac magnetic resonance.

• 

  Figure 3. Nomogram to predict the occurrence of LGE-based myocardial fibrosis risk in DMD. (A) Tuning parameter lambda selection in the LASSO model used cross-validation for determining the minimal mean-squared error (the left dotted line) and the one standard error of the minimal mean-squared error (the right dotted line). (B) LASSO coefficient profiles of 29 variables were produced versus the log (lambda) sequence and 3 nonzero coefficient variables were selected. (C) Nomogram to predict myocardial fibrosis risk in DMD, with data taken from an 8-year-old child with DMD as an example. The steps for using the interactive dynamic nomogram are as follows: find the position of each risk factor on the corresponding axis (the higher blue wave crest indicates the more concentrated data distribution), draw a red dotted line to the points axis for the value of individual score, add the scores from the three risk factors, and finally draw a red solid line from the total point axis to determine the myocardial fibrosis probabilities of DMD. LGE, late gadolinium enhancement; DMD, Duchenne muscular dystrophy; BFP, body fat percentage; EATVI, epicardial adipose tissue volume indexed; LASSO, least…

• 

  Figure 4. Development and validation of the nomogram. (A) Calibration plots evaluating the actual probability versus predicted probability of the nomogram in the training set, internal validation set, and external validation set. (B) The AUCs evaluating discrimination of the nomogram, age, BFP, and EAT volume in all sets. (C) Decision curve analysis reflecting the good clinical net benefit of the nomogram in all sets. DMD, Duchenne muscular dystrophy; BFP, body fat percentage; AUC, area under the curve; EATVI, epicardial adipose tissue volume indexed.