Cardiac computer tomography-derived radiomics in assessing myocardial characteristics at the connection between the left atrial appendage and the left atrium in atrial fibrillation patients

Figures

• 

  Figure 1. The measurement of myocardial thickness at the junction of the LAA and the LA. (A–C) three consecutive cross-sectional images of myocardium at the junction of the LAA and the LA was shown. The thickness of myocardium on three layers was measured, and Figure 1A was determined to be the thickest myocardium layer. LA and LAA are colored. (D) A sagittal image displayed the thickest part of the myocardium, indicated by the crossroads. It was located by positioning the marker on the myocardium of the aforementioned cross-section.

• 

  Figure 2. Measurement of LAA opening. (A) A coronal image of the LAA opening. The red line was along the plane of the LAA opening. (B) A sagittal image of the LAA opening. (C) A transverse image of the LAA opening.

• 

  Figure 3. Measurement of the long diameter, short diameter, circumference, and area of the LAA opening. (A) On the plane depicting the LAA opening in Figure 2C, the longest diameter of the LAA opening was measured, and a line segment perpendicular to it was drawn to represent the short diameter of the LAA opening. (B) On the same image of A, the circumference and area of the LAA opening was automatically calculated by outlining the opening of the LAA.

• 

  Figure 4. Measurement of LAA volume, LA volume, and LAA depth. (A) The post-processing software automatically calculate the volume of the LAA and the LA. LAA was segmented at the connection between the LAA and the LA. (B) LAA was obtained separately. The length from the tip of the LAA to the midpoint of the LAA opening was measured as the depth of the LAA.

• 

  Figure 5. ROI delineation. On the axial image of cardiac CT, the myocardium at the junction of the LAA and the LA is delineated (green). The length of ROI was 1 cm in total.

• 

  Figure 6. Flowchart of patient recruitment and study design. AF, Atrial fibrillation.

• 

  Figure 7. LAASO regression model dimensionality reduction. Lambda(λ) in the shrinkage coefficient represented the adjustment parameter. The vertical axis represented the mean square error, and the horizontal axis represents λ.

• 

  Figure 8. Selected features and weights. Fourteen features were extracted, which were showed on X-axis. The weights of features were displayed on Y-axis. The positive or negative weights indicate positive or negative correlation with AF.

• 

  Figure 9. The nomogram of a predictive model combining radiomics and clinical features.

• 

  Figure 10. The ROC curve. (A) The ROC curve of the radiomics model; (B) The ROC curve of the clinical feature model; (C) The ROC curve of the model combining radiomics and clinical features.

• 

  Figure 11. Calibration curves of combining radiomics and clinical features. When the solid line (predicted situation) is closer to the dashed line (actual situation), the calibration effect of the model is better.

• 

  Figure 12. Decision curve. The Y-axis represents the net benefit and the X-axis represents the threshold probability. The gray curve represents the hypothesis that all patients develop AF events, and the black curve represents the hypothesis that no patients develop AF.