Convolutional Neural Network for the Detection of End-Diastole and End-Systole Frames in Free-Breathing Cardiac Magnetic Resonance Imaging

Figures

• 

  Figure 1. The changing position of LV in ED frame in six cardiac cycles.

• 

  Figure 2. The changing LV center (red) along with the diaphragm (green) over multiple cardiac cycles.

• 

  Figure 3. The block diagram of CNN for the heart localization.

• 

  Figure 4. The input images (b) and the corresponding binary masks (a) produced by the trained network. The center of the output image is the same as that of the input image.

• 

  Figure 5. The block diagram of the LV localization.

• 

  Figure 6. Input image (a) and labeled image (b) of the STACOM 2011 data and free-breathing CMR data.

• 

  Figure 7. The result of the LV ellipse fitting in one cardiac frame: (a) the center of the endocardial contour; (b) region growing result; (c) blood edge; (d) convex hull curve; (e) LV ellipse fitting curve and the LV center.

• 

  Figure 8. Respiratory motion and cardiac motion signal detection. The horizontal axis represents the cardiac image frames numbered 1 to 84. The vertical axis represents the relative position which is the distance between the LV center and the boundary of the image.

• 

  Figure 9. The detection at the end-expiration. The blue circle denotes the end-expiration. Frames (A) and (B) are the first frame and last frame in the end-expiration.

• 

  Figure 10. The CMR images of the apical slice (a) and the basal slice (b).

• 

  Figure 11. Representative example of the segmented LV endocardial contours of the ED frame (a) and the ES frame (b) by the CNN model from the first slice (apex) to the tenth slice (basal) in the end-expiration.