Patient‐Specific In Silico Prediction of Outcomes of Partial Continuous‐Flow LVAD Treatment in Peripartum Cardiomyopathy

Figures

• 

  FIGURE 1. (a) Short‐ and long‐axis views of biventricular meshed geometries of six PPCM patients illustrating the wall thickness of the left and right ventricular free wall and septal wall. In each image, the LV is on the right, and the RV is on the left. Ejection fraction is indicated by the number in the patient code. (b) Schematic of coupling of biventricular geometry with circulatory system and LVAD [22]. RM, mitral valve resistance, RA, aortic valve resistance, CSA, systemic arterial compliance, RSYS, systemic arterial resistance, CSV, systemic venous compliance, RT, tricuspid valve resistance, RP, pulmonary valve resistance, CPA, pulmonary arterial compliance, RP, pulmonary valve resistance, CPV, pulmonary venous compliance, LVAD, left ventricular assist device.

• 

  FIGURE 2. (a) Calibrated in silico LV passive filling curves for each patient with volume normalised to the unloaded volume V 0 (Table 3). Compliance of the LV myocardium increased with an increase in ejection fraction (represented by the numerical value in the model code). The model for patient N06D with the lowest ejection fraction of 6% had the stiffest LV myocardium. (b) Solution convergence of the patient‐specific models for multiple cardiac cycles expressed as mean squared error for LV pressure (top) and volume (bottom) of the last two cardiac cycles versus LVAD speed. The mean squared error is based on all pressure and volume values from the LV pressure‐volume loops of the 5th and 6th cardiac cycle for models without LVAD support (LVAD speed of 0 RPM) and the 9th and 10th cardiac cycle for models with LVAD support.

• 

  FIGURE 3. Comparison of in vivo and FE‐predicted (Model) values for (a) LV volume at end‐diastole (EDV) and end‐systole (ESV) and (b) LV pressure at end‐diastole (EDP). The models accurately replicated in vivo end‐diastolic and end‐systolic LV volumes but overestimated end‐diastolic LV pressure for patients with EF > 20% and underestimated the LV pressure for patients with EF < 20%.

• 

  FIGURE 4. (a) Pressure‐volume loops for the LV (solid lines) and RV (dashed lines) versus LVAD speed for all patients. Apart from R45A, the LV pressure‐volume curve shifted to the left and downwards, whereas the RV pressure‐volume curve shifted to the right after implementing the LVAD support. (b) Changes in the LV ejection fraction (EF) with an increase in LVAD speeds. The zero revolutions per minute (RPM) on the horizontal axis represents models without the LVAD. Increasing the LVAD speeds led to a steady increase in EF in some models, while there was a decrease or marginal change in EF in others. (* Indicates significant differences between the average ejection fraction for models without and with LVAD at each speed, p < 0.05).

• 

  FIGURE 5. Variation of LV stroke volume contributions with LVAD speed for each PPCM model (zero revolutions per minute (RPM) represents models without LVAD). The LV and LVAD jointly contributed to stroke volume in the four models with EF = 18% and higher, while the LVAD was the only contributor in the two models with the lowest EF of 12% and 6%.

• 

  FIGURE 6. Probability density curves showing the LV myofibre stress distribution for each patient at (a) end‐diastole (ED) and (b) end‐systole (ES). LV myofibre stress distribution was different for models without LVAD support but similar for different LVAD speeds for each patient except R45A. For that patient, similar stress distributions were predicted with and without LVAD at end‐diastole. The peak probability density was higher with than without LVAD support and increased with increasing LVAD speed.

• 

  FIGURE 7. Representative contour plots illustrating two main observations for the biventricular myofibre stress in PPCM with increasing LVAD speed at end‐diastole (ED) and end‐systole (ES). Model R45A illustrates the first observation of a decrease in myofibre stress with an increase in LVAD speed at end‐diastole, but an increase in myofibre stress with an increase in LVAD speed from 8000 to 12,000 rpm at end‐systole. This observation applies to models R45A and N39A. Model N06D illustrates the second observation with a decrease in myofibre stress with increasing LVAD speed, both at end‐diastole and end‐systole, which applies to models R20A, N18A, N12A and N06D.

• 

  FIGURE 8. Patient mean myofibre stresses versus LVAD speed in LV and RV at end‐diastole (ED) and end‐systole (ES). The LVAD support led to a decrease in myofibre stress in the LV at end‐diastole (a) and end‐systole (b). In the RV, LVAD support decreased myofibre stress only at end‐systole (d) and not at end‐diastole (c). Zero revolutions per minute (RPM) represents models without the LVAD.

• 

  FIGURE 9. Representative contour plots displaying the two main trends of the biventricular myofibre strain distribution for the PPCM patients with and without the LVAD support at end‐diastole (ED) and end‐systole (ES). For patient R45A, end‐diastolic myofibre strains decreased from no LVAD to the LVAD at 8000 rpm and decreased further from 8000 to 12,000 rpm LVAD speed. At end‐systole, the myofibre strains decreased from no LVAD to the LVAD at 8000 rpm but increased marginally with increasing LVAD speed from 8000 to 12,000 rpm. For patient N06D, representative of all other patients, end‐diastolic and end‐systolic myofibre strain decreased from no LVAD to the LVAD at 8000 rpm and decreased further from 8000 to 12,000 rpm LVAD speed.