Characteristics of left ventricular dysfunction in repaired tetralogy of Fallot: A multi-institutional deep learning analysis of regional strain and dyssynchrony

Figures

• 

  Fig. 1. Deep learning synthetic strain analysis for a patient with tetralogy of Fallot, showing radial strain and strain rate maps and time curves derived from the inferred myocardial velocity. Using the strain and strain rate time curves, we calculate measures of dyssynchrony such as time to peak, strain-rate cross correlation, and uniformity ratio estimates. SSFP steady-state free precession.

• 

  Fig. 2. Sixteen-segment AHA bulls-eye plots of the mean peak radial strain and mean time to peak radial strain for patients with tetralogy of Fallot (n = 198) and healthy controls (n = 21). Relative to controls, patients with tetralogy of Fallot demonstrate reduced septal strain with compensation in the lateral wall. AHA American Heart Association, TOF tetralogy of Fallot.

• 

  Fig. 3. Hierarchical clustering of tetralogy of Fallot patients into unique contraction patterns using standardized regional strain and dyssynchrony measurements. Patient phonetic IDs are shown on the x-axis with color codes used to distinguish each phenotypic cluster. For each metric, the data are normalized and colored such that values greater than the mean are red, while values less than the mean are blue. We discovered multiple distinct contraction patterns from the regional measures of LV contraction strength and timing.

• 

  Fig. 4. Sixteen-segment AHA bulls-eye plots of the mean peak radial strain and mean time to peak radial strain for clusters 1, 2, 3, and 4. Clusters 1, 3, and 4 demonstrate decreased strains in the septal segments, with compensation in the lateral wall. In contrast, cluster 2 demonstrates globally reduced radial strains without compensation in the lateral wall. Cluster 1 was also characterized by significantly increased time to peak radial strain (indicates increased dyssynchrony) in the apical, mid, and basal septal segments. AHA American Heart Association.

• 

  Fig. 5. Representative patients from clusters 1, 2, 3, and 4. The cluster 1 patient demonstrates evidence of RV volume overload (increased RVEDVi) and septal dysfunction (increased dyssynchrony, decreased septal radial strain). Representative of cluster 2 is a patient with decreased LV function without evidence of RV volume overload or dyssynchrony. Representative of cluster 3 is a patient with adequate compensation in both the right and left ventricles. Representative of cluster 4 is a patient with mild apical dyssynchrony and septal dysfunction. LV left ventricle, RV right ventricle, EDVi end-diastolic volume, EF ejection fraction, RS radial strain, sdTTP standard deviation of time to peak.

• 

  Fig. 6. Kaplan-Meier survival curves and boxplots showing time since CMR evaluation until progression to PVR for patients in LV mechanical clusters 1, 2, 3, and 4. Patients with the cluster 1 contraction pattern, which was characterized by dyssynchrony and decreased radial strain in the anteroseptal and inferoseptal segments, showed a statistically significant decrease in survival relative to the well-compensated patients in clusters 3 and 4. Significance was assessed with pairwise log-rank tests. *p < 0.05; ***p < 0.001; NS not statistically significant, LV left ventricular, CMR cardiovascular magnetic resonance, PVR pulmonary valve replacement.

• 

  Fig. 7. Cox proportional hazards regression analysis showing that the cluster 1 contraction pattern (severe septal dysfunction) is the only LV metric associated with clinical progression to PVR, which is likely reflective of the concomitant RV dysfunction that occurs with this contraction pattern. Of note, when controlling for this pattern of LV dysfunction, no other LV metric remains independently associated with progression to PVR. LV left ventricle, RV right ventricle, PVR pulmonary valve replacement, EDVi indexed end diastolic volume, GRS global radial strain, EF ejection fraction, SVi indexed stroke volume, ESVi indexed end-systolic volume.