96 | Chapter 5 was included to account for large geometrical differences between patients and geometrical changes over time. Wall volumes were not included in the parameter subset because they were unidentifiable given the available measurements. Dependencies in strain were partially included by including strain rate and strain differences. Based on the used likelihood function, posterior distributions were estimated with a relatively wide variance (Figure 4), suggesting not all parameters are identifiable. The low reproducibility in some parameters (HDI 95% CI (0-79%)) is probably related to this unidentifiability. Heterogeneity in model parameters is, however, well preserved, suggesting that measurements that are sensitive to segment-averaged model parameters should also be included in the likelihood function. Further prospective studies could investigate the error propagation of dependent and independent uncertainties, whether all components of the likelihood are essential to include, and which other measurements should be included to increase the identifiability of the model parameters. Derived tissue properties were estimated more precise and reproducible compared to model parameters, suggesting that different parameter combinations can result in the same hemodynamic state. Mechanics of the three RV segments were modelled with the same mathematical equations, however, they have different interactions with the surrounding walls as shown in Figure 5. Compliance in the basal segment was estimated more precise compared to the other segments (Figure 6). This results from the non-linear behaviour of the model, as basal model parameters were differently estimated due to basal deformation abnormalities. Therefore, compliance in the basal segment was less correlated with the other segments. In this study, we used a single definition for myocardial contractility and compliance related to other more global definitions. There is no consensus on a single indicator for contractility and compliance, and often multiple (non-invasive) measures are used to get an impression. For contractility, the maximum pressure-time derivative dP/dtmax is the most commonly used index of contractility in the field of drug safety assessment.43 Although this measure is preload and afterload dependent, the regional stress-time derivative as local equivalent gives insight in the regional differences in RV contractile function. Other global measures have been proposed to bypass preload and afterload dependencies, such as dP/dtmax at a specific pressure43 or end-systolic pressure-volume relation44. New techniques might be useful for future validation of RV tissue properties, such as shear wave imaging45 to quantify cardiac stiffness. The gold-standard assessment of RV stiffness (inverse of compliance) is the end-diastolic pressure-volume relation.46 The local equivalent is the models material law describing the stress-sarcomere length relation. The actual amount of stress prescribed by this law depends on the sarcomere length during the cycle.19 Due to the complexity of the model, which includes mechanics based on sarcomere length, an accurate estimation of compliance is difficult. The compliance measure as used in this study only includes the compliance at the end diastolic sarcomere length and is therefore load-dependent. To obtain a load-independent measure, more information on the loading conditions should be included in the likelihood distribution. Case study and future research directions The two subjects included in the case study showed different behaviour over time. The first subject developed an abnormal basal RV deformation pattern during follow-up which was reflected in changes in estimated local tissue properties. The second subject did not develop clear deformation abnormalities, but did develop slight abnormal heterogeneity in tissue properties. In both cases, only small changes in estimations were observed from baseline
RkJQdWJsaXNoZXIy MTk4NDMw