Even though seemingly endless at the beginning, there is only so much you can do in five years of PhD studies. Importantly, the studies of this thesis are not providing finalised or completed forms of the presented methods, but are rather forming a foundation on which further developments, refinements, and clinical investigations can be performed. With such, there are several areas were future work can be envisaged and where continued research could lead to a matured clinical usage of the presented techniques.
SWE seems to have potential for improved atherosclerotic plaque characterisation, however, as indicated by the results of Study I-IV cardiovascular structures impose complex wave propagation response. Detailed technical investigations could thus be highly beneficial in understanding observed clinical response. Patient-specific modelling has been utilised to study acoustic response in atherosclerotic plaques [391-394], and experimental inverse modelling has also been attempted to describe mechanical plaque behaviour [399, 435]. Incorporating anisotropic or viscous tissue behaviour into such models could shed further light on the complex response of atherosclerotic SWE, and coupling pre-clinical data as such from Study III with histological validations could also be highly beneficial in deciphering defined tissue response. In particular SWE-based attenuation-mapping [436] could have specific potential for diagnostic plaque imaging, considering the importance of soft-tissue components on predicted plaque vulnerability.
If staying with the derived outputs of Study IV, the diagnostic abilities of SWE group and phase velocity should be studied in larger cohorts in order to clarify their clinical implications. Additionally, alternative image acquisition protocols should be explored to improve the frequency range over which phase velocities can be retrieved, as well as to map the robustness of the evaluated outputs. To this, shear wave tracking along tilted vessels or automated ECG triggered pushing could also be introduced to overcome potential practical issues related to clinical ultrasound elastography.
Looking ahead with regards to cardiovascular SWE, the transition from 2D to 3D imaging poses interesting technical challenges, and even though 3D strain imaging has been attempted by others [186, 196, 437] 3D ultrasound elastography plaque
imaging remains to be explored. Additionally, translation of SWE into myocardial assessment has clear clinical implications, especially considering the constitutive myocardial changes observed in heart failure patients [191], or following myocardial infarction [438]. Myocardial elastography is under exploration [439- 441], however developments to handle temporal cardiac movements and aggravated image quality are still required.
For cardiovascular flow assessment, νWERP has exciting potential in advancing clinical assessments into previously unexplored vasculatures. Study VII showcases only one example of where novel hemodynamic assessment improves our understanding of cardiovascular disease, but many other application areas could be explored. Renal failure has been coupled to hemodynamic alterations [442], intraventricular flow and pressure changes seem determining in the long-term development of hypertrophic cardiomyopathy [89, 90], and flow disruptions are intrinsically coupled to the initiation and development of atherosclerosis [141, 443, 444]. Using νWERP on cerebrovascular flows, and mapping plaque phenotype to hemodynamic environment would be a very exciting method translation, coupling the developments in Study V-VII to the clinical question explored in Study I-IV. With regards to the practical implementation of νWERP, development work is still required to create an automated and clinician-friendly user interface. The necessity of full-field data might also be slightly limiting in a clinical setting, wherefore translation into 2D flow assessments (as attempted with WERP in [64, 90]) could be explored. However, with the added clinical value of 4D flow MRI showcased in a number of studies [19, 22, 81, 218, 220, 413] and with 2D/3D Doppler ultrasound showing promising potential in preliminary studies [184-186] the clinical incorporation of full-field flow measurements could very well be realised in the near future.
The turbulence-driven behaviour in Study VI should also be explored in a clinical setting. Firstly, in-vivo validation of the method would be advisable, and secondly,
specific method behaviour as a function of covariance or mean field quality could also be mapped in detail. Furthermore, with incoherent flow associated to a range of cardiovascular pathologies, the method could be applied in novel clinical settings exploring turbulence-induced pulmonary hypertension, mapping post- stenotic relative pressure in relation to bicuspid aortic valve disease and thoracic aortic aneurysms, or again assessing turbulent energy dissipation around plaque- prone vasculatures.
Lastly some methodological features of νWERP could also be further evaluated, implementing a temporally changing virtual field (suitable for e.g. intraventricular flows where the geometrical domain varies over time), or evaluating the effect of different virtual fields or boundary conditions on method accuracy and robustness. For the multigrid reconstruction approach, a majority of the evaluation was performed on simplified numerical phantoms and further detailed validation in a
clinical setting could be beneficial. The introduced concept of domain splitting however opens up for interesting technical advancements, one being multithreaded parallelisation where full-field high-resolution reconstructions could be achieved by distributed multigrid solvers. Incorporating energy-dependent operators would be another possibility, especially if attempting an implementation towards spectral CT.