Bioengineered skin was initially developed for clinical application in reconstructive surgery, but nowadays it is also considered as a suitable skin model for research. Research applications range from assessment of cosmetic and pharmaceutical products to the modeling of pathological skin conditions. For example, psoriasis is a complicated skin disease that strongly concerns the scientific community because of its unclear pathogenesis and the current inability to cure it. The Pharmacogenomics Lab of ETH has been working on the development of psoriatic skin models in order to investigate the driving mechanisms of the disease and to later implement high-throughput drug screening to explore treatment possibilities. However, the manual production into transwell inserts results in unstable, self-contracting and inconsistent skin models, pain points that the Pharmacogenomics Lab would like to eliminate. A recent collaboration between the Product Development Group and the Pharmacogenomics Lab has been recently initiated in an attempt to standardize the fabrication of the psoriatic skin models. For this purpose, this Master Thesis focused on the investigation of the potentials and limitations of automated injection molding, a method developed by the Product Development Group, in the generation of consistent and representative dermo-epidermal models for psoriatic skin.
After a series of experiments, it was seen that in-mold fabrication achieves homogeneous and consistent dermal hydrogels that maintain their dimensions throughout the whole cultivation period regardless of the collagen concentration employed for the dermal matrix, while automated injection molding allows the utilization of collagen concentrations higher than the unstable 5mg/ml, traditionally used in the manual fabrication method. All collagen concentrations exhibited similar biological performance, but the optimum fibroblasts viability was achieved in case of 5mg/mL, meaning that the whole injection molding process should be faster and even simpler to fully prevail over manual fabrication of skin models and better fit the requirements of skin disease research. Keratinocytes viability and differentiation was similar for all mold designs, all collagen concentrations and both fabrication methods. Development of a first psoriatic model, consisting of psoriatic fibroblasts and healthy keratinocytes, did not show any direct effect of diseased fibroblasts on keratinocytes proliferation and differentiation as the epidermal layer of the psoriatic model was very similar to this one of the healthy model.
Evaluation of the automated seeding of keratinocytes for further standardization of the process and elongation of the psoriatic model’s cultivation period for a more thorough study should be the next steps. In the long-term, incorporation of all the psoriasis-related cells and molecules in a simple and fast way, as well as further advancements in the mold design to facilitate a direct high-throughput drug screening should be considered.

Objective: Estimators to accurately monitor ventricular relaxation, cardiac contractility and cardiac output (CO) during support of a left ventricular assist device (LVAD), based on the left ventricular pressure (LVP) and pump-intrinsic signals, were developed. The aim was to continuously monitor these cardiac parameters, hereby tracking the progression of disease and ensuring effectiveness of treatment.
Methods: Ventricular relaxation was estimated using the time constant of LVP decay, τ_p. Cardiac contractility was based on the maximum derivative of LVP, dP/dt_max versus the end-diastolic pressure (EDP). The CO was estimated using a hybrid monitoring scheme, that used the estimated pump flow (PF) in cardiac cycles with a closed aortic valve and a cardiovascular model in cycles with an open valve. The states of the cardiovascular model were estimated using an extended Kalman filter (EKF). The opening status of the aortic valve (AoV) was determined using a machine learning model that employed a quadratic discriminant analysis based on features of the LVP waveform. The estimators were evaluated using in-vitro and in vivo experiments on pigs that were earlier conducted within the Product Devel- opment Group Zurich (pd|z). The experiments contained preload, afterload, contractility and relaxation variations. The CO was evaluated in experiments from three control strategies (i.e. constant speed, physiological control and multi-objective control).
Results: The index τ_p uniformly resembled changes in left-ventricular relaxation. dP/dt_max versus EDP captured variations in cardiac contractility. Both were preload and afterload independent. The CO was estimated with a mean relative error of 11% compared to the true CO.
Conclusion: The feasibility and accuracy of using LVP based cardiac indices in LVAD assisted circulations was demonstrated.
Significance: The clinical implementation could provide clinicians with valuable and reliable information of the circulation of the patient, allowing for monitoring of progression of disease and improvement of treatment.