Pancreatic ductal adenocarcinoma is among the most lethal cancers, in part because its precursor lesions, pancreatic intraepithelial neoplasias, arise within the ductal epithelium, and are not visible with current imaging. Delivering a targeted treatment to this epithelial layer, while sparing the surrounding tissue and avoiding thermal damage with catheter based electroporation treatment, is a problem that conventional treatment techniques are poorly suited to solve. This thesis develops a computational framework to model and optimize a localized, low-voltage pulsed electric field protocol for catheter-based electroporation of the pancreatic duct.

A coupled finite element model was implemented in FEniCSx, solving the quasi-static charge conservation equation together with the Pennes bioheat equation under a field- and temperature-dependent conductivity. A steady-state approximation of the periodic thermal response was used to evaluate the high-frequency pulsing protocol. A Bayesian optimization routine, implemented in Optuna and parallelised with MPI, explored four design variables, applied voltage, pulse ratio, electrode gap, and electrode width. The objective was constructed to maximise electroporation of the ductal epithelium, penalise penetration into the deep tissue, and enforce a maximum steady-state temperature of $50^\circ$C.

Two sets of weights for the objective function was used for the wall-contact DuCath catheter and produced configurations achieving epithelial electroporation at maximum temperatures of $39.02^\circ$C and $44.73^\circ$C respectively, with the choice between them representing an explicit and tunable trade-off between coverage and deep tissue sparing. Under the same framework, no thermally safe electroporating configuration was found for a centrally positioned catheter across 50 trials of bayesian search, indicating that wall contact is not a marginal improvement but a qualitative change in the feasibility of the therapy.