报告题目:PDE-constrained electrode placement optimization for tumor treating fields: analysis and multi-fidelity computation
报告人:龚荣芳 教授 南京航空航天大学
报告时间:2026年11月24日 19:30-20:30
报告地点:腾讯会议 375-511-911
报告摘要:This talk addresses a PDE-constrained optimization problem for electrode placement, motivated by Tumor Treating Fields (TTFields) therapy. Since the potentials are prescribed on the electrodes, changing the layout moves the Dirichlet–Neumann partition of the boundary, so that the state space itself depends on the design parameters, which complicates both the analysis and simulations. We prove well-posedness of the forward problem, existence of an optimal electrode layout, and continuous dependence of the reduced functional on electrode positions. Transporting the electrode patches by a family of diffeomorphisms and pulling the weak formulation back to a fixed reference, we derive the shape derivative with respect to the electrode positions by the averaged adjoint method, together with the Riemannian gradient that serves as the search direction for the optimization. In discretization, remeshing is computationally cheap but weakens the continuous dependence of both the discrete objective and its gradient on electrode positions. Since layout updates alter the finite‑element space-including its dimensionality and mesh topology—the objective may cease to improve, leading to premature termination of iterations. By contrast, mesh moving preserves this continuous dependence but comes with substantial computational overhead; furthermore, mesh deformation progressively degrades element quality, at which point remeshing becomes unavoidable. Therefore, our approach predominantly employs mesh moving combined with a Riemannian limited‑memory BFGS (L-BFGS) scheme, with remeshing triggered upon significant deterioration of mesh quality. Numerical experiments verify the accuracy of the derivative and show that the method can give an optimal electrode layout.
报告人简介
南航约炮
教授、博士生导师,担任CSIAM反问题与成像专委会委员、中国仿真学会不确定性系统分析与仿真专委会委员、江苏省计算数学分会常务理事等。研究领域为数学物理反问题、医学成像。主持国自然、江苏省自然等基金多项,在NM、IP、IPI、CMAME、JCM等期刊上发表SCI论文40多篇。