论文和技术资料

Electrochemotherapy (ECT) combines chemotherapy with the delivery of short, high-voltage electric pulses that transiently permeabilize cell membranes (a phenomenon known as electroporation), enabling enhanced intracellular drug delivery. Treatment success depends on complete tumor coverage with an electric field strength exceeding the electroporation threshold. Because electric field distribution is strongly influenced by patient anatomy, electrode placement, and nonlinear tissue properties, treatment planning relies on patient-specific numerical modeling. However, electrode positioning is still largely performed manually through repeated simulations, making the process time-consuming and highly dependent on operator experience.

A simulation-driven optimization framework that automates treatment planning by coupling COMSOL Multiphysics with MATLAB via LiveLink for MATLAB was developed [1]. Instead of using computationally intensive global optimization methods, the framework iteratively updates electrode positions based on the spatial distribution of the computed electric field, enabling efficient convergence to an optimal treatment plan. The algorithm is currently designed for ECT of vertebral tumors using a transpedicular approach.

Patient-specific vertebral models were reconstructed from CT images using 3D Slicer and Materialise 3-matic, then imported into COMSOL. A spherical tumor, surrounding block of adipose tissue, and two needle electrodes were added to complete the model geometry (Figure 1A). Initial electrode positions were determined from the CT images. Electric field distribution was computed using the Electric Currents interface of the AC/DC Module by solving the stationary current conservation equation (Figure 1B). Tissue conductivity was modeled as a nonlinear function of the local electric field using smoothed Heaviside functions to capture conductivity changes associated with electroporation.

The optimization algorithm was implemented entirely in MATLAB, with COMSOL serving as the finite-element simulation engine. LiveLink for MATLAB enabled direct control of the COMSOL model throughout the optimization process (Figure 2). At each iteration, MATLAB updated electrode geometry and pulse voltage in the model, executed a COMSOL study, retrieved the resulting electric field distribution, identified undertreated regions within the tumor, computed corrective electrode displacements, and initiated the next simulation. This closed-loop workflow continued until complete tumor coverage was achieved or predefined hardware limits were reached. Apart from the initial setup, the framework required no manual interaction during the optimization process.

The workflow was evaluated on 108 patient-derived vertebral models covering multiple vertebral levels, tumor sizes, and tumor locations. The optimization converged in an average of 4.9 iterations and required an average of 71 s to produce a treatment plan. Complete tumor coverage was achieved in most cases (Figure 1C), and the remaining models reached near-complete coverage after voltage adjustment. By exploiting the spatial information contained in the electric field solution, rather than relying on computationally intensive genetic algorithms, the framework substantially reduced computation time while minimizing operator intervention.

This work demonstrates how COMSOL Multiphysics can be integrated into an automated, simulation-driven optimization workflow through LiveLink for MATLAB. Using COMSOL as a finite element solver embedded within an external decision-making algorithm enables efficient, iterative treatment planning while preserving the flexibility of MATLAB for optimization. Although demonstrated for electrochemotherapy of vertebral tumors, the proposed workflow is readily transferable to other electroporation-based treatments.