We developed a dedicated computational framework by coupling the lattice-Boltzmann-method (LBM) modeling and the particle-swarm-optimization (PSO) algorithm to search optimal strategies of magnetic nanoparticle (MNP) injection for hyperthermia-based cancer treatment. Two simplified tumor models were considered: a circular model representing geometrically regular tumors and an elliptic model representing geometrically irregular tumors, both sharing the same area. The temperature distribution in the tumor and its surrounding healthy tissue was predicted by solving the Pennes’ bio-heat transfer equation (PBHTE). Both single- and multi-site injection strategies were explored. The results suggest that the multi-site injection strategies generally work well, while the single-site injection strategy fails even on the simplest circular tumor model. The more the injection sites, the better the performance. In particular, when the number of injection sites reaches eight, all temperature requirements can be nearly 100% satisfied in both tumor models. Whether or not including the minimum dose requirement in the objective function only affects the optimization results by less than 2%. The thermal dose was also assessed by considering both temperature and heat exposure time. It was found that the optimal multi-site injection strategies perform reasonably well for both tumor models. Although the setting is only two dimensional and the optimization is on very simplified tumor models, the framework adopted in this present study works well and can provide useful insights into magnetic hyperthermia treatment.@en