上海科技大学知识管理系统

Magnetic resonance thermometry imaging (MRTI) is essen-tial for non-invasive temperature monitoring in thermal therapies, such as high-intensity focused ultrasound (HIFU) and hyperthermia, where precise temperature control and high temporal resolution are critical. This paper presents a novel framework for 3D temperature field reconstruction from 2D tri-planar MR thermometry images, employing cross-view representation to effectively integrate tri-planar information and achieve accurate temperature field estimation. By leveraging 2D-to-3D learning, our approach offers an innovative solution for reconstructing 3D thermometry fields. Given the scarcity of 3D MRI thermometry datasets, we generated a synthetic dataset using the Pennes bio-heat transfer model to simulate time-series temperature fields. Extensive experiments demonstrate the feasibility of the proposed method and achieved highly accurate results for 3D temperature field reconstruction.