Purpose Minimally invasive treatment of solid cancers, especially in the breast and liver, remains clinically challenging, despite a variety of treatment modalities, including radiofrequency ablation (RFA), microwave ablation or highintensity focused ultrasound. Each treatment modality has advantages and disadvantages, but all are limited by placement of a probe or US beam in the target tissue for tumor ablation and monitoring. The placement is difficult when the tumor is surrounded by large blood vessels or organs. Patientspecific image-based 3D modeling for thermal ablation simulation was developed to optimize treatment protocols that improve treatment efficacy. Methods A tissue-mimicking breast gel phantom was used to develop an image-based 3D computer-aided design (CAD) model for the evaluation of a planned RF ablation. First, the tissue-mimicking gel was cast in a breast mold to create a 3D breast phantom, which contained a simulated solid tumor. Second, the phantom was imaged in a medical MRI scanner using a standard breast imaging MR sequence. Third, the MR images were converted into a 3D CAD model using commercial software (ScanIP, Simpleware), which was input into another commercial package (COMSOL Multiphysics) for RFA simulation and treatment planning using a finite element method (FEM). For validation of the model, the breast phantom was experimentally ablated using a commercial (RITA) RFA electrode and a bipolar needle with an electrosurgi- Results A 3D CAD model, created from MR images of the complex breast phantom, was successfully integrated with an RFA electrode to perform FEM ablation simulation. The ablation volumes achieved both in the FEM simulation and the experimental test were equivalent, indicating that patientspecific models can be implemented for pre-treatment planning of solid tumor ablation. Conclusion A tissue-mimicking breast gel phantom and its MR images were used to perform FEM 3D modeling and validation by experimental thermal ablation of the tumor. Similar patient-specific models can be created from preoperative images and used to perform finite element analysis to plan radiofrequency ablation. Clinically, the method can be implemented for pre-treatment planning to predict the effect of an individual's tissue environment on the ablation process, and this may improve the therapeutic efficacy.