Registration of whole-body radiographic images is an important task in analysis of the disease progression and assessment of responses to therapies. Numerous registration algorithms have been successfully used in applications where differences between source and target images are relatively small. However, registration of whole-body CT scans remains extremely challenging for such algorithms as it requires taking large deformations of body organs and articulated skeletal motions into account. For registration problems involving large differences between source and target images, registration using biomechanical models has been recommended in the literature. Therefore, in this study, we propose a patient-specific nonlinear finite element model to predict the movements and deformations of body organs for the whole-body CT image registration. We conducted a verification example in which a patient-specific torso model was implemented using a suite of nonlinear finite element algorithms we previously developed, verified and successfully used in neuroimaging registration. When defining the patient-specific geometry for the generation of computational grid for our model, we abandoned the time-consuming hard segmentation of radiographic images typically used in patient-specific biomechanical modelling to divide the body into non-overlapping constituents with different material properties. Instead, an automated Fuzzy C-Means (FCM) algorithm for tissue classification was applied to assign the constitutive properties at finite element mesh integration points. The loading was defined as a prescribed displacement of the vertebrae (treated as articulated rigid bodies) between the two CT images. Contours of the abdominal organs obtained by warping the source image using the deformation field within the body predicted using our patient-specific finite element model differed by only up to only two voxels from the actual organs’ contours in the target image. These results can be regarded as encouraging step in confirming feasibility of conducting accurate registration of whole-body CT images using nonlinear finite element models without the necessity for time-consuming image segmentation when building patient-specific finite element meshes.