Abstract:
Computer vision (CV) algorithms and mechanic models have been used to track the deformation of soft organs and to predict the locations of their internal structures. Before applying these algorithms to clinical trials, validation studies are required to check their accuracy and robustness. In this paper we present a testing platform with both hardware and software implementations for such a purpose. The hardware consists of a custom-built force inducing apparatus, and a dual vision system with both mono and stereo vision cameras. The software includes a combination of CV techniques and in-house implemented Finite Element Modelling (FEM) solvers accelerated by the parallel computing engine CUDA of NVidia. In the experiments on an abdominal biopsy phantom, the displacements of tumours are detected by the vision system and used to validate the FEM solvers. Initial results show good agreements between the FEM solvers and that detected from the vision system in two of three tumours. The prediction discrepancy in one tumour requires further investigation by imposing more accurate boundary conditions to the FEM solvers. In summary the construction units of a prototype platform were implemented. With future improvements the platform can be used for elasticity experiments for soft objects.