Improving Tissue Elasticity Imaging Using A KALMAN Filter-Based Non-Rigid Motion Tracking Algorithm

Abstract

Imaging biomechanical properties of biological tissues under deformation has strong implications to many medical applications such as cancer detection and surgery planning. Quantifying the elasticity of soft tissue using an optical sensor is particularly attractive because it is non-invasive and easy to operate. However, the current computing method is plagued by the existence of noises in the two-frame optical flow solutions. In this thesis, a Kalman filter based tracking algorithm is examined, aiming to improve the quality of cumulative motion over a long video sequence. The proposed method is robust and is capable of handling non-rigid motion that is typical of soft tissue. Experiments of using videos of four rat tissue specimen subject to a biomechanical tensile test indicates that the proposed tracking method is very promising in generating a smooth, accurate, and continuous multiframe motion field. This type of multi-frame motion data not only allows us to compute a more accurate individual strain elastogram, but also provide valuable information for calibrating a series of relative strain images over the entire deformation process.