Breast Morphology Recognition and Control Based on Uncalibrated Vision

Breast cancer has become the cancer with the highest incidence rate in the world. Robot-assisted puncture is gradually being applied to breast puncture biopsy procedures. In this paper, to address the problem of lesion drift existing in the current work of breast puncture robots, a breast tissue manipulation system based on uncalibrated vision is designed to manipulate breast tissue to move the puncture target to a position that is easy to puncture, optimizing the puncture path to reduce procedure failure rates. Firstly, the Canny operator and SURF algorithm extract the target contour and features from the breast images captured by the camera. An improved algorithm combining stochastic Hough transform and direct least squares is proposed to enhance the accuracy of the target localization in the images. Secondly, uncalibrated vision is introduced into breast tissue manipulation, the image-deformation Jacobi matrix is derived from the system model. Then, the online estimation of the image-deformation Jacobi matrix is carried out using the Unscented Kalman Filter (UKF) algorithm. The robotic arm is controlled by the vision controller to accomplish the manipulation task. Simulations verify the high performance of the control scheme based on the UKF algorithm for online estimation of the image-deformation Jacobian matrix. Finally, breast tissue manipulation experimental results show that the manipulation system successfully shifts the breast tissue deformation target to the desired target position, and the manipulation accuracy meets the clinical requirements. The breast tissue morphology control is achieved, which verifies the effectiveness and feasibility of the design in this paper.