RANSAC algorithm has been implemented for the detection of micro-tools (biopsy metallic needles) inserted in human tissue, and this method has the capacity to detect the position of the needle in real time. However, RANSAC algorithm depends on randomly selected models so that the robustness is poor. Therefore, a Kalman filter has been added in order to increase the stability, and realize an application in a dynamic situation. A constant velocity model was chosen for the Kalman filter. The result from RANSAC algorithm (estimate tip position of needle) was set as one part of the measurement vector. Because the speed of insertion is unknown, the speckle tracking method was added to measure the inserting speed, which is another measurement for Kalman filter. The predicted tip position of Kalman filter was compared with that of RANSAC result. The simulation has been done with the simulated volume in both static situation and dynamic situation. The result shows that in static situation, the root mean square error (RMSE) of Kalman was reduced 44.79% compared with RANSAC; in dynamic situation, the RMSE of Kalman filter was reduce by 26.5%.