Automatic Diagnosis by Compact Portable Ultrasound Robot: State Estimation of Internal Organs with Steady-State Kalman Filter

In recent years, research has been very active on the development of artificial intelligence, robot technology, and support for proper image acquisition in ultrasound diagnosis. A conventional problem using robot technology is that robots themselves are large and complicated mechanisms. If a robot is large, there is a restriction where it can be used; that is, a certain amount of space is necessary. In consideration of these constraints, in this research, we developed a compact medical robot holding an ultrasound probe that can easily perform at-home diagnosis that compensates for organ movement. When the robot automatically diagnoses organs, it is necessary to scan organs with the ultrasound probe over the same cross-section always aligned with the center of the organ. Based on the present research, in order to compensate for the movement of the phantom with movement of the ultrasound probe in the ultrasound images, the movement of the phantom in the ultrasound images is analyzed. As a method, Kalman filter model with linear Gaussian noise is applied to position observations obtained by template matching, and system noise and observation noise are estimated in object state estimation. We also constructed a steady-state Kalman filter with asymptotic stability using solutions from the Riccati equation. Furthermore, verification experiments were carried out with the model on dataset acquired in previous research, and the position and velocity of the phantom were analyzed.