We propose a robustness enhancement methodology of motion compensation for non-invasive HIFU theragnostic system. The methodology is a combination of US-based visual tracking and skin-motion tracking by optical sensor. In this paper, we report the kidney motion model, which estimates kidney position from the measured skin position moving during respirations. Before the treatment, skin position and kidney position are measured simultaneously. To measure skin position, external markers are attached on patient's abdomen, and tracked by camera. To measure motion of kidney, we apply template-matching scheme for the stream of US images. Therefore, the correlation between measured skin position and kidney position is computed. During treatment, kidney position is estimated from the measured skin position using the correlation.