Detection and prediction of lung tumor motion with fluoroscopy

A new radiotherapy adjunct system capable of tracking tumor motion due to breathing has been proposed. The properties of tumor motion with breathing were determined from fluoroscopic video recorded during patient simulation. Real-time patient motion during treatment was detected with noninvasive motion sensing and the immediate-future tumor position was predicted in correlation with the pre-determined motion properties to guide the treatment field. Here, the authors report their solutions to some critical problems with this system. The pincushion distortion of the fluoroscopic imaging system was corrected with a 5/sup th/ order two-dimensional polynomial unwarping function based on the measurement with a fine lead-grid phantom. Temporal Fourier transformation of the fluoroscopic video provided a static view of the motion field. Regions with different motion ranges can be quickly measured. In order to obtain real time tumor position with noninvasive motion sensing, a calibration technique has also been developed. Phantom simulations revealed that the motion monitoring system with prediction capability can minimize phase-introduced errors and allow more conformal treatment fields. Accurate calibrations of the radiographic imaging system and the motion detection system can significantly improve the systematic uncertainties.