[2D01-05] 放射線医学と生物影響

[2D02] Three-dimensional CT image prediction from multiple points tracking for delay compensation in lung radiotherapy

*Michel Pohl¹, Mitsuru Uesaka¹, Kazuyuki Demachi¹, Ritu Bhusal Chhatkuli², Akihiro Haga³ (1. Univ. of Tokyo, 2. NIRS, 3. Tokushima Univ.)

キーワード：Optical flow, Lung cancer, Radiotherapy, Prediction

We propose a method to predict 3D CT data of a patient with lung cancer to compensate of the delay of the treatment machine. This will lead to more accurate estimation of the tumor position in real-time and thus to more accurate dose delivery. Indeed, we would like to ultimately decrease the amount of irradiation of healthy tissues which may cause unwanted complications. First, we calculate the optical flow and use it to track several points inside the body, representing markers. Linear regression is used to predict the motion of these points. The motion of each individual pixel is recovered from linear regression and the predicted images are reconstructed using forward-warping. The cross correlation between predicted and original images is 0.795+/-0.126.