Accuracy improvement of time delay correction method for PET-based tumor tracking

The world's first open type PET, named OpenPET, is being developed at the National Institute of Radiological Sciences. We are aiming to employ the OpenPET in radiotherapy and to track the respiratory motion of a tumor in the thoracoabdominal region of a patient. By using PET images, we expect that the tumor itself can be directly visualized without using radio-opaque markers. Our prototype OpenPET system can output reconstructed images at about 2 frames per second with about a 2 s delay; this delay is mainly due to the reconstruction calculation time. Therefore we presented a time delay correction method with belt type piezoelectric sensor for tumor tracking for the OpenPET with support vector regression method (SVR) in the last IEEE NSS-MIC meeting. In this paper, we improved accuracy of this method. In the modified method, we added body surface displacement velocity as a secondary independent variable of the regression line so that it represents hysteresis effect. In addition, to supplement SVR's prediction posterior probability, we introduced hidden Markov model (HMM). As the result, the mean and maximum tracking error was decreased 1.25 mm and 4.70 mm in the most effective case, respectively.