Prediction of respiratory motion using state space models

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K08703
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Medical Physics and Radiological Technology
• Research Institution: Teikyo University
• Principal Investigator: Kotoku Jun'ichi 帝京大学, 医療技術学部, 教授 (70450195)
• Co-Investigator(Kenkyū-buntansha): 芳賀 昭弘 徳島大学, 大学院医歯薬学研究部(医学系), 教授 (30448021)
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: データ同化 / 状態空間モデル

Outline of Final Research Achievements

The objective of this study is to build a reasonable predictive model for respiratory motion. To accomplish this aim, we combined the probabilistic model and non-linear dynamics to represent a complex behavior of respiratory motion. First, we measured respiratory motion using a depth camera. The accuracy of the system was within 1 mm. Then, time series data of respiratory motion was modeled as a harmonic oscillator with Hamilton's equation rewritten as a symplectic form. The entire measuring system was modeled as a state space model with a particle filter, which filters measured data and predicts the future. The prediction accuracy of our model was within 20 percent for four seconds ahead. The proposed approach is to predict future respiratory motion.

Free Research Field

医学物理