Signal processing for predicting an input time series signal and application thereof to predict position of an affected area during radiotherapy

A method performed by a radiotherapy apparatus for irradiating a radiation on an affected area of a patient, comprising:

• generating the radiation;
• measuring a position of the affected area;
• assuming an input time series signal, which represents values at respective time points until a first time point, as a time series signal whose cycle length, which is a length of a cycle, varies with respect to time to carry out a correlation analysis between part of the input time series signal and the input time series signal, calculating multiple cycle lengths, which are respectively at different time points, as a cycle length fluctuation of the input time series signal and reliability of the cycle length fluctuation using a result of the correlation analysis, predicting a cycle length of the time series signal until a second time point after the first time point using the cycle length fluctuation and the reliability of the cycle length fluctuation and calculating, using the result of the correlation analysis and the predicted cycle length, in-phase time information, which represents a time point before the second time point in the input time series signal, and reliability of the in-phase time information, a phase at the time point in the input time series signal becoming same as a phase at the second time point in the time series signal;
• estimating model design information for predicting a value of the time series signal at the second time point and accuracy of the prediction of the value of the time series signal at the second time point using the input time series signal, the time series information of a variation per unit time of the value and the calculated in-phase time information and the calculated reliability of the in-phase time information, the accuracy being based on the estimated model design information;
• and
• producing, using the input time series signal, the calculated in-phase time information and the calculated reliability of the in-phase time information and the estimated model design information, a prediction model for predicting the value of the time series signal at the second time point and predicting and outputting the value of the time series signal at the second time point using the prediction model, wherein
• said method includes predicting a position of the affected area at the second time point and estimating accuracy information of the prediction position using, as the input time series signal, a time series signal regarding the position of the affected area measured by said measure.

A non-transitory computer-readable recording medium storing a signal processing program for causing, in a radio radiotherapy apparatus for irradiating a radiation on an affected area of a patient comprising a radiation generator for generating the radiation, a measure for measuring a position of the affected area, and a signal processing apparatus for predicting a value of a time series signal at a second time point after a first time point, a computer to implement a prediction function that is stored, the signal processing program causing the computer to implement:

• a cycle fluctuation analysis function for assuming an input time series signal, which represents values at respective time points until the first time point, as the time series signal whose cycle length, which is a length of a cycle, varies with respect to time to carry out a correlation analysis between part of the input time series signal and the input time series signal, calculating multiple cycle lengths, which are respectively at different time points, as a cycle length fluctuation of the input time series signal and reliability of the cycle length fluctuation using a result of the correlation analysis, predicting a cycle length of the time series signal until the second time point using the cycle length fluctuation and the reliability of the cycle length fluctuation and calculating, using the result of the correlation analysis and the predicted cycle length, in-phase time information, which represents a time point before the second time point in the input time series signal, and reliability of the in-phase time information, a phase at the time point in the input time series signal becoming same as a phase at the second time point in the time series signal;
• a prediction accuracy estimation function for estimating model design information for predicting a value of the time series signal at the second time point and accuracy of the prediction of the value of the time series signal at the second time point using the input time series signal, the time series information of a variation per unit time of the value and the in-phase time information and the reliability of the in-phase time information calculated by said cycle fluctuation analysis function, the accuracy being based on the estimated model design information;
• and
• a model production function for producing, using the input time series signal, the in-phase time information and the reliability of the in-phase time information calculated by said cycle fluctuation analysis function and the model design information estimated by said prediction accuracy estimation function, a prediction model for predicting the value of the time series signal at the second time point and predicting and outputting the value of the time series signal at the second time point using the prediction model, wherein
• said signal processing apparatus is adapted to predict a position of the affected area at the second time point and estimate accuracy information of the prediction position using, as the input time series signal, a time series signal regarding the position of the affected area measured by said measure.