Respiratory movement predicting method

Abstract

The invention discloses a respiratory movement predicting method, which comprises the following steps: (1) inputting a status characteristic set; (2) acquiring a real-time respiratory signal f(t) anda status characteristic R(t) thereof; (3) setting up a similarity constraint condition of the status characteristic R(t) and a status characteristic R(k) through a probability likelihood model, and setting up a continuity constraint condition of adjacent status characteristics R(ti) and R(ti+1) through a probability prior model; (4) screening characteristic elements meeting the conditions in the step (3) from a candidate set through a maximum posteriori model, and predicating a respiratory signal; and (5) outputting a respiratory signal f(t+ Deltat) delayed for Deltat. The method adopts the maximum posteriori algorithm to construct the models and makes full use of local and global respiratory characteristics to set up the probability model. Compared with the conventional linear predictionmodel, the model can better predict the real respiratory movement condition, has small average prediction error and improves the respiratory movement predicting accuracy so as to improve radiotherapyeffect.

Description

technical field [0001] The invention relates to a method for predicting respiratory movement in radiotherapy, in particular to a method for tracking and predicting respiratory signals in radiotherapy. Background technique [0002] Radiation therapy for thoracic and abdominal organ tumors is largely influenced by respiratory motion. During therapy, tumor tissue changes dynamically over time. The resulting uncertainty has become the main problem faced by current radiotherapy, and the methods and means to solve it are far from perfect. The traditional radiotherapy mode based on conformal intensity-modulated technology usually adopts isocenter displacement, breathing training, gated radiation and other technologies to deal with the above situations. Addressing this situation with image-guided radiotherapy requires analysis of how the tumor responds to respiratory motion. [0003] The currently used tumor tracking algorithms mainly include: linear estimation algorithm (LP) and...

AI Technical Summary

Problems solved by technology

[0003] The currently used tumor tracking algorithms mainly include: linear estimation algorithm (LP) and linear extrapolation algorithm (LE); among them, the linear estimation algorithm (LP) algorithm calculates the tumor position X t Expressed as a linear combination of the first n+1 states, for a set of training samples X 1 , X 2 ,...X n , coefficient a k The mean square error can be minimized by solving a set of linear equations. It is not easy to choose the optimal number of historical states. Linear estimation is only suitable for systems with small delays; the linear extrapolation algorithm (LE) maintains a constant speed when assuming signal changes Under the condition of estimating the change speed of the signal by using the sample values ​​at the last two moments, the tumor movement can only be regarded as a constant in a very short time. Therefore, the linear extrapolation algorithm (LE) algorithm is only suitable for small delay system

[0004] The main problems of the above-mentioned traditional prediction and tracking technologies are: without using training samples and prior knowledge, the local linearization assumption or non-regularized mathematical modeling of the object is performed; when the shape of the object changes randomly, it cannot be effectively approximated. target, resulting in large calculation errors

Images