Premium
TH‐AB‐303‐02: An Investigation of Respiratory Signal Parameters for Multiple‐Step Ahead Prediction of Surrogate Motion
Author(s) -
Zawisza I,
Ren L,
Yin F
Publication year - 2015
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4926157
Subject(s) - weighting , mean squared error , computer science , principal component analysis , algorithm , artificial intelligence , correlation , tracking (education) , signal (programming language) , pattern recognition (psychology) , mathematics , statistics , medicine , psychology , pedagogy , geometry , radiology , programming language
Purpose: Target tracking or gating the radiation beam using real‐time imaging and surrogate motion monitoring methods are employed for dynamic tracking treatment or respiratory‐gated radiotherapy. This method requires tumor motion prediction far enough in advance. This study investigated the effect of various respiratory motion parameters on the prediction accuracy of tumor motion in a newly developed prediction algorithm. Methods: The algorithm takes a one‐dimensional surrogate signal of amplitude versus time, which is further divided into three components: training, input, and analysis components used as a validation against the prediction. The prediction algorithm consists of three major steps: (1)extracting top‐ranked subcomponents from training component which best‐match the input component; (2)calculating weighting factors from these best‐matched subcomponents; (3)fusing the proceeding optimal subcomponents with assigned weighting factors to form prediction. The prediction algorithm was examined using respiratory signals obtained from 30 simulations for prediction algorithm parameter optimization, and 555 phantom and patient data from the respiratory positioning management device. The analysis and input components were calculated for both a full and half respiratory cycle. Performance is assessed on correlation and root mean square error (RMSE) between prediction and analysis component. Results: Average correlation between prediction and analysis component was 0.720±0.390, 0.727±0.383, 0.535±0.454, 0.725±0.397, and 0.725±0.398 for full respiratory cycle prediction and 0.789±0.398, 0.800±0.385, 0.426±0.562, 0.784±0.389, and 0.784±0.389 for half respiratory cycle prediction for equal, relative, pattern, derivative equal and derivative relative weighting methods, respectively. Wilcoxon signed‐rank test (p‐test) between the full and half respiratory cycle correlations for each algorithm Result in statistically highly significant results (p<0.1%), in favor of a half respiratory cycle prediction in all cases except for Pattern Method. Conclusion: The prediction algorithms are effective in estimating surrogate motion multiple‐steps in advance. Statistical analysis indicates an advantage in using a half cycle prediction. Relative weighting method shows the best prediction accuracy.