Premium
The effect of respiratory motion variability and tumor size on the accuracy of average intensity projection from four‐dimensional computed tomography: An investigation based on dynamic MRI
Author(s) -
Cai Jing,
Read Paul W.,
Sheng Ke
Publication year - 2008
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 - 10.1118/1.2982245
Subject(s) - imaging phantom , maximum intensity projection , magnetic resonance imaging , projection (relational algebra) , nuclear medicine , medical imaging , radiation treatment planning , intensity (physics) , sagittal plane , iterative reconstruction , tomography , medicine , radiation therapy , radiology , computer science , physics , algorithm , optics , angiography
Composite images such as average intensity projection (AIP) and maximum intensity projection (MIP) derived from four‐dimensional computed tomography (4D‐CT) images are commonly used in radiation therapy for treating lung and abdominal tumors. It has been reported that the quality of 4D‐CT images is influenced by the patient respiratory variability, which can be assessed by the standard deviation of the peak and valley of the respiratory trajectory. Subsequently, the resultant MIP underestimates the actual tumor motion extent. As a more general application, AIP comprises not only the tumor motion extent but also the probability that the tumor is present. AIP generated from 4D‐CT can also be affected by the respiratory variability. To quantitate the accuracy of AIP and develop clinically relevant parameters for determining suitability of the 4D‐CT study for AIP‐based treatment planning, real time sagittal dynamic magnetic resonance imaging (dMRI) was used as the basis for generating simulated 4D‐CT. Five‐minute MRI scans were performed on seven healthy volunteers and eight lung tumor patients. In addition, images of circular phantoms with diameter 1, 3, or 5 cm were generated by software to simulate lung tumors. Motion patterns determined by dMRI images were reproduced by the software generated phantoms. Resorted dMRI using a 4D‐CT acquisition method (RedCAM) based on phantom or patient images was reconstructed by simulating the imaging rebinning processes. AIP images and the corresponding color intensity projection (CIP) images were reconstructed from RedCAM and the full set of dMRI for comparison. AIP similarity indicated by the Dice index between RedCAM and dMRI was calculated and correlated with respiratory variability ( v ) and tumor size ( s ) . The similarity of percentile intrafractional motion target area (IMTA), defined by the area that the tumor presented for a given percentage of time, and MIP‐to‐percentile IMTA similarity as a function of percentile were also determined. As a result, AIP similarity depends on both respiratory variability and tumor sizes. The AIP similarity correlated linearly with the respiratory variability normalized by tumor sizes ( R 2 equal to 0.82 and 0.91 for the phantom study and the patient study, respectively). For both studies, MIP derived from RedCAM was close to the area that the tumor presented 90% or more of the time and missed the region where the tumor appeared less than 10% of the time. In conclusion, the accuracy of composite images such as AIP and MIP derived from 4D‐CT to define the tumor motion and position is affected by patient‐specific respiratory variability and tumor sizes. Based on our study, normalized respiratory variability appears to be a pertinent parameter to assess the suitability of a 4D‐CT image set for AIP‐based treatment planning.