Statistical modeling of CTV motion and deformation for IMRT of early-stage rectal cancer

Int J Radiat Oncol Biol Phys. 2014 Nov 1;90(3):664-72. doi: 10.1016/j.ijrobp.2014.06.040. Epub 2014 Aug 20.

Abstract

Purpose: To derive and validate a statistical model of motion and deformation for the clinical target volume (CTV) of early-stage rectal cancer patients.

Methods and materials: For 16 patients, 4 to 5 magnetic resonance images (MRI) were acquired before each fraction was administered. The CTV was delineated on each MRI. Using a leave-one-out methodology, we constructed a population-based principal component analysis (PCA) model of the CTV motion and deformation of 15 patients, and we tested the model on the left-out patient. The modeling error was calculated as the amount of the CTV motion-deformation of the left-out-patient that could not be explained by the PCA model. Next, the PCA model was used to construct a PCA target volume (PCA-TV) by accumulating motion-deformations simulated by the model. A PCA planning target volume (PTV) was generated by expanding the PCA-TV by uniform margins. The PCA-PTV was compared with uniform and nonuniform CTV-to-PTV margins. To allow comparison, geometric margins were determined to ensure adequate coverage, and the volume difference between the PTV and the daily CTV (CTV-to-PTV volume) was calculated.

Results: The modeling error ranged from 0.9 ± 0.5 to 2.9 ± 2.1 mm, corresponding to a reduction of the CTV motion-deformation between 6% and 60% (average, 23% ± 11%). The reduction correlated with the magnitude of the CTV motion-deformation (P<.001, R=0.66). The PCA-TV and the CTV required 2-mm and 7-mm uniform margins, respectively. The nonuniform CTV-to-PTV margins were 4 mm in the left, right, inferior, superior, and posterior directions and 8 mm in the anterior direction. Compared to uniform and nonuniform CTV-to-PTV margins, the PCA-based PTV significantly decreased (P<.001) the average CTV-to-PTV volume by 128 ± 20 mL (49% ± 4%) and by 35 ± 6 mL (20% ± 3.5%), respectively.

Conclusions: The CTV motion-deformation of a new patient can be explained by a population-based PCA model. A PCA model-generated PTV significantly improved sparing of organs at risk compared to uniform and nonuniform CTV-to-PTV margins.

Publication types

  • Research Support, Non-U.S. Gov't
  • Validation Study

MeSH terms

  • Anatomic Landmarks / anatomy & histology
  • Female
  • Humans
  • Magnetic Resonance Imaging / methods
  • Male
  • Models, Statistical*
  • Movement
  • Neoplasm Staging
  • Organ Sparing Treatments / methods
  • Organ Sparing Treatments / statistics & numerical data
  • Organs at Risk / anatomy & histology
  • Organs at Risk / radiation effects
  • Principal Component Analysis / methods*
  • Radiation Injuries / prevention & control
  • Radiotherapy Planning, Computer-Assisted / methods
  • Radiotherapy, Intensity-Modulated / methods
  • Radiotherapy, Intensity-Modulated / statistics & numerical data*
  • Rectal Neoplasms / pathology
  • Rectal Neoplasms / radiotherapy*
  • Tumor Burden*