A Novel Technique for Predicting Which Head and Neck Patients Will Require Adaptive Therapy

Reviewer: Nathan Jones DO
Abramson Cancer Center of the University of Pennsylvania
Last Modified: September 25, 2008

Presenter: Carley Harris
Presenter's Affiliation: Thompson Cancer Survival Center, Knoxville, TN
Type of Session: Scientific



  • With the advent of IGRT, it is now feasible to actively measure geometric changes in tumor volume and normal tissues
  • Patients treated for head and neck malignancies often undergo significant changes in both tumor volume and normal anatomy as a result of tumor shrinkage and treatment-associated weight loss
  • These geometric changes can both compromise tumor dose and give excessive normal tissue dose
  • The purpose of this study is to develop an IGRT classification tool to predict which patients will require adaptation of the original treatment plan

Materials and Methods

Materials and Methods:

  • MVCT images were collected from 43 patients undergoing radiation treatment for head and neck malignancy
  • Data was collected regarding tumor response, weight loss, and anatomical changes over the course of treatment
  • The center of each vertebral body from C1-C5 was selected as an anatomical landmark, and cross-sectional measurements were taken at each level
  • Trend detection method was applied and data were classified as either normal or abnormal
  • Kernel regression was utilized to predict those patients who would most benefit from adaptive planning
  • Once the database exists, the data is used to predict which patients will require replanning after the first 8 fractions



  • 30,000 cross-sectional measurements were taken from 1450 MVCT images of 43 patients
  • All patients demonstrated a decrease in the cross-sectional volume, but only 15 demonstrated significant anatomic changes
  • Approximately 1/3 of patients (35%) experienced weight loss of >10%, and 14% of patients experienced a major reduction in tumor volume
  • The ability to immobilize the patient was noted to decrease substantially as the tumor regressed or as a patient lost weight and the aquaplast no longer conformed to the patient
  • The kernel regression predicted an abnormal class for 15 patients after 8 fractions, and 13 of them actually required replanning
  • No patients required adaptive replanning among those that were predicted to be normal by kernel regression

Author's Conclusions

Author’s Conclusions:

  • For patients undergoing irradiation for head and neck malignancy, this kernel regression model can accurately and automatically predict which patients will require replanning following the first 8 fractions of treatment

Clinical/Scientific Implications

Clinical/Scientific Implications:

  • Adaptive radiation therapy has the potential to significantly improve radiation delivery for a subset of head and neck patients
  • These data are promising and suggest an automated method of assessing early into treatment which patients will benefit from the adaptive approach
  • With this relatively small patient set, it is difficult to know the actual sensitivity of this method, and a study with a larger independent patient population would be useful to validate these findings
  • Additional areas for future study include determination of the time point in the treatment when replanning would be most beneficial, and the ultimate therapeutic implications of adaptive therapy