skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: A model for predicting lung cancer response to therapy

Abstract

Purpose: Volumetric computed tomography (CT) images acquired by image-guided radiation therapy (IGRT) systems can be used to measure tumor response over the course of treatment. Predictive adaptive therapy is a novel treatment technique that uses volumetric IGRT data to actively predict the future tumor response to therapy during the first few weeks of IGRT treatment. The goal of this study was to develop and test a model for predicting lung tumor response during IGRT treatment using serial megavoltage CT (MVCT). Methods and Materials: Tumor responses were measured for 20 lung cancer lesions in 17 patients that were imaged and treated with helical tomotherapy with doses ranging from 2.0 to 2.5 Gy per fraction. Five patients were treated with concurrent chemotherapy, and 1 patient was treated with neoadjuvant chemotherapy. Tumor response to treatment was retrospectively measured by contouring 480 serial MVCT images acquired before treatment. A nonparametric, memory-based locally weight regression (LWR) model was developed for predicting tumor response using the retrospective tumor response data. This model predicts future tumor volumes and the associated confidence intervals based on limited observations during the first 2 weeks of treatment. The predictive accuracy of the model was tested using a leave-one-out cross-validation technique withmore » the measured tumor responses. Results: The predictive algorithm was used to compare predicted verse-measured tumor volume response for all 20 lesions. The average error for the predictions of the final tumor volume was 12%, with the true volumes always bounded by the 95% confidence interval. The greatest model uncertainty occurred near the middle of the course of treatment, in which the tumor response relationships were more complex, the model has less information, and the predictors were more varied. The optimal days for measuring the tumor response on the MVCT images were on elapsed Days 1, 2, 5, 9, 11, 12, 17, and 18 during treatment. Conclusions: The LWR model accurately predicted final tumor volume for all 20 lung cancer lesions. These predictions were made using only 8 days' worth of observations from early in the treatment. Because the predictions are accurate with quantified uncertainty, they could eventually be used to optimize treatment.« less

Authors:
 [1];  [2];  [3];  [4];  [4];  [4];  [2]
  1. Department of Radiation Oncology, Thompson Cancer Survival Center, Knoxville, TN (United States) and Department of Nuclear Engineering, University of Tennessee, Knoxville, TN (United States)
  2. Department of Radiation Oncology, Thompson Cancer Survival Center, Knoxville, TN (United States)
  3. Department of Nuclear Engineering, University of Tennessee, Knoxville, TN (United States)
  4. Department of Radiation Oncology, M. D. Anderson Cancer Center Orlando, Orlando, FL (United States)
Publication Date:
OSTI Identifier:
20944708
Resource Type:
Journal Article
Journal Name:
International Journal of Radiation Oncology, Biology and Physics
Additional Journal Information:
Journal Volume: 67; Journal Issue: 2; Other Information: DOI: 10.1016/j.ijrobp.2006.09.051; PII: S0360-3016(06)03237-8; Copyright (c) 2007 Elsevier Science B.V., Amsterdam, Netherlands, All rights reserved; Country of input: International Atomic Energy Agency (IAEA); Journal ID: ISSN 0360-3016
Country of Publication:
United States
Language:
English
Subject:
62 RADIOLOGY AND NUCLEAR MEDICINE; ACCURACY; ALGORITHMS; CARCINOMAS; CHEMOTHERAPY; COMPUTERIZED TOMOGRAPHY; FORECASTING; IMAGES; LUNGS; PATIENTS; RADIATION DOSES; RADIOTHERAPY; VALIDATION

Citation Formats

Seibert, Rebecca M, Ramsey, Chester R, Department of Nuclear Engineering, University of Tennessee, Knoxville, TN, Hines, J Wesley, Kupelian, Patrick A, Langen, Katja M, Meeks, Sanford L, and Scaperoth, Daniel D. A model for predicting lung cancer response to therapy. United States: N. p., 2007. Web. doi:10.1016/j.ijrobp.2006.09.051.
Seibert, Rebecca M, Ramsey, Chester R, Department of Nuclear Engineering, University of Tennessee, Knoxville, TN, Hines, J Wesley, Kupelian, Patrick A, Langen, Katja M, Meeks, Sanford L, & Scaperoth, Daniel D. A model for predicting lung cancer response to therapy. United States. https://doi.org/10.1016/j.ijrobp.2006.09.051
Seibert, Rebecca M, Ramsey, Chester R, Department of Nuclear Engineering, University of Tennessee, Knoxville, TN, Hines, J Wesley, Kupelian, Patrick A, Langen, Katja M, Meeks, Sanford L, and Scaperoth, Daniel D. 2007. "A model for predicting lung cancer response to therapy". United States. https://doi.org/10.1016/j.ijrobp.2006.09.051.
@article{osti_20944708,
title = {A model for predicting lung cancer response to therapy},
author = {Seibert, Rebecca M and Ramsey, Chester R and Department of Nuclear Engineering, University of Tennessee, Knoxville, TN and Hines, J Wesley and Kupelian, Patrick A and Langen, Katja M and Meeks, Sanford L and Scaperoth, Daniel D},
abstractNote = {Purpose: Volumetric computed tomography (CT) images acquired by image-guided radiation therapy (IGRT) systems can be used to measure tumor response over the course of treatment. Predictive adaptive therapy is a novel treatment technique that uses volumetric IGRT data to actively predict the future tumor response to therapy during the first few weeks of IGRT treatment. The goal of this study was to develop and test a model for predicting lung tumor response during IGRT treatment using serial megavoltage CT (MVCT). Methods and Materials: Tumor responses were measured for 20 lung cancer lesions in 17 patients that were imaged and treated with helical tomotherapy with doses ranging from 2.0 to 2.5 Gy per fraction. Five patients were treated with concurrent chemotherapy, and 1 patient was treated with neoadjuvant chemotherapy. Tumor response to treatment was retrospectively measured by contouring 480 serial MVCT images acquired before treatment. A nonparametric, memory-based locally weight regression (LWR) model was developed for predicting tumor response using the retrospective tumor response data. This model predicts future tumor volumes and the associated confidence intervals based on limited observations during the first 2 weeks of treatment. The predictive accuracy of the model was tested using a leave-one-out cross-validation technique with the measured tumor responses. Results: The predictive algorithm was used to compare predicted verse-measured tumor volume response for all 20 lesions. The average error for the predictions of the final tumor volume was 12%, with the true volumes always bounded by the 95% confidence interval. The greatest model uncertainty occurred near the middle of the course of treatment, in which the tumor response relationships were more complex, the model has less information, and the predictors were more varied. The optimal days for measuring the tumor response on the MVCT images were on elapsed Days 1, 2, 5, 9, 11, 12, 17, and 18 during treatment. Conclusions: The LWR model accurately predicted final tumor volume for all 20 lung cancer lesions. These predictions were made using only 8 days' worth of observations from early in the treatment. Because the predictions are accurate with quantified uncertainty, they could eventually be used to optimize treatment.},
doi = {10.1016/j.ijrobp.2006.09.051},
url = {https://www.osti.gov/biblio/20944708}, journal = {International Journal of Radiation Oncology, Biology and Physics},
issn = {0360-3016},
number = 2,
volume = 67,
place = {United States},
year = {Thu Feb 01 00:00:00 EST 2007},
month = {Thu Feb 01 00:00:00 EST 2007}
}