Premium
TU‐FF‐A4‐06: Image‐Based Modeling of Tumor Shrinkage Or Growth: Towards Adaptive Radiation Therapy of Head‐And‐Neck Cancer
Author(s) -
Chao M,
Xie Y,
Xing L
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 - 10.1118/1.2962661
Subject(s) - voxel , feature (linguistics) , image registration , artificial intelligence , computer science , scale invariant feature transform , interpolation (computer graphics) , computer vision , imaging phantom , image scaling , pattern recognition (psychology) , feature extraction , image (mathematics) , image processing , nuclear medicine , medicine , philosophy , linguistics
Purpose: Understanding the kinetics of tumor growth/shrinkage represents a critical step in quantitative assessment of therapeutics and realization of adaptive radiation therapy (ART). We establish a novel framework for image‐based modeling of tumor change and demonstrate its performance. Method and Materials: Due to the non‐conservation of tissue, similarity‐based deformable models are not suitable for describing the tumor growth/shrinkage process. Under the hypothesis that the tissue features in the tumor volume or the boundary region are partially preserved, we model the tumor kinetics by a two‐step procedure: (1) auto‐detection of homologous tissue features shared by the planning CT and subsequent on‐treatment CBCT images using the Scale Invariance Feature Transformation (SIFT) method; (2) establishment of voxel‐to‐voxel correspondence between two input images for the remaining spatial points by a basis spline interpolation. The correctness of the tissue feature correspondence is doubly assured by a bi‐directional association procedure, in which the SIFT features are mapped from planning CT to CBCT and reversely. Only the associations common to both mappings are used in BSpline interpolation. A number of synthetic digital phantom experiments and five clinical HN cases are used to assess the performance of the proposed technique. Results: Image contents of the digital phantoms are modified in various ways. It is found the proposed technique can identify any of the changes faithfully. The subsequent feature‐guided BSpline interpolation reproduces the “ground truth” with an accuracy better than 1.3mm. For the clinical cases, the new algorithm works reliably for a volume change less than 30%, suggesting the time span between two consequent imaging sessions should not be unreasonably far away in order for the model to function properly. Conclusion: An image‐based tumor kinetic model has been developed to better understand the tumor response to radiation therapy. The technique provides a solid foundation for future head‐and‐neck ART.