PURPOSE: To describe a novel method of analyzing partial volume effects of normal tissues to radiation. With this approach, principal component analysis (PCA) is used to efficiently describe the variance in cumulative dose-volume histogram (cDVH) morphology. The independent features of cDVHs that describe the largest variance are then investigated regarding complication risk. METHODS AND MATERIALS: Principal component analysis was used to describe the variance in the morphology of normal tissue cDVHs, irrespective of complication, by summarizing the largest source of variation within the first principal component (PC), the next largest in the second PC, and so on. Plots relating the most meaningful PCs were constructed. Ideally, cDVHs associated with a complication would yield PC values that could be easily segregated from cDVHs without a complication. Two data sets were evaluated with this approach: 90 parotid gland cDVHs (36 with complications) and 203 liver cDVHs (19 with complications). RESULTS: Ninety-four percent and 80% of the variation in cDVH morphology was described with two PCs for the parotid gland and the liver data sets, respectively. Plots of the first and second PC values on a Cartesian plane for both data sets revealed "clusters." For the parotid gland, one cluster contained PCs from parotid gland cDVHs with complications, and the other primarily contained PCs from cDVHs without complications. The first PC value, corresponding to a larger volume treated with 10-60 Gy (2 Gy per fraction), was more likely to be larger in parotid gland cDVHs associated with complications than those without complications. In the plots of PC values of liver cDVHs, whole liver radiation cDVHs were segregated from the other cDVHs. There was a trend for cDVHs with a higher first PC, corresponding to increased volume treated with approximately 10-40 Gy (1.5 Gy b.i.d.), to be associated with increased risk of complication. For partial liver radiation cDVHs there was a trend for cDVHs with a higher first PC, corresponding to an increased volume treated with 5-50 Gy, to be associated with a complication. For each data set, logistic regression modeling revealed that the first PC was significantly associated with a complication developing (p < 0.02). CONCLUSIONS: Principal component analysis can be used to summarize the variance in parallel normal tissue cDVHs, and it can help segregate cDVHs at high or low risk for complications.
PURPOSE: To describe a novel method of analyzing partial volume effects of normal tissues to radiation. With this approach, principal component analysis (PCA) is used to efficiently describe the variance in cumulative dose-volume histogram (cDVH) morphology. The independent features of cDVHs that describe the largest variance are then investigated regarding complication risk. METHODS AND MATERIALS: Principal component analysis was used to describe the variance in the morphology of normal tissue cDVHs, irrespective of complication, by summarizing the largest source of variation within the first principal component (PC), the next largest in the second PC, and so on. Plots relating the most meaningful PCs were constructed. Ideally, cDVHs associated with a complication would yield PC values that could be easily segregated from cDVHs without a complication. Two data sets were evaluated with this approach: 90 parotid gland cDVHs (36 with complications) and 203 liver cDVHs (19 with complications). RESULTS: Ninety-four percent and 80% of the variation in cDVH morphology was described with two PCs for the parotid gland and the liver data sets, respectively. Plots of the first and second PC values on a Cartesian plane for both data sets revealed "clusters." For the parotid gland, one cluster contained PCs from parotid gland cDVHs with complications, and the other primarily contained PCs from cDVHs without complications. The first PC value, corresponding to a larger volume treated with 10-60 Gy (2 Gy per fraction), was more likely to be larger in parotid gland cDVHs associated with complications than those without complications. In the plots of PC values of liver cDVHs, whole liver radiationcDVHs were segregated from the other cDVHs. There was a trend for cDVHs with a higher first PC, corresponding to increased volume treated with approximately 10-40 Gy (1.5 Gy b.i.d.), to be associated with increased risk of complication. For partial liver radiationcDVHs there was a trend for cDVHs with a higher first PC, corresponding to an increased volume treated with 5-50 Gy, to be associated with a complication. For each data set, logistic regression modeling revealed that the first PC was significantly associated with a complication developing (p < 0.02). CONCLUSIONS: Principal component analysis can be used to summarize the variance in parallel normal tissue cDVHs, and it can help segregate cDVHs at high or low risk for complications.
Authors: Søren M Bentzen; Louis S Constine; Joseph O Deasy; Avi Eisbruch; Andrew Jackson; Lawrence B Marks; Randall K Ten Haken; Ellen D Yorke Journal: Int J Radiat Oncol Biol Phys Date: 2010-03-01 Impact factor: 7.038
Authors: Charlie C Pan; Brian D Kavanagh; Laura A Dawson; X Allen Li; Shiva K Das; Moyed Miften; Randall K Ten Haken Journal: Int J Radiat Oncol Biol Phys Date: 2010-03-01 Impact factor: 7.038
Authors: Yun Liang; Karen Messer; Brent S Rose; John H Lewis; Steve B Jiang; Catheryn M Yashar; Arno J Mundt; Loren K Mell Journal: Int J Radiat Oncol Biol Phys Date: 2010-05-14 Impact factor: 7.038
Authors: Issam El Naqa; Joseph O Deasy; Yi Mu; Ellen Huang; Andrew J Hope; Patricia E Lindsay; Aditya Apte; James Alaly; Jeffrey D Bradley Journal: Acta Oncol Date: 2010-03-02 Impact factor: 4.089
Authors: Issam El Naqa; Jeffrey D Bradley; Patricia E Lindsay; Andrew J Hope; Joseph O Deasy Journal: Phys Med Biol Date: 2009-08-18 Impact factor: 3.609