












































Microsoft Word - Mikell


International Journal of Cancer Therapy and Oncology 
www.ijcto.org 

Copyright © Mikell et al.                                                                                      ISSN 2330-4049 

Characterization of tumor dose heterogeneity for 90Y microsphere 

therapies using voxel‐ based dosimetry   

Justin Mikell1, Firas Mourtada2, Armeen Mahvash1, S Cheenu Kappadath1 
 

1The University of Texas M. D. Anderson Cancer Center, Houston, TX, USA. 
2Christiana Care, Newark, DE, USA.  

 
Received March 19, 2014; Published Online April 08, 2014 

[Presented at the Young Investigator’s Symposium at the 2014 Annual Meeting of  
Southwest Chapter of American Association of Physicists in Medicine (AAPM) in San Antonio, Texas, USA] 

Conference Proceeding 

Abstract 

Purpose: Dosimetry for 90Y microsphere therapies (YMT) 

with Standard (SM) and Partition (PM) models provide only 

uniform dose estimates to tumor and liver. Our objective is 

to calculate tumor dose heterogeneity, known to effect re-

sponse, using voxel-based dosimetry and investigate the lim-

itations of SM and PM. 

 

Methods: Voxel-based dosimetry was performed on 17 YMT 

patients using Monte Carlo DOSXYZnrc. 90Y activity and 

tissue/density distributions were based on quantitative 90Y 

bremsstrahlung SPECT/CT. Tumors (n=31), liver, and treat-

ment lobe/segments were segmented on diagnostic CT or 

MR. Dose volume histograms (DVH) were created for tu-

mors and normal liver. Bland-Altman analysis compared 

voxel-based mean absorbed doses to tumor and liver with 

SM and PM. Tumor and normal liver absorbed dose hetero-

geneity were investigated through metrics: integral uni-

formity (IU), D10/D90, COV. Correlations of heterogeneity 

with voxel-based mean doses and volumes were evaluated.  

 

Results: Heterogeneity metrics (mean ± 1σ) for tumor dose 

were COV = 0.48 ± 0.28, D10/D90 = 4.7 ± 3.9, and IU = 0.8 ± 

0.18. Heterogeneity metrics correlated with tumor volume (r 

> 0.58) but not tumor mean doses (r < 0.20). Voxel-based 

tumor mean doses correlated with PM (r = 0.84) but not SM 

(r = 0.08). Both yielded poor limits of agreement with of 83 ± 

174 and -28 ± 181 Gy, respectively. Normal liver heterogene-

ity metrics (mean ± 1σ) were COV = 0.83 ± 0.29, D10/D90 =  

12 ± 15, and IU = 0.97 ± 0.03. Only D10/D90 (r = 0.49) corre-

lated with mean normal liver absorbed dose. Voxel-based 

normal liver/lobe mean doses correlated with PM (r = 0.96), 

but had poor limits of agreement (26 ± 29 Gy). 

 

Conclusion: Tumor doses have high levels of heterogeneity 

that increase with volume but are independent of dose. 

Voxel-based DVH and dose heterogeneity metrics will pro-

mote accurate characterization of tumor response following 

YMT. 
[Research Support: NIH/NCI R01 CA138986] 
-------------------------------------------------------- 

Key Results:  

 
FIG. 1: This figure shows the tumor dose heterogeneity (max 

to min range of individual absorbed dose tumor estimates in 

beige shadow), together with the mean voxel-based tumor 

absorbed dose (red) and those for the PM (blue) and SM 

(green).  

High levels of heterogeneity are observed. The PM and SM are 

unable to encompass the heterogeneity which is represented 

as the shaded area for each tumor. The PM tracks the mean 

voxel-based dose better than the SM. Unfortunately, the PM 

overestimates for 28/31 with a large mean difference of 83 ± 

174 Gy. 

 

Presenting author: Justin Mikell; The University of Texas M.D. 

Anderson Cancer Center, Houston, TX, USA. 

 

Cite this article as: 

Mikell J, Mourtada F, Mahvash A, Kappadath SC. Characterization 

of tumor dose heterogeneity for 90Y microsphere therapies using 

voxel- based dosimetry. Int J Cancer Ther Oncol 2014; 
2(2):020228. DOI: 10.14319/ijcto.0202.28 

 

http://ijcto.org/index.php/IJCTO/index
http://dx.doi.org/10.14319/ijcto.0202.28


 2  Mikell et al. : Characterization of tumor dose heterogeneity                       International Journal of Cancer Therapy and Oncology 
                      www.ijcto.org 

Copyright © Mikell et al.                                                                                      ISSN 2330-4049 

 

 
FIG. 2: A plot of mean tumor voxel-based dose versus those from PM and SM showing that the PM is a better surrogate for true mean doses 

than the SM. No correlation between voxel-based and SM was found. Voxel-based and PM were correlated, but notice that the PM overesti-

mates the true mean dose by a factor of ~2.

 
TABLE 1: Correlations of heterogeneity metrics with voxel-based mean absorbed doses. Checkered cells indicate that no correlation was found. 

D10 and D90 are derived from cumulative dose volume histograms. D90 is the minimum dose in the hottest 90% of tumor volume. Similarly, 

D10 is the minimum dose in the hottest 10% of the tumor volume. COV is the coefficient of variation (= standard deviation/mean). 

non- 

uniformity  

metric 

Linear fits of heterogeneity metrics (correlation coefficient) 

versus voxel-based mean dose (Gy) versus volume (cc) 

tumor (N=31) 

y = mx + b (corr coef) 

normal liver (N=17) 

y = mx + b (corr coef) 

tumor (N=31) 

y = mx + b (corr coef) 

normal liver (N=17) 

y = mx + b (corr coef) 

COV   7.0E-4x + 0.34 (0.68)  

max/min 2.0E-1x + 15 (0.32) 

D10/D90 4.1E-1x – 3.6 (0.49) 9.4E-3x + 2.7 (0.64) -9.6E-3x + 25 (0.44) 

IU  4.0E-4x + 0.71 (0.59)  

 
 

TABLE 2: Bland Altman analyses of dosimetry models with voxel-based mean absorbed dose as truth. Biases calculated as PM or SM minus 

voxel-based mean absorbed dose. Both SM and PM overestimate the normal liver dose compared to voxel-based means. 

 Bland Altman analysis with voxel-based mean dose as 

truth 

tumor (N=31) 

bias ± 2σ (Gy) 

normal liver (N=17) 

bias ± 2σ (Gy) 

Standard Model -28.1 ± 180.7 40.1 ± 26.8 

Partition Model 83.4 ± 174.4 25.7 ± 29.3 

 

 

 

 


