Pilot Study: Texture Analysis of PET Imaging Demonstrates Changes in 18F-FDG Uptake of the Brain After Prophylactic Cranial Irradiation

Article Figures & Data

TABLE 1

Texture Analysis of Whole-Brain–Segmented ¹⁸F-FDG PET Datasets

Texture feature	P
Angular second moment	0.100
Contrast	0.043
Correlation	0.992
Sum of squares: variance	0.035
Inverse difference moment	0.168
Sum average	0.963
Sum variance	0.042
Sum entropy	0.396
Entropy	0.229
Difference variance	0.233
Difference entropy	0.039
Information on measurement of correlation 1	0.239
Information on measurement of correlation 2	0.247

P values refer to probability of statistically significant difference in each texture feature, when comparing pre- and post-PCI image sets.

TABLE 2

Texture Analysis of Regionally Segmented Datasets

Texture feature	R frontal	R parietal	R temporal	L frontal	L parietal	L temporal
Angular second moment	0.292	0.212	0.173	0.449	0.176	0.335
Contrast	0.720	0.757	0.472	0.833	0.637	0.763
Correlation	0.821	0.232	0.557	0.984	0.609	0.989
Sum of squares: variance	0.651	0.630	0.483	0.850	0.384	0.740
Inverse difference moment	0.214	0.400	0.430	0.922	0.332	0.328
Sum average	0.133	0.010	0.892	0.221	0.047	0.501
Sum variance	0.618	0.416	0.557	0.884	0.351	0.760
Sum entropy	0.478	0.205	0.576	0.846	0.744	0.296
Entropy	0.322	0.272	0.197	0.451	0.145	0.254
Difference variance	0.286	0.238	0.842	0.852	0.550	0.351
Difference entropy	0.452	0.437	0.790	0.953	0.785	0.574
Information on measurement of correlation 1	0.465	0.448	0.790	0.478	0.138	0.409
Information on measurement of correlation 2	0.921	0.980	0.061	0.497	0.034	0.109

Values refer to probability of statistically significant difference in each texture feature (P value), when comparing pre- and post-PCI image sets.

TABLE 3

Relevant GLCM Texture Features, Originally Described by Haralick et al. (10)

θ) corresponds to entry in GLCM, designating probability of pixel with gray-level of i being distance d pixels away from neighboring pixel with gray-level of j in θ direction. In this study, we fixed d at 1 pixel and averaged over available directions for θ. N_g represents number of gray levels. x and y are row and column indices of GLCM, and is probability of 2 indices summing to , whereas is probability that difference 2 indices is ; μ is mean of p.