willaime et.al. Texture features and proliferation
by Mikalai Budzevich
1. GLSZM grey-level size-zone matrix, 13 total
1.1. Cupture information about the size and intensities of the clusters of voxels
2. NGTDM, 4 total
2.1. local intensities differences between each voxel and its 26 neighbors
3. SUV, mean and max over 60 minutes
4. FOS, first order statistics
5. Analysis
5.1. Repeatability: ability of the features to give consistent results for the same subject scanned twice without an intervening treatment
5.1.1. Only features with high repeatability, ICC >=0.7
5.1.2. 5 cases , Shapiro-Wilk test
5.1.3. Very High repeatability
5.1.3.1. Coefficient of Variability
5.1.3.2. InfCo2
5.1.3.3. Contrast from GLCM
5.1.3.4. dissimilarity
5.1.3.5. complexity
6. Response after 60 days
7. FLT-PET
7.1. uptake is correlated with cell proliferation Ki-67 labeling index
8. 28 descriptors
9. heterogeneity > tumor growth and treatment failure
10. Initially 15 Patients
10.1. histologically proven breast cancer
10.2. at least one lesion with minimum diameter 2.5 cm
10.3. was staged at II and IV according AJCC
10.4. had not received any treatment
10.5. Only 12 were in analysis
11. PET Scan
11.1. ECAT 962/HR+ Siemens
11.2. Time = 95 min
11.3. Dynamic PET, planed dose 370 MBq
11.4. 31 discrete frames
11.5. OSEM reconstruction 360 iter, 6 subsets
11.6. FWHM 5.0 mm
11.7. voxel reconstructed size 2.651x2.651x2.420
11.8. ROI and VOI - manual
12. GLCM
12.1. computed on dynamic PET images averaged over 32-45 min
12.2. GCLM - 13 orientations 2 directions , distance - one , 26 neighbors. Texture from all GLCM were averaged over 13 directions