Authors:
B. Ganeshan1, K. A. Miles2, R. C. D. Young1, C. R. Chatwin1; 1Falmer/UK, 2Brighton/UK
Methods and Materials
Measuring Liver Texture
A feasibility study comprised 27 patients undergoing CT after primary resection of colorectal cancer followed by an evaluation group of a further 32 patients. Hepatic CT perfusion and survival data were available for both groups with fluorodeoxyglucose PET results accessible in the evaluation group.
Texture was assessed within regions of interest (ROIs) manually constructed over apparently normal liver whilst excluding major blood vessels and fat.
Texture analysis comprised: a) band-pass image filtration using Laplacian of Gaussian filter (Figure 1) to highlight image features at different spatial frequencies (Table 1 & Figure 2) between 0.5 (fine detail) and 2.5 (coarse features),
![](https://epos.myesr.org/posterimage/esr/ecr2007/11892/media/165236?maxheight=300&maxwidth=300)
Fig.
Sigma (σ) | Texture type | Filter Width (approx. no. of pixels) |
0.5 | Fine | 2 |
1.0 | 4 |
1.5 | Medium | 6 |
2.0 | 10 |
2.5 | Coarse | 12 |
Table 1. Filter sigma value and the corresponding width of the filter (pixels)
![](https://epos.myesr.org/posterimage/esr/ecr2007/11892/media/165273?maxheight=300&maxwidth=300)
Fig.
and b) assessment of the relative contributions made to the image by features at two different spatial frequencies, expressed as filter ratios (e.g. 0.5/2.5).
Texture Quantification
![](https://epos.myesr.org/posterimage/esr/ecr2007/11892/media/168823?maxheight=300&maxwidth=300)
Fig.
![](https://epos.myesr.org/posterimage/esr/ecr2007/11892/media/168828?maxheight=300&maxwidth=300)
Fig.
![](https://epos.myesr.org/posterimage/esr/ecr2007/11892/media/165594?maxheight=300&maxwidth=300)
Fig.