Table 1 Studies that evaluated the impact of variation in imaging parameters on HCC texture quantification

Perrin et al. [9]

CECT

254 features: GLCM,

GLRLM,

LBP,

ACM,

IH,

FD

Contrast Injection rate (CIR)

Change in CIR 0.15 ml/s (range 0–2.5)

68/254features reproducible when variation CIR < 15%

50/254 features reproducible with variations of 50%

Quantification of features reduced as variability in CIR increases.

Pixel resolution

Pixel resolution difference 7.27% (range 0–30.8%)

34/254 features reproducible with 15% variation in resolution.

>  60 features reproducible with resolution variation < 5%

Quantification of features reduced as variability in pixel resolution increases

Scanner model

75/254 features reproducible with same scanner and 35/254 with different scanner

Quantification of features reduced when > 1 scanner is used

Solomon et al. [10]

CECT

23 GLCM-features:

Contrast, correlation,

energy, homogeneity, entropy

Reconstruction algorithms:

Different reconstruction algorithm

Contrast:

32% lower with MBIR than with FBP

MBIR and ASIR significantly improved the quantification of texture features.

MBIR, FBP and ASIR

Correlation: 37% higher with MBIR than FBP

Energy: not significantly affected by algorithm

Radiation dose had no significant effect on texture features

Radiation dose

Homogeneity: 15% higher with MBIR than FBP

Entropy: unaffected

No significant impact on texture features

Mayerhoefer et al. [15]

3 T MRI

GLCM, GLRLM, IH, ARM, WAV based features

NA, TR, TE, SBW and pixel resolution

NA, TR, TE, SBW and pixel resolution at different values

Clinical resolution (MTX = 32 X 32; pixel size = 0.88 mm²):

GLCM and GLRLM more sensitive to changes in NA, TR, SBW, TE than IH, ARM and WAV.

Lower resolution: Sensitivity of all features to NA, TR, TE and SBW reduced

GLCM derived features were most robust to variations