Pancreas adenocarcinoma CT texture analysis: comparison of 3D and 2D tumor segmentation techniques.

PURPOSE: To determine equivalency of multi-slice 3D CTTA and single slice 2D CTTA of pancreas adenocarcinoma. METHODS: This retrospective study was research ethics board approved. Untreated pancreas adenocarcinomas were segmented on CT in 128 consecutive patients. Tumor segmentation was compared using two techniques: 3D segmentation by contouring all visible tumor in a 3D volume, and 2D segmentation using only a single axial image. First-order CTTA features including mean, minimum, maximum Hounsfield units (HU), standard deviation, skewness, kurtosis, entropy, and second-order gray-level co-occurrence matrix (GLCM) features homogeneity, contrast, correlation, entropy and dissimilarity were extracted. Median values were compared using the Mann-Whitney U test with Holm-Bonferroni correction. Kendall's Rank Correlation Tau assessed for correlation, and agreement was calculated using intraclass correlation coefficients (ICC) using a two-way model with single rating and absolute agreement. Statistical significance defined as P < 0.05. RESULTS: The median values of CTTA features differed significantly between 3 and 2D segmentations for all of the evaluated features except for mean attenuation, standard deviation and skewness (P = 0.2979 each). 3D and 2D segmentations had moderate correlation for mean attenuation (R = 0.69, P < 0.01), while all other features demonstrated poor to fair correlation. Agreement between 3 and 2D segmentations was good for mean attenuation (ICC: 0.87, P < 0.01), moderate for minimum (ICC: 0.65, P < 0.01) and standard deviation (ICC: 0.56, P < 0.01), and poor for all other features. CONCLUSION: While pancreas adenocarcinoma CTTA features obtained using 3D and 2D segmentation have multiple associations with clinically relevant outcomes, these segmentation techniques are likely not interchangeable other than for mean HU.

Diagnostic Accuracy of Single-Phase Computed Tomography Texture Analysis for Prediction of LI-RADS v2018 Category.

OBJECTIVE: The aim of this study was to determine if texture analysis can classify liver observations likely to be hepatocellular carcinoma based on the Liver Imaging Reporting and Data System (LI-RADS) using single portal venous phase computed tomography. METHODS: This research ethics board-approved retrospective cohort study included 64 consecutive LI-RADS observations. Individual observation texture analysis features were compared using Kruskal-Wallis and 2 sample t tests. Logistic regression was used for prediction of LI-RADS group. Diagnostic accuracy was assessed using receiver operating characteristic curves and Youden method. RESULTS: Multiple texture features were associated with LI-RADS including the mean HU (P = 0.003), median (P = 0.002), minimum (P = 0.010), maximum (P = 0.013), standard deviation (P = 0.009), skewness (P = 0.007), and entropy (P < 0.001). On logistic regression, LI-RADS group could be predicted with area under the curve, sensitivity, and specificity of 0.98, 96%, and 100%, respectively. CONCLUSIONS: Texture analysis features on portal venous phase computed tomography can identify liver observations likely to be hepatocellular carcinoma, which may preclude the need to recall some patients for additional multiphase imaging.

Hypovascular pancreas head adenocarcinoma: CT texture analysis for assessment of resection margin status and high-risk features.

OBJECTIVES: To determine if CT texture analysis features are associated with hypovascular pancreas head adenocarcinoma (PHA) postoperative margin status, nodal status, grade, lymphovascular invasion (LVI), and perineural invasion (PNI). METHODS: This Research Ethics Board-approved retrospective cohort study included 131 consecutive patients with resected PHA. Tumors were segmented on preoperative contrast-enhanced CT. Tumor diameter and texture analysis features including mean, minimum and maximum Hounsfield units, standard deviation, skewness, kurtosis, and entropy and gray-level co-occurrence matrix (GLCM) features correlation and dissimilarity were extracted. Two-sample t test and logistic regression were used to compare parameters for prediction of margin status, nodal status, grade, LVI, and PNI. Diagnostic accuracy was assessed using receiver operating characteristic curves and Youden method was used to establish cutpoints. RESULTS: Margin status was associated with GLCM correlation (p = 0.012) and dissimilarity (p = 0.003); nodal status was associated with standard deviation (p = 0.026) and entropy (p = 0.031); grade was associated with kurtosis (p = 0.031); LVI was associated with standard deviation (p = 0.047), entropy (p = 0.026), and GLCM correlation (p = 0.033) and dissimilarity (p = 0.011). No associations were found for PNI (p > 0.05). Logistic regression yielded an area under the curve of 0.70 for nodal disease, 0.70 for LVI, 0.68 for grade, and 0.65 for margin status. Optimal sensitivity/specificity was as follows: nodal disease 73%/72%, LVI 72%/65%, grade 55%/83%, and margin status 63%/66%. CONCLUSIONS: CT texture analysis features demonstrate fair diagnostic accuracy for assessment of hypovascular PHA nodal disease, LVI, grade, and postoperative margin status. Additional research is rapidly needed to identify these high-risk features with better accuracy. KEY POINTS: • CT texture analysis features are associated with pancreas head adenocarcinoma postoperative margin status which may help inform treatment decisions as a negative resection margin is required for cure. • CT texture analysis features are associated with pancreas head adenocarcinoma nodal disease, a poor prognostic feature. • Indicators of more aggressive pancreas head adenocarcinoma biology including tumor grade and LVI can be diagnosed using CT texture analysis with fair accuracy.