Study population
Fig. 1 is a flowchart that summarizes the inclusion process in this study.
[6-10].
The final cohort included 23 patients with 28 HCCs and 25 patients with 28 AP shunts (5 – 20mm).
Five patients with HCCs and 3 patients with AP shunts had two lesions each.
CT technique
All patients underwent quadriphasic liver CT,
including precontrast phase,
arterial phase,
portal venous phase,
delayed phase images.
Image analysis
1.
Qualitative and quantitative analysis
1) Morphologic features [1]
- Size: the longest dimension on axial images
- Location: ‘subcapsular’ or ‘non-subcapsular’
- Shape: ‘Irregular’ or ‘round or oval’
- Margin: ‘well-defined’ or ‘ill-defined’
- Coexistent HCC: present or absent
2) Visual assessment for attenuation of lesions
- Classified as hypo,
iso,
or hyper compared with the liver parenchyma
3) Quantitative analysis
- The attenuation of each lesion and liver parenchyma was measured using ROIs and the lesion-to-liver attenuation ratios were calculated,
comparing the attenuation of lesion divided by that of liver parenchyma
2.
Evaluation of the added value of precontrast phase for differentiating small HCC from AP shunt
Two gastrointestinal radiologists independently reviewed the CT images; two reading sessions for two image sets (triphasic and quadriphasic CT images) with 4-weeks interval.
Imaging features that favors HCC or AP shunt are listed in Table 1 [1,
3].
Five-point confidence scale was used as follows: 1,
definitely HCC; 2,
probably HCC; 3,
indeterminate; 4,
probably AP shunt; 5,
definitely AP shunt.
Statistical analysis
Differences in the imaging features between HCC and AP shunt
- Univariate analysis (chi-square test,
Fisher’s exact test,
Student’s t-test,
and Mann-Whitney test).
- Multivariate logistic regression analysis.
Diagnostic predictive values
- McNemar test
Analysis of underdiagnosis and overdiagnosis
- Overdiagnosis: confidence levels of ≤3 that were confirmed as AP shunt
- Underdiagnosis: confidence levels of ≥4 that were confirmed as HCC
Interobserver agreement: kappa (κ) statistics.
A p value less than 0.05 was considered significant.