RPS 1816 - Liver tumours: advanced imaging, radiomics and treatment effects

Purpose:

To develop and validate a radiomics nomogram for the pretreatment prediction of transcatheter arterial chemoembolisation (TACE) refractoriness with pre-therapeutic dynamic CT in patients with hepatocellular carcinoma without metastasis.

Methods and materials:

This study included 80 patients with HCC treated with TACE, except for any vascular involvement, from July 2016-November 2018, after exclusion. The datasets were split into a training set (80%) and a test set (20%) for feature selection and we did 10-fold cross-validation. 40 radiomic features were extracted from arterial-phase CT using LIFEx. A lasso regression model was used for radiomics signature selection and the selected signatures were validated using a Mann-Whitney U test. The radiomics nomogram was based on the multivariate logistic regression model incorporated the rad score (radiomics score), CT image factors (enhancement, homogeneous arterial, heterogeneous portal, capsule, Bilobar, and multiple), and clinical factors (age, sex, and logAFP). For optimising the radiomics nomogram, multiple backward stepwise elimination methods were investigated. We also evaluated the radiomics nomogram using ROC and a confusion matrix. All statistical analyses were performed using RStudio software.

Results:

The rad score, which consisted of GLZLM (grey-level zone length matrix), LZLGE (long-zone low grey-level emphasis), and GLZLM-GLNU (grey-level nonuniformity), was significantly associated with TACE refractioness (p<0.05) and a higher odds ratio (≥1.0). Predictors contained in the individualised prediction nomogram included rad score and Biolbar. The radiomics nomogram showed a good predicting performance: 0.91 AUC, 0.75 sensitivity, 0.91 specificity, 0.75 precision, and 0.87 accuracy.

Conclusion:

This study presents a radiomics nomogram that incorporates the rad score and CT image factor (Biolbar), which can be conveniently used to facilitate the pretreatment prediction of TACE refractoriness.

Limitations:

A retrospective study.

Ethics committee approval

Informed consent was waived due to the retrospective nature of the analyses.

Funding:

No funding was recieved for this work.

Purpose:

Recent studies have demonstrated that virtual monoenergetic images (VMIs) improve soft tissue contrast and conspicuity for hypodense liver lesions. This study investigates whether this translates into a benefit for washout assessment of arterially hyper-enhancing liver lesions in delayed-phase CT images.

Methods and materials:

Our retrospective study was approved by the institutional review board. 60 arterially hyper-enhancing lesions in 30 patients (age: 65±9yrs, M/W: 20/10) were included. All patients underwent multi-phase contrast-enhanced SDCT (unenhanced, arterial, portal-venous, and delayed) for HCC work-up. MRI, CEUS, or biopsy within 3 months served as a standard of reference to determine the lesions as hepatocellular carcinomas (HCC) or non-HCCs. VMIs (40-200keV, 10keV-increment) and conventional images (CI) were reconstructed. Regions-of-interest were placed in lesions, liver parenchyma, and muscle on delayed phase; washout (HU_liver-HU_lesion) and the lesion-to-liver ratio (LLR) were calculated. Visual analysis was performed on 40, 60, 80keV, and CI by 3 radiologists. On a 5-point Likert scale, conspicuity and confidence of washout-evaluation were assessed. Statistical assessment was performed using ANOVA adjusted for multiple comparisons and a Wilcoxon test.

Results:

The lesion size was 3.0±2.4 cm. Irrespective of reconstruction technique and keV-levels, washout and LLR were significantly higher on delayed phase images in HCC (n=50) compared to non-HCC lesions (n=10) (p<0.05). In HCC lesions, washout and LLR were significantly higher in VMIs with 40keV compared to CI and higher VMI keV-levels (p<0.05; e.g. washout in 40keV vs CI: 26.2±22.8 vs 11.8±8.5). Subjective analysis confirmed the best washout-evaluation (p<0.05) and confidence of evaluation in 40keV (p>0.05).

Conclusion:

By increasing the lesion contrast, VMIs with low keV-values from SDCT, especially using 40keV, significantly improve the qualitative and quantitative assessment of washout in hyper-enhancing liver lesions.

Limitations:

A retrospective, single-centre study.

Ethics committee approval

IRB-approval given.

Funding:

No funding was received for this work.

Purpose:

To analyse the valence of an IVIM model and ADC value for the characterisation of solid focal liver lesions (FLLs) using a volumetric approach in 4 groups: benign vs malignant, FNH vs adenoma, hypervascular benign vs hypervascular malignant, and hypervascular malignant vs hypervascular malignant.

Methods and materials:

100 patients with 104 FLLs (77 malignant-27 benign) underwent liver 1.5T MRI for routine examination sequences using IVIM diffusion-weighted imaging with 11 b-values (0-800 s/mm2). The apparent diffusion coefficient (ADC) and IVIM-derived parameters, such as pure diffusion coefficient (D), pseudo diffusion coefficient (Ds), perfusion fraction (f), and their product (fDs) were computed with a volumetric approach and compared with the groups. A ROC curve analysis was performed to assess their diagnostic value.

Results:

ADC and D were significantly lower in the malignant group than in the benign group (ADC mean: 1.38±0.36 vs 1.60±0.58; D mean 0.88±0.29 vs 1.05±0.30) × 10-3 mm2/s, and in the hypervascular malignant than hypervascular benign (ADC mean 1.38±0.33 vs 1.60±0.72; D mean: 0.90±0.21 vs 1.05±0.29) × 10-3 mm2/s. Ds did not show statistical differences. FDs were significantly lower in the hypovascular malignant than in the hypervascular malignant group (mean 9.96±1.03 vs 12.7±8.5) × 10-3 mm2/s. F was also lower in the hypovascular malignant than in the hypervascular malignant group

Conclusion:

Compared with ADC, the IVIM model values using a volumetric approach improves the characterisation of FLLs.

Limitations:

A retrospective study and a selection bias could be possible. A larger patient cohort may be useful to validate our results.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

To evaluate the intra-individual rate of occurrence and recurrence of hepatocellular carcinoma (HCC) in patients with hepatitis C virus (HCV) after sustained virologic response to direct-acting antivirals (DAA).

Methods and materials:

We retrospectively included consecutive HCV patients with sustained virologic response to DAA from 2015-2019 imaged with CT or MRI. We excluded patients lacking pre-DAA and post-DAA imaging follow-up. A combined reference standard (imaging and pathology) was used for HCC diagnosis. Two radiologists reviewed imaging studies, categorised each lesion, and assessed treatment response according to LI-RADS v2018. Intra-individual differences in HCC occurrence and recurrence were evaluated with the McNemar test. The incidence rate of HCC after DAA therapy was calculated.

Results:

The final cohort included 65 patients (mean age: 70±10 years old). Overall, pre-DAA and post-DAA mean follow-ups were 47±26 months, 25±25 months, and 24±15 months, respectively (pre-DAA vs post-DAA: p=0.472). HCC occurrence was demonstrated in a significantly larger number of patients before DAA than after DAA (n=48 (73.8%) vs n=25 (38.5%), respectively; p=0.002). After DAA, HCC occurred with a mean interval of 18±16 months and 60% (n=15) of HCC occurred after one year. Conversely, HCC recurrence was similar before and after DAA (n=5 (7.7%) vs n=12 (18.5%), respectively; p=0.092).

Conclusion:

HCC occurrence is significantly reduced in patients with a sustained virologic response to DAA. Conversely, HCC recurrence is not significantly affected by DAA.

Limitations:

The retrospective study design, small sample, and lack of cohort of HCV patients not treated with DAA.

Ethics committee approval

IRB-approved. Informed consent was waived.

Funding:

No funding was received for this study.

Purpose:

To determine whether texture features on pretreatment contrast material–enhanced computed tomographic (CT) images combined with an analysis of sarcopenia can predict the overall survival (OS) in patients with advanced hepatocellular carcinoma (HCC) treated with sorafenib.

Methods and materials:

We retrospectively evaluated 54 patients with advanced hepatocellular (HCC) treated with sorafenib from September 2008-August 2013. The inclusion criteria for this study were available pre-treatment computed tomography (CT) at the same hospital and available OS. 14 patients met these criteria and were enrolled in the study. CT texture and sarcopenia analysis were performed on pretreatment portal venous images. We analysed CT texture first level features (mean, standard deviation, entropy, mean of positive pixels, kurtosis, and skewness) at different anatomic scales ranging from fine to coarse texture. Sarcopenia was defined by a reduced skeletal muscle index calculated from an L3 section CT image. Statistical analysis was performed with SPSS 24.

Results:

The normality distribution of continuous variables was assessed using a Shapiro-Wilk test and a univariate analysis was performed via the appropriate test (Chi-squared, Mann-Whitney U, and a 2-tail paired t-test) in order to identify variables which affected overall survival. Sarcopenia and 4 texture features were significant (p<0.05): mean, entropy, mpp, and skewness. These factors were combined to create a hybrid texture-clinical score, using coefficients derived from a multivariate Cox proportional-hazards regression model. The performance of the model was calculated using the Kaplan-Meier survival analysis (log rank=24.763, p<0.001).

Conclusion:

Texture analysis combined to sarcopenia analysis can help predict overall survival in patients with advanced hepatocellular carcinoma (HCC) treated with sorafenib.

Limitations:

The small population study.

Ethics committee approval

n/a

Funding:

No funding was received for this work.