Diagnostic performance of Gd-EOB-DTPA-enhanced MRI for evaluation of liver dysfunction: a multivariable analysis of 3T MRI sequences

Radiomics and nomogram of magnetic resonance imaging for preoperative prediction of microvascular invasion in small hepatocellular carcinoma

BACKGROUND

Microvascular invasion (MVI) of small hepatocellular carcinoma (sHCC) (≤ 3.0 cm) is an independent prognostic factor for poor progression-free and overall survival. Radiomics can help extract imaging information associated with tumor pathophysiology.

AIM

To develop and validate radiomics scores and a nomogram of gadolinium ethoxybenzyl-diethylenetriamine pentaacetic acid (Gd-EOB-DTPA)-enhanced magnetic resonance imaging (MRI) for preoperative prediction of MVI in sHCC.

METHODS

In total, 415 patients were diagnosed with sHCC by postoperative pathology. A total of 221 patients were retrospectively included from our hospital. In addition, we recruited 94 and 100 participants as independent external validation sets from two other hospitals. Radiomics models of Gd-EOB-DTPA-enhanced MRI and diffusion-weighted imaging (DWI) were constructed and validated using machine learning. As presented in the radiomics nomogram, a prediction model was developed using multivariable logistic regression analysis, which included radiomics scores, radiologic features, and clinical features, such as the alpha-fetoprotein (AFP) level. The calibration, decision-making curve, and clinical usefulness of the radiomics nomogram were analyzed. The radiomic nomogram was validated using independent external cohort data. The areas under the receiver operating curve (AUC) were used to assess the predictive capability.

RESULTS

Pathological examination confirmed MVI in 64 (28.9%), 22 (23.4%), and 16 (16.0%) of the 221, 94, and 100 patients, respectively. AFP, tumor size, non-smooth tumor margin, incomplete capsule, and peritumoral hypointensity in hepatobiliary phase (HBP) images had poor diagnostic value for MVI of sHCC. Quantitative radiomic features (1409) of MRI scans) were extracted. The classifier of logistic regression (LR) was the best machine learning method, and the radiomics scores of HBP and DWI had great diagnostic efficiency for the prediction of MVI in both the testing set (hospital A) and validation set (hospital B, C). The AUC of HBP was 0.979, 0.970, and 0.803, respectively, and the AUC of DWI was 0.971, 0.816, and 0.801 (P < 0.05), respectively. Good calibration and discrimination of the radiomics and clinical combined nomogram model were exhibited in the testing and two external validation cohorts (C-index of HBP and DWI were 0.971, 0.912, 0.808, and 0.970, 0.843, 0.869, respectively). The clinical usefulness of the nomogram was further confirmed using decision curve analysis.

CONCLUSION

AFP and conventional Gd-EOB-DTPA-enhanced MRI features have poor diagnostic accuracies for MVI in patients with sHCC. Machine learning with an LR classifier yielded the best radiomics score for HBP and DWI. The radiomics nomogram developed as a noninvasive preoperative prediction method showed favorable predictive accuracy for evaluating MVI in sHCC.

MELIF, a Fully Automated Liver Function Score Calculated from Gd-EOB-DTPA-Enhanced MR Images: Diagnostic Performance vs. the MELD Score

In the management of patients with chronic liver disease, the assessment of liver function is essential for treatment planning. Gd-EOB-DTPA-enhanced MRI allows for both the acquisition of anatomical information and regional liver function quantification. The objective of this study was to demonstrate and evaluate the diagnostic performance of two fully automatically generated imaging-based liver function scores that take the whole liver into account. T1 images from the native and hepatobiliary phases and the corresponding T1 maps from 195 patients were analyzed. A novel artificial-intelligence-based software prototype performed image segmentation and registration, calculated the reduction rate of the T1 relaxation time for the whole liver (rrT1_liver) and used it to calculate a personalized liver function score, then generated a unified score—the MELIF score—by combining the liver function score with a patient-specific factor that included weight, height and liver volume. Both scores correlated strongly with the MELD score, which is used as a reference for global liver function. However, MELIF showed a stronger correlation than the rrT1_liver score. This study demonstrated that the fully automated determination of total liver function, regionally resolved, using MR liver imaging is feasible, providing the opportunity to use the MELIF score as a diagnostic marker in future prospective studies.

Assessment of Liver Function With MRI: Where Do We Stand?

Liver disease and hepatocellular carcinoma (HCC) have become a global health burden. For this reason, the determination of liver function plays a central role in the monitoring of patients with chronic liver disease or HCC. Furthermore, assessment of liver function is important, e.g., before surgery to prevent liver failure after hepatectomy or to monitor the course of treatment. Liver function and disease severity are usually assessed clinically based on clinical symptoms, biopsy, and blood parameters. These are rather static tests that reflect the current state of the liver without considering changes in liver function. With the development of liver-specific contrast agents for MRI, noninvasive dynamic determination of liver function based on signal intensity or using T1 relaxometry has become possible. The advantage of this imaging modality is that it provides additional information about the vascular structure, anatomy, and heterogeneous distribution of liver function. In this review, we summarized and discussed the results published in recent years on this technique. Indeed, recent data show that the T1 reduction rate seems to be the most appropriate value for determining liver function by MRI. Furthermore, attention has been paid to the development of automated tools for image analysis in order to uncover the steps necessary to obtain a complete process flow from image segmentation to image registration to image analysis. In conclusion, the published data show that liver function values obtained from contrast-enhanced MRI images correlate significantly with the global liver function parameters, making it possible to obtain both functional and anatomic information with a single modality.

T1 reduction rate with Gd-EOB-DTPA determines liver function on both 1.5 T and 3 T MRI

Magnetic resonance T1 mapping before and after Gd-EOB-DTPA administration allows quantification of the T1 reduction rate as a non-invasive surrogate marker of liver function. A major limitation of T1 relaxation time measurement is its dependency on MRI field strengths. Since T1 reduction rate is calculated as the relative shortening of T1 relaxation time before and after contrast administration, we hypothesized that the T1 reduction rate is comparable between 1.5 and 3 T. We thus compared liver T1 relaxation times between 1.5 and 3 T in a total of 243 consecutive patients (124, 1.5 T and 119, 3 T) between 09/2018 and 07/2019. T1 reduction rates were compared between patients with no cirrhosis and patients with cirrhosis Child-Pugh A-C. There was no significant difference of T1 reduction rate between 1.5 and 3 T in any patient group (p-value 0.126-0.861). On both 1.5 T and 3 T, T1 reduction rate allowed to differentiate between patients with no cirrhosis and patients with liver cirrhosis Child A-C (p < 0.001). T1 reduction rate showed a good performance to predict liver cirrhosis Child A (AUC = 0.83, p < 0.001), Child B (AUC = 0.83, p < 0.001) and Child C (AUC = 0.92, p < 0.001). In conclusion, T1 reduction rate allows to determine liver function on Gd-EOB-DTPA MRI with comparable values on 1.5 T and 3 T.

Diagnostic Accuracy of Artificial Intelligence Based on Imaging Data for Preoperative Prediction of Microvascular Invasion in Hepatocellular Carcinoma: A Systematic Review and Meta-Analysis

Background

The presence of microvascular invasion (MVI) is considered an independent prognostic factor associated with early recurrence and poor survival in hepatocellular carcinoma (HCC) patients after resection. Artificial intelligence (AI), mainly consisting of non-deep learning algorithms (NDLAs) and deep learning algorithms (DLAs), has been widely used for MVI prediction in medical imaging.

Aim

To assess the diagnostic accuracy of AI algorithms for non-invasive, preoperative prediction of MVI based on imaging data.

Methods

Original studies reporting AI algorithms for non-invasive, preoperative prediction of MVI based on quantitative imaging data were identified in the databases PubMed, Embase, and Web of Science. The quality of the included studies was assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) scale. The pooled sensitivity, specificity, positive likelihood ratio (PLR), and negative likelihood ratio (NLR) were calculated using a random-effects model with 95% CIs. A summary receiver operating characteristic curve and the area under the curve (AUC) were generated to assess the diagnostic accuracy of the deep learning and non-deep learning models. In the non-deep learning group, we further performed meta-regression and subgroup analyses to identify the source of heterogeneity.

Results

Data from 16 included studies with 4,759 cases were available for meta-analysis. Four studies on deep learning models, 12 studies on non-deep learning models, and two studies compared the efficiency of the two types. For predictive performance of deep learning models, the pooled sensitivity, specificity, PLR, NLR, and AUC values were 0.84 [0.75–0.90], 0.84 [0.77–0.89], 5.14 [3.53–7.48], 0.2 [0.12–0.31], and 0.90 [0.87–0.93]; and for non-deep learning models, they were 0.77 [0.71–0.82], 0.77 [0.73–0.80], 3.30 [2.83–3.84], 0.30 [0.24–0.38], and 0.82 [0.79–0.85], respectively. Subgroup analyses showed a significant difference between the single tumor subgroup and the multiple tumor subgroup in the pooled sensitivity, NLR, and AUC.

Conclusion

This meta-analysis demonstrates the high diagnostic accuracy of non-deep learning and deep learning methods for MVI status prediction and their promising potential for clinical decision-making. Deep learning models perform better than non-deep learning models in terms of the accuracy of MVI prediction, methodology, and cost-effectiveness.

Systematic Review Registration

https://www.crd.york.ac.uk/PROSPERO/display_record.php? RecordID=260891, ID:CRD42021260891.

Comparison of Conventional Gadoxetate Disodium-Enhanced MRI Features and Radiomics Signatures With Machine Learning for Diagnosing Microvascular Invasion

OBJECTIVE. This study aimed to determine the best model for predicting microvascular invasion (MVI) of hepatocellular carcinoma (HCC) using conventional gadolinium-ethoxybenzyl-diethylenetriamine pentaacetic acid (gadoxetate disodium)-enhanced MRI features and radiomics signatures with machine learning. MATERIALS AND METHODS. This retrospective study included 269 patients with a postoperative pathologic diagnosis of HCC. Gadoxetate disodium-enhanced MRI features were assessed, including T1 relaxation time, tumor margin, tumor size, peritumoral enhancement, peritumoral hypointensity, and ADC. Radiomics models were constructed and validated by machine learning. The least absolute shrinkage and selection operator (LASSO) was used for feature selection, and radiomics-based LASSO models were constructed with six classifiers. Predictive capability was assessed using the ROC AUC. RESULTS. Histologic examination confirmed MVI in 111 (41.3%) of the 269 patients. ADC value, nonsmooth tumor margin, and 20-minute T1 relaxation time showed diagnostic accuracy with AUC values of 0.850, 0.847, and 0.846, respectively (p < .05 for all). A total of 1395 quantitative imaging features were extracted. In the hepatobiliary phase (HBP) model, the support vector machine (SVM), extreme gradient boosting (XGBoost), and logistic regression (LR) classifiers showed greater diagnostic efficiency for predicting MVI, with AUCs of 0.942, 0.938, and 0.936, respectively (p < .05 for all). CONCLUSION. ADC value, nonsmooth tumor margin, and 20-minute T1 relaxation time show high diagnostic accuracy for predicting MVI. Radiomics signatures with machine learning can further improve the ability to predict MVI and are best modeled during HBP. The SVM, XGBoost, and LR classifiers may serve as potential biomarkers to evaluate MVI.