Preoperatively Identify the Microvascular Invasion of Hepatocellular Carcinoma with the Restricted Spectrum Imaging

Decision Fusion Model for Predicting Microvascular Invasion in Hepatocellular Carcinoma Based on Multi-MR Habitat Imaging and Machine-Learning Classifiers

RATIONALE AND OBJECTIVES

Accurate prediction of microvascular invasion (MVI) in hepatocellular carcinoma (HCC) is crucial for guiding treatment. This study evaluates and compares the performance of clinicoradiologic, traditional radiomics, deep-learning radiomics, feature fusion, and decision fusion models based on multi-region MR habitat imaging using six machine-learning classifiers.

MATERIALS AND METHODS

We retrospectively included 300 HCC patients. The intratumoral and peritumoral regions were segmented into distinct habitats, from which radiomics and deep-learning features were extracted using arterial phase MR images. To reduce feature dimensionality, we applied intra-class correlation coefficient (ICC) analysis, Pearson correlation coefficient (PCC) filtering, and recursive feature elimination (RFE). Based on the selected optimal features, prediction models were constructed using decision tree (DT), K-nearest neighbors (KNN), logistic regression (LR), random forest (RF), support vector machine (SVM), and XGBoost (XGB) classifiers. Additionally, fusion models were developed utilizing both feature fusion and decision fusion strategies. The performance of these models was validated using the area under the receiver operating characteristic curve (ROC AUC), calibration curves, and decision curve analysis.

RESULTS

The decision fusion model (VOI-Peri10-1) using LR and integrating clinicoradiologic, radiomics, and deep-learning features achieved the highest AUC of 0.808 (95% confidence interval [CI]: 0.807-0.912) in the test cohort, with good calibration (Hosmer-Lemeshow test, P > 0.050) and clinical net benefit.

CONCLUSION

The LR-based decision fusion model integrating clinicoradiologic, radiomics, and deep-learning features shows promise for preoperative prediction of MVI in HCC, aiding in patient outcome predictions and personalized treatment planning.

Histogram analysis of continuous-time random walk and restrictive spectrum imaging for identifying hepatocellular carcinoma and intrahepatic cholangiocarcinoma

Background

To compare the ability and potential additional value of various diffusion models, including continuous-time random walk (CTRW), restrictive spectrum imaging (RSI), and diffusion-weighted imaging (DWI), as well as their associated histograms, in distinguishing the pathological subtypes of liver cancer.

Methods

40 patients with liver cancer were included in this study. Histogram metrics were derived from CTRW (D, α, β), RSI (f1, f2, f3), and DWI (ADC) parameters across the entire tumor volume. Statistical analyses included the Chi-square test, independent samples t-test, Mann-Whitney U test, ROC, logistic regression, and Spearman correlation.

Results

Patients with hepatocellular carcinoma exhibited higher values in f1 median, f1 20th, f1 40th, and f1 60th compared to patients with intrahepatic cholangiocarcinoma, whereas Dmean, Dmedian, D40th, D60th, and D80th percentiles were lower (P<0.05). Among the individual histogram parameters, f1 40th percentile demonstrated the highest accuracy (AUC = 0.717). Regarding the combined and single models, the total combined model exhibited the best diagnostic performance (AUC = 0.792). Although RSI showed higher diagnostic efficacy than CTRW (AUC = 0.731, 0.717), the combination of CTRW and RSI further improved diagnostic performance (AUC = 0.787), achieving superior sensitivity and specificity (sensitivity = 0.72, specificity = 0.80).

Conclusion

CTRW, RSI, and their corresponding histogram parameters demonstrated the ability to distinguish between pathological subtypes of liver cancer. Moreover, whole-lesion histogram parameters provided more comprehensive statistical insights compared to mean values alone.

Gadoxetic Acid-Enhanced MRI-Based Radiomic Models for Preoperative Risk Prediction and Prognostic Assessment of Proliferative Hepatocellular Carcinoma

RATIONALE AND OBJECTIVES

Proliferative hepatocellular carcinoma (HCC) is associated with high invasiveness and poor prognosis. This study aimed to investigate the preoperative risk prediction and prognostic value of different radiomics models and a nomogram for proliferative HCC.

MATERIALS AND METHODS

Patients were randomly divided into a training cohort (n = 156) and a validation cohort (n = 66) in a 7:3 ratio. Original and delta (the different value between imaging features extracted from two different phases) radiomics features were extracted from T1-weighted imaging (T1WI), arterial, and hepatobiliary phases to construct models using different machine learning algorithms. Logistic regression was used to select clinical independent risk factors. A nomogram was constructed by integrating the optimal radiomics model score with independent risk factors. The diagnostic efficacy and clinical utility of the models were assessed. Subsequently, patients were stratified into high-risk and low-risk subgroups based on radiomics model scores and nomogram scores, and both recurrence-free survival (RFS) and overall survival (OS) were evaluated.

RESULTS

Multivariate logistic regression analysis showed that BCLC stage and combined radscore were independent predictors of proliferative HCC. The area under the curve (AUC) of the nomogram incorporating these factors was 0.838 and 0.801 in the training and validation cohorts, respectively, with good predictive performance. Multivariate Cox regression analysis shows that the delta radiomics model (DR)-predicted proliferative HCC can independently predict RFS and OS, with scores from the delta radiomics model performing best in prognostic risk stratification.

CONCLUSION

The nomogram can effectively predict proliferative HCC, while different radiomics models and the nomogram can offer varying prognostic stratification values.

The Diagnostic Accuracy Between Radiomics Model and Non-radiomics Model for Preoperative of Microvascular Invasion of Solitary Hepatocellular Carcinoma: A Systematic Review and Meta-analysis

RATIONALE AND OBJECTIVES

Microvascular invasion (MVI) is a key prognostic factor for hepatocellular carcinoma (HCC). The predictive models for solitary HCC could potentially integrate more comprehensive tumor information. Owing to the diverse findings across studies, we aimed to compare radiomic and non-radiomic methods for preoperative MVI detection in solitary HCC.

MATERIALS AND METHODS

Articles were reviewed from databases including PubMed, Embase, Web of Science, and the Cochrane Library until April 7, 2023. The pooled sensitivity, specificity, positive likelihood ratio (PLR), and negative likelihood ratio (NLR) were calculated using a random-effects model within a 95% confidence interval (CI). Diagnostic accuracy was assessed using summary receiver-operating characteristic curves and the area under the curve (AUC). Meta-regression and Z-tests identified heterogeneity and compared the predictive accuracy. Subgroup analyses were performed to compare the AUC of two methods according to study type, study design, tumor size, modeling methods, and imaging modality.

RESULTS

The analysis incorporated 26 studies involving 3539 patients with solitary HCC. The radiomics models showed a pooled sensitivity and specificity of 0.79 (95%CI: 0.72-0.85) and 0.78 (95%CI: 0.73-0.82), with an AUC at 0.85 (95%CI: 0.82-0.88). Conversely, the non-radiomics models had sensitivity and specificity of 0.74 (95%CI: 0.65-0.81) and 0.88 (95%CI: 0.82-0.92) and an AUC of 0.88 (95%CI: 0.85-0.91). Subgroups with preoperative MRI, larger tumors, and functional imaging had higher accuracy than those using preoperative CT, smaller tumors, and conventional imaging.

CONCLUSION

Non-radiomic methods outperformed radiomic methods, but high heterogeneity calls across studies for cautious interpretation.

Detecting microvascular invasion in hepatocellular carcinoma using the impeded diffusion fraction technique to sense macromolecular coordinated water

OBJECTIVES

Impeded diffusion fraction (IDF) is a novel and promising diffusion-weighted imaging (DWI) technique that allows for the detection of various diffusion compartments, including macromolecular coordinated water, free diffusion, perfusion, and cellular free water. This study aims to investigate the clinical potential of IDF-DWI in detecting microvascular invasion (MVI) in hepatocellular carcinoma (HCC).

METHODS

66 patients were prospectively included. Metrics derived from IDF-DWI and the apparent diffusion coefficient (ADC) were calculated. Multivariate logistic regression was employed to identify clinical risk factors. Diagnostic performance was evaluated using the area under the receiver operating characteristics curve (AUC-ROC), the area under the precision-recall curve (AUC-PR), and the calibration error (cal-error). Additionally, a power analysis was conducted to determine the required sample size.

RESULTS

The results suggested a significantly higher fraction of impeded diffusion (FID) originating from IDF-DWI in MVI-positive HCCs (p < 0.001). Moreover, the ADC was found to be significantly lower in MVI-positive HCCs (p = 0.019). Independent risk factors of MVI included larger tumor size and elevated alpha-fetoprotein (AFP) levels. The nomogram model incorporating ADC, FID, tumor size, and AFP level yielded the highest diagnostic accuracy for MVI (AUC-PR = 0.804, AUC-ROC = 0.783, cal-error = 0.044), followed by FID (AUC-PR = 0.693, AUC-ROC = 0.760, cal-error = 0.060) and ADC (AUC-PR = 0.570, AUC-ROC = 0.651, cal-error = 0.164).

CONCLUSION

IDF-DWI shows great potential in noninvasively, accurately, and preoperatively detecting MVI in HCC and may offer clinical benefits for prognostic prediction and determination of treatment strategy.

Non-invasive assessment of response to transcatheter arterial chemoembolization for hepatocellular carcinoma with the deep neural networks-based radiomics nomogram

BACKGROUND

Transcatheter arterial chemoembolization (TACE) is a mainstay treatment for intermediate and advanced hepatocellular carcinoma (HCC), with the potential to enhance patient survival. Preoperative prediction of postoperative response to TACE in patients with HCC is crucial.

PURPOSE

To develop a deep neural network (DNN)-based nomogram for the non-invasive and precise prediction of TACE response in patients with HCC.

MATERIAL AND METHODS

We retrospectively collected clinical and imaging data from 110 patients with HCC who underwent TACE surgery. Radiomics features were extracted from specific imaging methods. We employed conventional machine-learning algorithms and a DNN-based model to construct predictive probabilities (RScore). Logistic regression helped identify independent clinical risk factors, which were integrated with RScore to create a nomogram. We evaluated diagnostic performance using various metrics.

RESULTS

Among the radiomics models, the DNN_LASSO-based one demonstrated the highest predictive accuracy (area under the curve [AUC] = 0.847, sensitivity = 0.892, specificity = 0.791). Peritumoral enhancement and alkaline phosphatase were identified as independent risk factors. Combining RScore with these clinical factors, a DNN-based nomogram exhibited superior predictive performance (AUC = 0.871, sensitivity = 0.844, specificity = 0.873).

CONCLUSION

In this study, we successfully developed a deep learning-based nomogram that can noninvasively and accurately predict TACE response in patients with HCC, offering significant potential for improving the clinical management of HCC.

The value of varying diffusion curvature MRI for assessing the microvascular invasion of hepatocellular carcinoma

PURPOSE

Varying diffusion curvature (VDC) MRI is an emerging diffusion-weighted imaging (DWI) technique that can capture non-Gaussian diffusion behavior and reflect tissue heterogeneity. However, its clinical utility has hardly been evaluated. We aimed to investigate the value of the VDC technique in noninvasively assessing microvascular invasion (MVI) in hepatocellular carcinoma (HCC).

METHODS

74 patients with HCCs, including 39 MVI-positive and 35 MVI-negative HCCs were included into this prospective study. Quantitative metrics between subgroups, clinical risk factors, as well as diagnostic performance were evaluated. The power analysis was also carried out to determine the statistical power.

RESULTS

MVI-positive HCCs exhibited significantly higher VDC-derived structural heterogeneity measure, D1 (0.680 ± 0.100 × 10-3 vs 0.572 ± 0.148 × 10-3 mm2/s, p = 0.001) and lower apparent diffusion coefficient (ADC) (1.350 ± 0.166 × 10-3 vs 1.471 ± 0.322 × 10-3 mm2/s, p = 0.0495) compared to MVI-negative HCCs. No statistical significance was observed for VDC-derived diffusion coefficient, D0 between the subgroups (p = 0.562). Tumor size (odds ratio (OR) = 1.242) and alpha-fetoprotein (AFP) (OR = 2.527) were identified as risk factors for MVI. A predictive nomogram was constructed based on D1, ADC, tumor size, and AFP, which exhibited the highest diagnostic accuracy (AUC = 0.817), followed by D1 (AUC = 0.753) and ADC (AUC = 0.647). The diagnostic performance of the nomogram-based model was also validated by the calibration curve and decision curve.

CONCLUSION

VDC can aid in the noninvasive and preoperative diagnosis of HCC with MVI, which may result in the clinical benefit in terms of prognostic prediction and clinical decision-making.

Preoperative Grading of Rectal Cancer with Multiple DWI Models, DWI-Derived Biological Markers, and Machine Learning Classifiers

Background: this study aimed to utilize various diffusion-weighted imaging (DWI) techniques, including mono-exponential DWI, intravoxel incoherent motion (IVIM), and diffusion kurtosis imaging (DKI), for the preoperative grading of rectal cancer. Methods: 85 patients with rectal cancer were enrolled in this study. Mann-Whitney U tests or independent Student's t-tests were conducted to identify DWI-derived parameters that exhibited significant differences. Spearman or Pearson correlation tests were performed to assess the relationships among different DWI-derived biological markers. Subsequently, four machine learning classifier-based models were trained using various DWI-derived parameters as input features. Finally, diagnostic performance was evaluated using ROC analysis with 5-fold cross-validation. Results: With the exception of the pseudo-diffusion coefficient (Dp), IVIM-derived and DKI-derived parameters all demonstrated significant differences between low-grade and high-grade rectal cancer. The logistic regression-based machine learning classifier yielded the most favorable diagnostic efficacy (AUC: 0.902, 95% Confidence Interval: 0.754-1.000; Specificity: 0.856; Sensitivity: 0.925; Youden Index: 0.781). Conclusions: utilizing multiple DWI-derived biological markers in conjunction with a strategy employing multiple machine learning classifiers proves valuable for the noninvasive grading of rectal cancer.

The Feasibility of Using Tri-Exponential Intra-Voxel Incoherent Motion DWI for Identifying the Microvascular Invasion in Hepatocellular Carcinoma

Purpose

To assess the effectiveness of tri-exponential Intra-Voxel Incoherent Motion (tri-IVIM) MRI in preoperatively identifying microvascular invasion (MVI) in hepatocellular carcinoma (HCC).

Patients and Methods

In this prospective study, 67 patients with HCC were included. Metrics from bi-exponential IVIM (bi-IVIM) and tri-IVIM were calculated. Subgroup comparisons were analyzed using the independent Student's t-test or Mann-Whitney U-test. Logistic regression was performed to explore clinical risk factors. Diagnostic performance was assessed using receiver operating characteristic (ROC) curves, calibration curves and decision curve analysis.

Results

MVI-positive HCCs exhibited significantly lower true diffusion coefficient (Dt) from bi-IVIM, as well as fast-diffusion coefficients (Df) and slow-diffusion coefficients (Ds) from tri-IVIM, compared to MVI-negative HCCs (p < 0.05). Tumor size and alpha-fetoprotein (AFP) were identified as risk factors. The combination of tri-IVIM-derived metrics (Ds and Df) yielded higher diagnostic accuracy (AUC = 0.808) compared to bi-IVIM (AUC = 0.741). A predictive model based on a nomogram was constructed using Ds, Df, tumor size, and AFP, resulting in the highest diagnostic accuracy (AUC = 0.859). Decision curve analysis indicated that the constructed model, provided the highest net benefit by accurately stratifying the risk of MVI, followed by tri-IVIM and bi-IVIM.

Conclusion

Tri-IVIM can provide information on perfusion and diffusion for evaluating MVI in HCC. Additionally, tri-IVIM outperformed bi-IVIM in identifying MVI-positive HCC. By integrating clinical risk factors and metrics from tri-IVIM, a predictive nomogram exhibited the highest diagnostic accuracy, potentially aiding in the noninvasive and preoperative assessment of MVI.