Preoperative Microvascular Invasion Prediction to Assist in Surgical Plan for Single Hepatocellular Carcinoma: Better Together with Radiomics

Multitask deep learning for prediction of microvascular invasion and recurrence-free survival in hepatocellular carcinoma based on MRI images

BACKGROUND AND AIMS

Accurate preoperative prediction of microvascular invasion (MVI) and recurrence-free survival (RFS) is vital for personalised hepatocellular carcinoma (HCC) management. We developed a multitask deep learning model to predict MVI and RFS using preoperative MRI scans.

METHODS

Utilising a retrospective dataset of 725 HCC patients from seven institutions, we developed and validated a multitask deep learning model focused on predicting MVI and RFS. The model employs a transformer architecture to extract critical features from preoperative MRI scans. It was trained on a set of 234 patients and internally validated on a set of 58 patients. External validation was performed using three independent sets (n = 212, 111, 110).

RESULTS

The multitask deep learning model yielded high MVI prediction accuracy, with AUC values of 0.918 for the training set and 0.800 for the internal test set. In external test sets, AUC values were 0.837, 0.815 and 0.800. Radiologists' sensitivity and inter-rater agreement for MVI prediction improved significantly when integrated with the model. For RFS, the model achieved C-index values of 0.763 in the training set and ranged between 0.628 and 0.728 in external test sets. Notably, PA-TACE improved RFS only in patients predicted to have high MVI risk and low survival scores (p < .001).

CONCLUSIONS

Our deep learning model allows accurate MVI and survival prediction in HCC patients. Prospective studies are warranted to assess the clinical utility of this model in guiding personalised treatment in conjunction with clinical criteria.

Preoperative Three-Dimensional Morphological Tumor Features Predict Microvascular Invasion in Hepatocellular Carcinoma

RATIONALE AND OBJECTIVES

The study was designed to evaluate microvascular invasion (MVI) using three-dimensional (3D) morphological indicators prior to surgery.

MATERIALS AND METHODS

This retrospective study included 156 patients with hepatocellular carcinoma (HCC) at our hospital from 2017 to 2018. Through thin-layer CT scanning and 3D reconstruction, the tumor surface inclination angles can be quantitatively analyzed to determine the surface irregularity rate (SIR), which serves as a comprehensive assessment method for tumor irregularity based on preoperative 3D morphological evaluation. Univariate and multivariate logistic regression analyses were employed to investigate the correlation with MVI.

RESULTS

The SIR was related to MVI (OR: 10.667, P < 0.001). Multivariate logistic regression analysis showed that the SIR was an independent risk factor for MVI. The area under the receiver operating characteristic curve (ROC) of prediction model composed of the morphological indicator SIR was 0.831 (95% confidence interval: 0.759-0.895).

CONCLUSION

The preoperative 3D morphological indicator SIR of a tumor is an accurate predictor of MVI, providing a valuable tool in clinical decision-making.

Establishment and Validation of the LI-RADS Morphologic Type II Hepatocellular Carcinoma Early Recurrence Risk Scoring System

BACKGROUND

Tumor morphology links to early recurrence of hepatocellular carcinoma. Controversy exists regarding the recurrence risk of Liver Imaging Reporting and Data System morphologic Type II hepatocellular carcinoma. This study aims to explore risk factors for early recurrence of Type II hepatocellular carcinoma.

METHODS

Retrospective analysis of hepatocellular carcinoma patients who underwent curative resection and preoperative contrast-enhanced MRI from June 2016 to June 2020. Our patients formed the development set, and hepatocellular carcinoma patients from the TCIA database served as validation. Univariable and multivariable Cox regression identified independent risk factors for early recurrence. A risk scoring system was established for risk stratification, and an early recurrence prediction model was developed and validated.

RESULTS

95 Type II hepatocellular carcinoma patients were in the development set, and 29 cases were in the validation set. Early recurrence rates were 33.7% and 37.9%, respectively. Multivariate analysis revealed age, histological grade, AFP, and intratumoral hemorrhage as independent risk factors for early recurrence. The model's diagnostic performance for early recurrence was AUC = 0.817 in the development set. A scoring system classified patients into low-risk (scores ≤ 3) and high-risk (scores > 3) groups. The high-risk group had significantly lower recurrence-free survival (40.0% vs 73.2%, P = 0.001), consistent with the validation set (25.0% vs 73.3%, P = 0.028).

CONCLUSIONS

The risk scoring system demonstrated excellent discrimination and predictive ability, aiding clinicians in assessing early recurrence risk and identifying high-risk individuals effectively.

MRI Features for Predicting Microvascular Invasion and Postoperative Recurrence in Hepatocellular Carcinoma Without Peritumoral Hypointensity

Purpose

To identify MRI features of hepatocellular carcinoma (HCC) that predict microvascular invasion (MVI) and postoperative intrahepatic recurrence in patients without peritumoral hepatobiliary phase (HBP) hypointensity.

Patients and Methods

One hundred and thirty patients with HCC who underwent preoperative gadoxetate-enhanced MRI and curative hepatic resection were retrospectively reviewed. Two radiologists reviewed all preoperative MR images and assessed the radiological features of HCCs. The ability of peritumoral HBP hypointensity to identify MVI and intrahepatic recurrence was analyzed. We then assessed the MRI features of HCC that predicted the MVI and intrahepatic recurrence-free survival (RFS) in the subgroup without peritumoral HBP hypointensity. Finally, a two-step flowchart was constructed to assist in clinical decision-making.

Results

Peritumoral HBP hypointensity (odds ratio, 3.019; 95% confidence interval: 1.071-8.512; P=0.037) was an independent predictor of MVI. The sensitivity, specificity, positive predictive value, negative predictive value, and AUROC of peritumoral HBP hypointensity in predicting MVI were 23.80%, 91.04%, 71.23%, 55.96%, and 0.574, respectively. Intrahepatic RFS was significantly shorter in patients with peritumoral HBP hypointensity (P<0.001). In patients without peritumoral HBP hypointensity, the only significant difference between MVI-positive and MVI-negative HCCs was the presence of a radiological capsule (P=0.038). Satellite nodule was an independent risk factor for intrahepatic RFS (hazard ratio,3.324; 95% CI: 1.733-6.378; P<0.001). The high-risk HCC detection rate was significantly higher when using the two-step flowchart that incorporated peritumoral HBP hypointensity and satellite nodule than when using peritumoral HBP hypointensity alone (P<0.001).

Conclusion

In patients without peritumoral HBP hypointensity, a radiological capsule is useful for identifying MVI and satellite nodule is an independent risk factor for intrahepatic RFS.

CT radiomics for noninvasively predicting NQO1 expression levels in hepatocellular carcinoma

Using noninvasive radiomics to predict pathological biomarkers is an innovative work worthy of exploration. This retrospective cohort study aimed to analyze the correlation between NAD(P)H quinone oxidoreductase 1 (NQO1) expression levels and the prognosis of patients with hepatocellular carcinoma (HCC) and to construct radiomic models to predict the expression levels of NQO1 prior to surgery. Data of patients with HCC from The Cancer Genome Atlas (TCGA) and the corresponding arterial phase-enhanced CT images from The Cancer Imaging Archive were obtained for prognosis analysis, radiomic feature extraction, and model development. In total, 286 patients with HCC from TCGA were included. According to the cut-off value calculated using R, patients were divided into high-expression (n = 143) and low-expression groups (n = 143). Kaplan-Meier survival analysis showed that higher NQO1 expression levels were significantly associated with worse prognosis in patients with HCC (p = 0.017). Further multivariate analysis confirmed that high NQO1 expression was an independent risk factor for poor prognosis (HR = 1.761, 95% CI: 1.136-2.73, p = 0.011). Based on the arterial phase-enhanced CT images, six radiomic features were extracted, and a new bi-regional radiomics model was established, which could noninvasively predict higher NQO1 expression with good performance. The area under the curve (AUC) was 0.9079 (95% CI 0.8127-1.0000). The accuracy, sensitivity, and specificity were 0.86, 0.88, and 0.84, respectively, with a threshold value of 0.404. The data verification of our center showed that this model has good predictive efficiency, with an AUC of 0.8791 (95% CI 0.6979-1.0000). In conclusion, there existed a significant correlation between the CT image features and the expression level of NQO1, which could indirectly reflect the prognosis of patients with HCC. The predictive model based on arterial phase CT imaging features has good stability and diagnostic efficiency and is a potential means of identifying the expression level of NQO1 in HCC tissues before surgery.

Predicting Microvascular Invasion in Hepatocellular Carcinoma Using CT-based Radiomics Model

Background Prediction of microvascular invasion (MVI) may help determine treatment strategies for hepatocellular carcinoma (HCC). Purpose To develop a radiomics approach for predicting MVI status based on preoperative multiphase CT images and to identify MVI-associated differentially expressed genes. Materials and Methods Patients with pathologically proven HCC from May 2012 to September 2020 were retrospectively included from four medical centers. Radiomics features were extracted from tumors and peritumor regions on preoperative registration or subtraction CT images. In the training set, these features were used to build five radiomics models via logistic regression after feature reduction. The models were tested using internal and external test sets against a pathologic reference standard to calculate area under the receiver operating characteristic curve (AUC). The optimal AUC radiomics model and clinical-radiologic characteristics were combined to build the hybrid model. The log-rank test was used in the outcome cohort (Kunming center) to analyze early recurrence-free survival and overall survival based on high versus low model-derived score. RNA sequencing data from The Cancer Image Archive were used for gene expression analysis. Results A total of 773 patients (median age, 59 years; IQR, 49-64 years; 633 men) were divided into the training set (n = 334), internal test set (n = 142), external test set (n = 141), outcome cohort (n = 121), and RNA sequencing analysis set (n = 35). The AUCs from the radiomics and hybrid models, respectively, were 0.76 and 0.86 for the internal test set and 0.72 and 0.84 for the external test set. Early recurrence-free survival (P < .01) and overall survival (P < .007) can be categorized using the hybrid model. Differentially expressed genes in patients with findings positive for MVI were involved in glucose metabolism. Conclusion The hybrid model showed the best performance in prediction of MVI. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Summers in this issue.

Data-Driven Assisted Decision Making for Surgical Procedure of Hepatocellular Carcinoma Resection and Prognostic Prediction: Development and Validation of Machine Learning Models

Background: Currently, surgical decisions for hepatocellular carcinoma (HCC) resection are difficult and not sufficiently personalized. We aimed to develop and validate data driven prediction models to assist surgeons in selecting the optimal surgical procedure for patients. Methods: Retrospective data from 361 HCC patients who underwent radical resection in two institutions were included. End-to-end deep learning models were built to automatically segment lesions from the arterial phase (AP) of preoperative dynamic contrast enhanced magnetic resonance imaging (DCE-MRI). Clinical baseline characteristics and radiomic features were rigorously screened. The effectiveness of radiomic features and radiomic-clinical features was also compared. Three ensemble learning models were proposed to perform the surgical procedure decision and the overall survival (OS) and recurrence-free survival (RFS) predictions after taking different solutions, respectively. Results: SegFormer performed best in terms of automatic segmentation, achieving a Mean Intersection over Union (mIoU) of 0.8860. The five-fold cross-validation results showed that inputting radiomic-clinical features outperformed using only radiomic features. The proposed models all outperformed the other mainstream ensemble models. On the external test set, the area under the receiver operating characteristic curve (AUC) of the proposed decision model was 0.7731, and the performance of the prognostic prediction models was also relatively excellent. The application web server based on automatic lesion segmentation was deployed and is available online. Conclusions: In this study, we developed and externally validated the surgical decision-making procedures and prognostic prediction models for HCC for the first time, and the results demonstrated relatively accurate predictions and strong generalizations, which are expected to help clinicians optimize surgical procedures.

A Novel Online Calculator to Predict Risk of Microvascular Invasion in the Preoperative Setting for Hepatocellular Carcinoma Patients Undergoing Curative-Intent Surgery

BACKGROUND

The presence of microvascular invasion (MVI) has been highlighted as an important determinant of hepatocellular carcinoma (HCC) prognosis. We sought to build and validate a novel model to predict MVI in the preoperative setting.

METHODS

Patients who underwent curative-intent surgery for HCC between 2000 and 2020 were identified using a multi-institutional database. Preoperative predictive models for MVI were built, validated, and used to develop a web-based calculator.

RESULTS

Among 689 patients, MVI was observed in 323 patients (46.9%). On multivariate analysis in the test cohort, preoperative parameters associated with MVI included α-fetoprotein (AFP; odds ratio [OR] 1.50, 95% confidence interval [CI] 1.23-1.83), imaging tumor burden score (TBS; hazard ratio [HR] 1.11, 95% CI 1.04-1.18), and neutrophil-to-lymphocyte ratio (NLR; OR 1.18, 95% CI 1.03-1.35). An online calculator to predict MVI was developed based on the weighted β-coefficients of these three variables ( https://yutaka-endo.shinyapps.io/MVIrisk/ ). The c-index of the test and validation cohorts was 0.71 and 0.72, respectively. Patients with a high risk of MVI had worse disease-free survival (DFS) and overall survival (OS) compared with low-risk MVI patients (3-year DFS: 33.0% vs. 51.9%, p < 0.001; 5-year OS: 44.2% vs. 64.8%, p < 0.001). DFS was worse among patients who underwent an R1 versus R0 resection among those patients at high risk of MVI (R0 vs. R1 resection: 3-year DFS, 36.3% vs. 16.1%, p = 0.002). In contrast, DFS was comparable among patients at low risk of MVI regardless of margin status (R0 vs. R1 resection: 3-year DFS, 52.9% vs. 47.3%, p = 0.16).

CONCLUSION

Preoperative assessment of MVI using the online tool demonstrated very good accuracy to predict MVI.

Radiomics: A radiological evidence-based artificial intelligence technique to facilitate personalized precision medicine in hepatocellular carcinoma

The high postoperative recurrence rates in hepatocellular carcinoma (HCC) remain a major hurdle in its management. Appropriate staging and treatment selection may alleviate the extent of fatal recurrence. However, effective methods to preoperatively evaluate pathophysiologic and molecular characteristics of HCC are lacking. Imaging plays a central role in HCC diagnosis and stratification due to the non-invasive diagnostic criteria. Vast and crucial information is hidden within image data. Other than providing a morphological sketch for lesion diagnosis, imaging could provide new insights to describe the pathophysiological and genetic landscape of HCC. Radiomics aims to facilitate diagnosis and prognosis of HCC using artificial intelligence techniques to harness the immense information contained in medical images. Radiomics produces a set of archetypal and robust imaging features that are correlated to key pathological or molecular biomarkers to preoperatively risk-stratify HCC patients. Inferred with outcome data, comprehensive combination of radiomic, clinical and/or multi-omics data could also improve direct prediction of response to treatment and prognosis. The evolution of radiomics is changing our understanding of personalized precision medicine in HCC management. Herein, we review the key techniques and clinical applications in HCC radiomics and discuss current limitations and future opportunities to improve clinical decision making.