Page 156 - Read Online

Grewal et al. Art Int Surg 2023;3:217-32 https://dx.doi.org/10.20517/ais.2023.28 Page 223

accurately classifying adenoma 0.80-0.93/0.78-0.93 [87-91] . Another study distinguished combined
hepatocellular cholangiocarcinoma (C-HCC) from HCC with an AUC of 0.79-0.81 and outperformed
current diagnostic criteria when used in conjunction to differentiate intrahepatic cholangiocarcinoma from
[92]
HCC . Radiomic features have also been applied to distinguish primary and metastatic liver tumors, with
one study reporting an accuracy of 83% in differentiating metastatic liver disease from primary tumors .
[93]
Differentiating neoplastic portal vein involvement and thrombosis from benign portal vein thrombosis is
also essential in determining the underlying HCC’s resectability. Canellas et al. were able to reliably
[94]
distinguish neoplastic and bland thrombi using radiomics signatures based on thrombus density values .
Moreover, microinvasion, encompassing infiltration of portal vein, hepatic vein, or bile duct, is considered a
[95]
marker of poor prognosis after hepatic resection and transplantation . Several studies have also
demonstrated promising results of select CT and U/S-based radiomics signatures in accurately predicting
the presence of microvascular tumor invasion [96-99] . For ICC, in a study of 203 cases, radiomics-based models
demonstrated the ability to predict futile resections with greater AUC (0.804) than traditional clinical risk
factor-based models (AUC: 0.590) .
[100]
Characterization of tumor biology
Traditional biopsies are restricted to the small sample of tumor that is biopsied, which may not be
representative of the tumor at large [101,102] . With radiomics-based tumor grading, the entirety of the tumor is
evaluated when being graded, thus providing physicians with greater certainty regarding tumor
characteristics . Applications of radiomics in this regard have demonstrated promise, with studies
[103]
reporting that radiomics signatures strongly correlate with pathological grade of HCCs, allowing for rapid,
[104]
noninvasive tumor grade determination . Subsequent research has since further correlated radiomic
signatures with tumor Ki-67 status, demonstrating the utility of radiomics in assessing HCC proliferation
indices entirely noninvasively .
[105]
Treatment selection
Liver transplantation and liver resection represent potentially curative options for early-stage HCC;
however, most patients present with later-stage HCC, rendering them ineligible for these options. As such,
physicians and patients must choose from a variety of locoregional therapies, including radiofrequency
ablation (RFA), transarterial chemoembolization (TACE), and transarterial radioembolization (TARE), all
of which have particular advantages and disadvantages [106,107] . In recent years, there have been investigational
efforts to harness radiomics to aid in this challenging clinical decision-making process. For instance, one
group employed a DL-based radiomics strategy to develop nomograms for predicting progression-free
survival among HCC patients undergoing liver resection or RFA (c-index = 0.726 for RFA, 0.741 for
[108]
resection) . To identify future candidates for RFA, another group calculated a radiomics signature based
on textural features extracted from patients with significantly longer progression-free survival after RFA
(P = 0.008) . Suh et al. demonstrated the utility of a radiomics-based signature as a feasible barometer for
[109]
stratifying and determining patient suitability for liver resection vs TACE . In unresectable cases,
[110]
radiomics has also demonstrated efficacy in predicting the development of progressive disease following
TACE [111,112] . In this non-surgical cohort of patients, radiomics analysis has also demonstrated the ability to
accurately predict tumor immunoscores and immune phenotypes [113,114] . This data has been validated to
[114]
tumor responsiveness to novel immunotherapy and can help make treatment decisions .
Prognostication
Multiple studies have reported the value of applying radiomics to predict outcomes in hepatobiliary
diseases. These studies demonstrated that the addition of radiomics to conventional clinical variables
increases the accuracy of predicting early recurrence of disease, disease-specific recurrence, and long-term
mortality as compared to clinical variables alone with AUCs ranging from 0.59 to 0.91 [107,115,116] . For