Diagnostic efficacy of dynamic liver imaging using qualitative diagnostic algorithm versus LI-RADS v2018 lexicon for atypical versus classical HCC lesions: A decade of experience from a tertiary liver institute

Abstract

ObjectiveTo analyze and evaluate the diagnostic performance of conventional diagnostic (qualitative) imaging features versus LI-RADSv2018 lexicon for indeterminate and atypical Hepatocellular carcinoma (HCC) on dynamic liver imaging with reference to histopathology. Patients and methodsThis retrospective study (June 2009–June 2019) evaluated the performance characteristics of conventional imaging findings, versus the Liver Imaging Reporting and Data System (LIRADS) v2018, for interpretation of indeterminate and atypical HCC, in patients who underwent subsequent histopathological analysis (gold standard). A total of 100,457 dynamic hepatobiliary CT and MR examinations were performed over ten years at our institute. Using current international imaging guidelines, 3218 patients were found to have suspected liver cancer lesions on imaging. Classical enhancement pattern of typical HCC was seen in 2916 of these patients. These patients did not require further biopsy. We enrolled, the remaining (n = 302) patients, who underwent biopsy, into our study group. Two radiologists, blinded to pathology findings, reviewed and classified these lesions, in consensus, according to LI-RADS® lexicon and as per ‘conventional’ (Indeterminate, Atypical HCC, Classical HCC, other malignancies) imaging. The histopathology diagnosis was considered as the final diagnosis. Alpha feto protein (AFP) levels amongst various subgroups were compared. Statistical analysis was performed to calculate the efficacy of LIRADS versus qualitative imaging parameters in comparison with histopathology. ResultsA total of n = 302 patients, [89 % men (n = 269), mean age 57.08 ± 12.43 years] underwent biopsy for suspected liver lesions. Qualitative imaging had 92.3 % (CI 88.53–94.91) sensitivity, 41.4 % (CI 25.51–59.26) specificity, positive predictive value (PPV) of 93.7 % (CI 90.11–96.02), negative predictive value (NPV) of 36.4 % (CI 22.19–53.38), positive likelihood ratio (PLHR) of 1.575 (CI 1.40–1.77) and negative likelihood ratio of (NLHR) 0.19 (CI 0.13-0.26). It correctly classified 87.4 % of lesions diagnosed on pathology. In comparison, LI-RADS was found to have 92 % sensitivity, 55.5 % specificity, 97 % PPV, 30.3 %, NPV, PLHR 2.068 (CI 1.62–2.64), NLHR 0.15 (CI 0.11-0.18) and 89.7 % diagnostic accuracy. A total of 38 patients (17 false negative, 21 false positive lesions) had discordant diagnoses on imaging versus histopathology. The kappa agreement between LIRADs and qualitative Imaging was found to be 0.77 ± .07 (p < 0.001). LIRADS and qualitative imaging collectively had 97 % sensitivity, 30 % specificity, 91.9 % PPV, 55.6 % NPV, PLHR of 1.39 (CI 1.27–1.51) and NLHR of 0.09 (0.048-0.19) which was better than, either reporting system, independently. ConclusionIt was observed that the LI-RADS v2018 lexicon with qualitative imaging as a combination technique added extra value in interpretation of atypical HCC or indeterminate lesions on dynamic CT and MRI compared to either as ‘stand- alone’ reporting systems.