Detectability of Automated Tumor-feeder Detection Software Using Angio-computed Tomography in Transarterial Chemoembolization for Hepatocellular Carcinoma

Abstract

Purpose: To evaluate the performance of automated tumor-feeder detection software in transarterial chemoembolization for hepatocellular carcinoma using angio-computed tomography.

Material and Methods: This was a retrospective study of 107 hepatocellular carcinomas in 74 patients who underwent selective transarterial chemoembolization between June 2021 and December 2022. Identification of tumor-feeding arteries on computed tomography during hepatic angiography images acquired prior to chemoembolization with angio-computed tomography was evaluated in two independent methods: analysis by automated tumor-feeder detection software and interpretation by radiologists. The sensitivity and positive predictive value of both were calculated, and the sensitivity was compared with the McNemar test. Differences with p <0.05 were considered statistically significant.

Results: Transarterial chemoembolization was applied to 107 hepatocellular carcinoma tumors fed by 114 arteries. No significant difference was observed in sensitivity between the software and the interpretation of radiologists (90.4% vs. 95.6%, p = 0.15). The positive predictive value for the software was 90.4%; that for the interpretation of radiologists was 86.8%.

Conclusions: The accuracy of automated tumor-feeder detection software applied to angio-computed tomography was comparable to that of radiologists.