Predicting early recurrence of hepatocellular carcinoma with texture analysis of preoperative MRI: a radiomics study

Abstract

Aim: To investigate the feasibility of using texture analysis in preoperative magnetic resonance imaging (MRI) to predict early recurrence (ER) in hepatocellular carcinoma (HCC) post-curative surgery.

Material and methods: Institutional review board was obtained. A retrospective review of all patients who underwent hepatectomy between 1 January 2007 and 31 December 2015 was performed. Inclusion criteria included preoperative MRI, tumour size ≥1 cm, new cases of HCC. Exclusion criteria included loss to follow-up, ruptured HCCs, movement artefacts, and previous hepatectomy or interval adjuvant therapy. Patients were divided into ER and late or no recurrence (LNR) groups. ER was defined as new foci of HCC within 730 days of curative surgery. Radiomics feature extraction was performed on T2, diffusion-weighted imaging (DWI), T1 arterial, and T1 portovenous acquisitions on MATLAB (Mathworks, Matick, MA, USA). The MaZda software was used to analyse 290 texture parameters and PRTools was used for feature selection.

Results: Fifty patients (43 male, mean age 67 years) were divided into ER (n=20) and LNR (n=30) groups. Serum alpha-fetoprotein level (p=0.026), serum ɣ-glutamyltranspeptidase (p=0.014), Child-Pugh score (p=0.02) and the presence of vascular invasion (gross and/or microvascular, p=0.025) were found to be statistically significant different between the two groups. Parameters S(4,0)SumVarnc, S(0,3)SumOfSqs, and S(1,1)DifVarnc of the equilibrium phase were most accurate, achieving 84%, 82%, and 78% accuracy, respectively.

Conclusion: Texture analysis of preoperative MRI has the potential to predict ER of HCC with up to 84% accuracy using an appropriate, single texture analysis parameter. Future studies are needed to validate these findings.