# Preoperative contrast-enhanced CT prediction of distinct vascular patterns in solitary early-stage hepatocellular carcinoma and its prognostic value

## TL;DR

This study uses CT scans to predict vascular patterns in early liver cancer, helping identify patients at higher risk of cancer recurrence.

## Contribution

A combined model using clinical and CT features effectively identifies vascular patterns and predicts recurrence in early-stage hepatocellular carcinoma.

## Key findings

- The combined model achieved an AUC of 0.784 in training and 0.794 in external validation for predicting V/M+ status.
- A V/M+ score ≥34 and tumor size ≥60 mm were significant predictors of HCC recurrence.
- The model helps stratify patients into high- and low-risk groups for 2-year recurrence-free survival.

## Abstract

To investigate the value of qualitative and quantitative contrast-enhanced CT (CECT) features for noninvasive identification of two distinct vascular patterns, vessels that encapsulate tumor clusters (VETC) and/or microvascular invasion (MVI), in solitary early-stage (BCLC 0-A) hepatocellular carcinoma (HCC) and assess their prognostic implications.

We retrospectively included 347 patients with solitary early-stage HCC who underwent preoperative CECT and subsequent resection at two centers. Patients were divided into V/M+ (MVI and/or VETC positive, n = 174) and VM− (both MVI and VETC negative, n = 173) groups based on histopathology. Four predictive models (clinical, CT quantitative, CT qualitative, and combined) integrating clinical and CECT features were developed and validated for identifying V/M+ status. The optimal model was further applied to predict 2-year recurrence-free survival (RFS). Sensitivity analysis was performed using propensity score matching (PSM). Models’ performance was evaluated and compared using AUC analyses and DeLong tests.

The combined model [serum AFP ≥ 200 ng/mL, non-smooth tumor margin, internal arteries, and lower tumor-to-liver density ratio in the portal venous phase (P-TLR)] achieved optimal predictive performance for V/M + HCC, with training AUC of 0.784 and 0.782 pre- and post-PSM, and external validating AUC of 0.794. A derived V/M+ score stratified patients, with higher scores associated with significantly shorter 2-year RFS. V/M+ score ≥ 34 and tumor size ≥ 60 mm were significant predictors of HCC recurrence (p < 0.05).

The combined model integrating clinical and CECT-based features, enables non-invasive assessment of V/M status in early-stage solitary HCC and effectively stratifies patients according to recurrence risk.

Specific CT-based qualitative and quantitative features are associated with a distinct vascular pattern of BCLC stage 0-A HCC. The developed combined model and derived V/M+ score offer a reliable tool for clinicians to predict V/M + HCC and patients’ 2-year RFS.

Specific CECT-based qualitative and quantitative features are associated with V/M + HCC at the BCLC stage 0-A.The developed combined model offers a reliable tool for clinicians to identify V/M + HCC.The derived V/M+ score helps stratify HCC patients into high- and low-risk groups for 2-year RFS, facilitating personalized management of HCC.

Specific CECT-based qualitative and quantitative features are associated with V/M + HCC at the BCLC stage 0-A.

The developed combined model offers a reliable tool for clinicians to identify V/M + HCC.

The derived V/M+ score helps stratify HCC patients into high- and low-risk groups for 2-year RFS, facilitating personalized management of HCC.