CT-Based Nested Habitats Analysis for Early Recurrence Prediction and Risk Stratification in Hepatocellular...

The purpose of this study was to develop and validate a computed tomography (CT)-based nested habitats analysis for identifying aggressive tumor subregions and predicting early recurrence in patients with hepatocellular carcinoma (HCC).

Patients from three institutions were allocated to a training cohort (n = 372) and an internal validation cohort (n = 160) at a 7:3 ratio. An external validation cohort (n = 169) from a fourth institution was included. Venous-phase CT images underwent nested habitats analysis to locate aggressive subregions. First, a support vector machine classified tumors on the basis of global radiomic features. Then, local features were extracted to construct probability maps, from which aggressive micro-regions were identified using k-means clustering. Features from the highest-risk micro-regions were integrated to generate a nested habitats score. Model performance was evaluated with the area under the curve (AUC) and Kaplan-Meier survival analysis.

The nested habitats score demonstrated strong predictive ability for early recurrence, achieving AUCs of 0.832 (95% CI 0.778-0.885) in the training cohort, 0.896 (95% CI 0.833-0.959) in the internal validation cohort, and 0.833 (95% CI 0.762-0.905) in the external validation cohort. In multivariable Cox regression, the nested habitats score remained an independent predictor of recurrence-free survival (RFS) (P < 0.05), along with alkaline phosphatase, macrotrabecular-massive HCC, sex, and intratumoral tertiary lymphoid structures. Kaplan-Meier analysis further confirmed significantly shorter RFS among patients with high nested habitats scores or high nomogram-predicted risk (P < 0.05).

The CT-based nested habitats analysis effectively captures intratumoral heterogeneity and accurately predicts early recurrence in HCC. This technique enables precise postoperative risk stratification.