Macroscopic fractal dynamics characterize the "physical-metabolic" dual barriers and systemic immune exhaustion associated with primary resistance to immunotherapy in liver metastases.
Liver metastases are associated with systemic immune tolerance and primary resistance to immune checkpoint inhibitors (ICIs) by establishing complex physical and metabolic barriers within the tumor immune microenvironment (TIME). We developed a non-invasive macroscopic fractal dynamics framework to map these microenvironmental barriers across scales, aiming to predict ICI efficacy in colorectal cancer liver metastases (CRLM) and lung squamous cell carcinoma (SCC).
This single-center, retrospective, proof-of-concept cohort study consecutively enrolled 472 patients with CRLM or SCC liver metastases. Patients were divided into a training cohort (n=400, 2019-2024) and an independent validation cohort (n=72, 2025). Vascular fractal acceleration (Afd ) and metabolic fractal dimension (Df ) were extracted from contrast-enhanced magnetic resonance imaging (CE-MRI) and 18F-FDG PET, respectively. To eliminate baseline histological confounding, macroscopic fractal probes were Z-score normalized strictly within their respective histological cohorts. Cross-scale validation utilized digital pathology and platelet-poor plasma (PPP) cytokine profiling. An extreme gradient boosting (XGBoost) model was explicitly trained to predict a composite "High TIME Barrier" phenotype-defined by restricted CD8+ infiltration and low PD-L1 expression-to generate the Immuno-Radiomics Joint Score (IRJS). An exploratory survival analysis was subsequently conducted to evaluate its association with progression-free survival (PFS) among the 185 patients receiving ICI therapy. We evaluated early dynamic drift (ΔAfd ) at week 3 for its utility in tracking physical barrier remodeling.
CRLM and SCC displayed distinct fractal trajectories indicative of metabolic and physical barriers, respectively. High Afd correlated with dense fibrovascular stroma and severe spatial exclusion of CD8+ T cells. High Df corresponded to severe hypoxia, CD163-enriched macrophage infiltration, and systemic immune exhaustion, characterized by elevated circulating TGF-β and exhausted IFN-γ. The IRJS demonstrated strong diagnostic performance for the High TIME barrier phenotype (temporal validation AUC: 0.912). In the ICI sub-cohort, multivariable Cox regression confirmed that an increase in the continuous baseline IRJS was a robust, independent risk factor associated with primary ICI resistance and shorter PFS.
Macroscopic fractal dynamics offer a non-invasive, cross-scale method to evaluate the "physical-metabolic" dual microenvironmental barriers in liver metastases. The combined IRJS and dynamic ΔAfd tracking system show potential as exploratory, non-invasive surrogates to identify the systemic immune exhaustion phenotype. Pending external multi-center validation, these tools may generate hypotheses for associating macroscopic spatial barriers with primary ICI resistance and informing multidisciplinary interventions.
This single-center, retrospective, proof-of-concept cohort study consecutively enrolled 472 patients with CRLM or SCC liver metastases. Patients were divided into a training cohort (n=400, 2019-2024) and an independent validation cohort (n=72, 2025). Vascular fractal acceleration (Afd ) and metabolic fractal dimension (Df ) were extracted from contrast-enhanced magnetic resonance imaging (CE-MRI) and 18F-FDG PET, respectively. To eliminate baseline histological confounding, macroscopic fractal probes were Z-score normalized strictly within their respective histological cohorts. Cross-scale validation utilized digital pathology and platelet-poor plasma (PPP) cytokine profiling. An extreme gradient boosting (XGBoost) model was explicitly trained to predict a composite "High TIME Barrier" phenotype-defined by restricted CD8+ infiltration and low PD-L1 expression-to generate the Immuno-Radiomics Joint Score (IRJS). An exploratory survival analysis was subsequently conducted to evaluate its association with progression-free survival (PFS) among the 185 patients receiving ICI therapy. We evaluated early dynamic drift (ΔAfd ) at week 3 for its utility in tracking physical barrier remodeling.
CRLM and SCC displayed distinct fractal trajectories indicative of metabolic and physical barriers, respectively. High Afd correlated with dense fibrovascular stroma and severe spatial exclusion of CD8+ T cells. High Df corresponded to severe hypoxia, CD163-enriched macrophage infiltration, and systemic immune exhaustion, characterized by elevated circulating TGF-β and exhausted IFN-γ. The IRJS demonstrated strong diagnostic performance for the High TIME barrier phenotype (temporal validation AUC: 0.912). In the ICI sub-cohort, multivariable Cox regression confirmed that an increase in the continuous baseline IRJS was a robust, independent risk factor associated with primary ICI resistance and shorter PFS.
Macroscopic fractal dynamics offer a non-invasive, cross-scale method to evaluate the "physical-metabolic" dual microenvironmental barriers in liver metastases. The combined IRJS and dynamic ΔAfd tracking system show potential as exploratory, non-invasive surrogates to identify the systemic immune exhaustion phenotype. Pending external multi-center validation, these tools may generate hypotheses for associating macroscopic spatial barriers with primary ICI resistance and informing multidisciplinary interventions.