PET/CT-Based Intratumoral Heterogeneity and Multimodal Imaging for Tumor Diagnosis and Prognosis

Background \& Rationale: Intratumoral heterogeneity (ITH) is a recognized hallmark and critical driver of therapeutic resistance, rapid disease progression, and poor survival across solid malignancies, notably advanced non-small-cell lung cancer (NSCLC). While tissue biopsies provide histopathologic confirmation, they are invasive and subject to severe spatial sampling bias. Standard whole-tumor radiomics from single-modality CT often masks localized subregional ecological niches. 18F-FDG PET/CT unites glucose metabolic activity with anatomical resolution, offering an ideal non-invasive tool to map whole-tumor spatial and functional heterogeneity in vivo. Methods \& Design: In this multicenter, bidirectional (retrospective and prospective) observational cohort study, 900 advanced NSCLC patients receiving first-line systemic therapies (chemotherapy, anti-PD-1 immunotherapy, and EGFR-TKIs) were enrolled. Baseline 18F-FDG PET/CT scans obtained within 2 weeks prior to therapy were partitioned into distinct biological habitats using entropy-intensity clustering. IBSI-compliant radiomic features were extracted to develop treatment-specific imaging intratumoral heterogeneity (IITH) signatures. Diagnostic accuracy for staging/metastases and prognostic performance for 6-month progression status, progression-free survival (PFS), and overall survival (OS) were comprehensively validated across internal, external multicenter, and independent prospective cohorts, with biological explainability benchmarked against digital histopathology (HoVer-Net nuclear morphometrics) and transcriptomics.