Single-Subject [18F]FDG PET Metabolic Connectomics: Static and Dynamic Methods, Validation, and Clinical Readiness—A Systematic Review

Single-subject [18F]FDG PET networks are increasingly proposed as imaging biomarkers, but their edges and validation settings differ substantially. We systematically reviewed human studies generating participant-specific metabolic networks, separating spatial similarity, within-scan dynamics, across-visit covariance and learned representations. Scopus, PubMed/MEDLINE and Web of Science searches on 3 September 2026 yielded 1555 records and 840 after deduplication. Fifty-eight database reports and one additional preprint underwent full-text assessment; 52 reports were retained. Screening and extraction were machine-assisted, followed by competency-based verification by two coauthors; independent duplicate screening and a random exclusion audit were not documented. The revision re-examined 48 full texts, with limited re-verification for four previously assessed reports. The corpus contained 34 static, 16 within-scan dynamic/fPET, one across-visit longitudinal and one hybrid/model-derived report. Six source-reuse families encompassed 15 reports, without permitting a definitive count of independent cohorts. Selected internally validated comparisons favored graph features, whereas others showed classifier-dependent advantages or inferior performance; technical reproducibility was distinct from clinical validation. Static edges plausibly describe regional metabolic resemblance but do not establish direct communication. Clinical translation requires locked external validation, physiological and acquisition control, transparent provenance, and demonstrated improvement beyond conventional-PET and clinical models.