Artificial Intelligence and Multi-Omics Approaches in the Precision Management of Pulmonary Hypertension: From Early Diagnosis to Therapeutic Stratification

Figures

• 

  Figure 1. Established clinical pathway and investigational AI support across the pulmonary hypertension patient journey. Solid blue boxes and arrows denote guideline-directed or reference-standard clinical steps, whereas dashed orange boxes denote investigational AI support. Right-heart catheterisation (RHC) at an expert PH centre remains the reference standard for haemodynamic confirmation and classification. AI evidence maturity is shown qualitatively at each stage and does not imply clinical approval or readiness for autonomous use. Clinician oversight applies throughout. This figure was created with the assistance of artificial intelligence and subsequently reviewed and curated by the authors.

• 

  Figure 2. Multimodal AI architecture for precision management of pulmonary hypertension. The schematic shows how heterogeneous clinical, physiological, imaging, and molecular inputs are harmonised and fused into interpretable outputs for diagnosis, phenotyping, prognosis, and therapeutic stratification, with uncertainty estimates and clinician oversight. This figure was created with the assistance of artificial intelligence and subsequently reviewed and curated by the authors. The figure operationalises the molecular pathophysiology-to-therapeutic stratification sequence by linking cross-modal phenotypes to candidate biological endotypes, prognosis, and treatment-response support. This schematic represents a proposed translational architecture; it does not imply that a clinically validated PH model currently integrates all depicted molecular, imaging, haemodynamic, and clinical layers.

• 

  Figure 3. Molecular-to-precision-therapy framework in pulmonary hypertension. Genetic and epigenetic susceptibility converges on dysregulated vascular, immune, and metabolic pathways that generate measurable multi-omics signatures. AI can integrate these molecular data with imaging and haemodynamics to define candidate biological endotypes and estimate prognosis or treatment response. Translation to precision therapy requires independent replication, mechanistic coherence, temporal stability, calibration, and prospective demonstration of clinical utility and treatment-effect modification; candidate endotypes are therefore not current prescribing rules. This figure was created with the assistance of artificial intelligence and subsequently reviewed and curated by the authors. The molecular evidence underpinning this framework is strongest in PAH; extension to other PH groups remains investigational and requires disease-specific validation.