Emerging Frontiers in the Diagnosis of Deep Vein Thrombosis: From D-dimer to Artificial Intelligence-driven Imaging

Abstract

Deep vein thrombosis (DVT) remains a significant contributor to global morbidity and mortality, particularly
because of its association with life-threatening complications, such as pulmonary embolism and post-thrombotic
syndrome. This review critically explores advancements in DVT diagnostic strategies, including traditional
approaches, emerging imaging modalities, biomarker integration, and artificial intelligence (AI)-driven innovations.
Conventional diagnostic pathways relying on clinical scores (e.g., Wells and Geneva), D-dimer assays, and duplex
ultrasonography, though widely used, exhibit limitations in terms of sensitivity, specificity, and adaptability
across special populations such as pregnant women and cancer patients. New imaging modalities, including
magnetic resonance venography, computed tomography venography, intravascular ultrasound, elastography,
and photoacoustic imaging, offer enhanced anatomical and functional insights, addressing gaps in thrombus
age characterization and venous outflow obstruction. AI tools leveraging machine learning, natural language
processing, and electronic health records are revolutionizing risk stratification, imaging interpretation, and decision
support. These technologies aim to reduce diagnostic uncertainty, minimize unnecessary interventions, and enable
personalized care for patients. Challenges persist in standardizing protocols, ensuring ethical AI deployment,
and validating novel biomarkers, such as urinary proteomics and thrombin generation profiles. Special emphasis
is placed on tailoring diagnostic algorithms for vulnerable subgroups and optimizing the timing of therapeutic
interventions. This review highlights the clinical implications of these advancements and underscores the need for
future translational research to bridge innovation and bedside applications.