Artificial intelligence in the diagnosis of deep vein thrombosis: A scoping review
Clinical Snapshot
PICO Framework
| P — Population | Adult patients with suspected or confirmed deep vein thrombosis (DVT), primarily lower extremity |
| I — Intervention | Artificial intelligence (AI) and machine learning algorithms applied to medical imaging (CT, MRI, ultrasound) for DVT diagnosis |
| C — Comparator | Implicit comparator: conventional expert image interpretation or reference standard diagnosis (e.g., duplex ultrasound, venography, clinical diagnosis) |
| O — Outcomes | Diagnostic accuracy metrics including sensitivity, specificity, precision, and accuracy of AI models for DVT detection across imaging modalities |
Bottom Line
This scoping review maps the emerging landscape of AI-assisted DVT diagnosis across ultrasound, MRI, and CT imaging, identifying only 11 eligible studies published between 2021 and 2025. While some AI models — particularly MRI-based and multi-dataset approaches — report impressive diagnostic metrics (sensitivity and specificity approaching 95–97%), the evidence base is too small, heterogeneous, and methodologically variable to support clinical adoption. Ultrasound-based AI models show the widest performance range (sensitivity 68–100%), reflecting the diversity of architectures and validation approaches rather than consistent clinical utility. No pooled estimates or confidence intervals are available, and GRADE certainty assessment was not performed. The authors appropriately conclude that AI should supplement, not replace, expert radiological interpretation. For Australian clinicians, compression duplex ultrasound remains the standard of care for DVT diagnosis, and no TGA-approved AI tool currently exists for this indication. Future research must prioritise prospective multicenter validation, standardised reference standards, explainable AI frameworks, and health economic analysis before AI-assisted DVT diagnosis can be responsibly integrated into clinical workflows.
Key Findings
P Value: Not reported
Effect Size: Ultrasound-based AI models: sensitivity 68–100%, specificity 70–100%; MRI-based models: sensitivity and specificity 95–97%; CT-based model: sensitivity 83%; Multi-dataset models: sensitivity, specificity, and precision ≥96%
Primary Outcome: Diagnostic accuracy of AI algorithms for DVT detection across CT, MRI, and ultrasound imaging modalities
Nnt Or Sensitivity: Sensitivity range for US-based AI: 68–100% (wide variation indicating heterogeneous model performance); MRI-based AI sensitivity: 95–97%; CT-based AI sensitivity: 83%. No NNT calculable from available data. No AUC or ROC data reported in abstract.
Confidence Interval: Not reported — no pooled estimates or confidence intervals provided; scoping review design precludes quantitative synthesis
Clinical Application
Clinical translation remains premature. The 11 included studies represent early-stage development and single-centre validation of diverse AI architectures. No included AI tool appears to have achieved regulatory approval or prospective clinical validation at scale. Integration into existing PACS/RIS workflows, explainability requirements, and cost-effectiveness have not been systematically evaluated. Radiologist oversight remains essential given the wide performance variability. In Australia, compression duplex ultrasound is the primary diagnostic modality for suspected DVT, consistent with RACGP and vascular surgery guidelines. CT venography is used selectively. AI-assisted ultrasound interpretation could theoretically address workforce pressures in regional and rural settings where specialist sonographer access is limited. However, no TGA-approved AI diagnostic tool for DVT is currently listed on the ARTG as of the knowledge cutoff. The MBS does not currently include specific item numbers for AI-assisted DVT diagnosis. Australian clinicians should note that the evidence base (11 studies, no multicenter validation) is insufficient to support adoption outside research settings. The RACGP and relevant specialist colleges would require prospective Australian validation data before endorsing AI-assisted DVT diagnosis in primary or emergency care settings. Adult patients undergoing imaging evaluation for suspected DVT, particularly in settings with high imaging volume where AI-assisted triage or detection could augment radiologist workflow. Most applicable to lower extremity DVT evaluated by compression ultrasound or CT venography.
Abstract
Deep vein thrombosis (DVT) is the formation of thrombi in the deep venous system, most often in the lower extremities. Although usually not life-threatening, DVT requires timely diagnosis to prevent complications such as pulmonary embolism and post-thrombotic syndrome. The growing demand for image interpretation has generated interest in applying artificial intelligence (AI) to automated DVT detection. This scoping review analyzes the performance of artificial intelligence in diagnosing DVT using computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound (US). We conducted a search across seven databases from inception to May 2025 using terms related to deep vein thrombosis, artificial intelligence, and machine learning. Eligible studies were limited to those evaluating DVT diagnosis using CT, MRI, or ultrasound. Two independent reviewers selected eligible studies, and quality was assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2). Eleven studies published between 2021 and 2025 met the inclusion criteria. Some of the AI algorithms included RetinaNet, Deep R-Belief Neural Networks, and Sooty Tern Optimization. US-based models were the most studied algorithms, with sensitivities and specificities ranging from 68 to 100% and 70-100%, respectively. The MRI-based model achieved sensitivities, specificities, and accuracies of 95% to 97%. One CT-based model demonstrated a sensitivity of 83%. Studies evaluated across multiple imaging datasets showed high sensitivities, specificities, and precision of 96% or higher. Future research should prioritize multicenter validation and integration of clinical factors. In addition, explainable frameworks capable of integrating multiple imaging datasets must be developed with attention to workflow efficiency and cost-effectiveness to support clinical translation. The results indicate that AI is best situated as a supplementary tool rather than a replacement for expert interpretation in DVT diagnosis.
References
- 1.Chen, T., Hisham Shunmugam, R., Tan, S. Y. Y., Goh, S. S. L., Yan, Y. Y., & Ng, K. H. (2026). Artificial intelligence in the diagnosis of deep vein thrombosis: A scoping review. PLoS ONE. https://doi.org/10.1371/journal.pone.0351558
Related Research
Physics in medicine and biology
Anatomically adaptive feature-wise linear modulation for deep learning-based low-dose CT denoising
28 July 2026
Science advances
Interpretable dynamic quantitative vascular morphometry features using SHAP for anti-angiogenic therapy response prediction
27 July 2026
Pediatric surgery international
Pediatric colonic diverticulitis: clinical presentation, management, and review of the literature
21 July 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service