EUVIMEDObjevte EUVIMED
Open access · CC BY-NC-ND 4.0

Multimodal Deep Learning for Predicting Deep Vein Thrombosis Risk After Total Knee Arthroplasty: A Clinical Study

Kejia Zhu, Hang Li, Hui Zhang, Biao Wang, Bin Shen iD, Yong Nie iD

Orthopaedic Surgery · 2026-08 · 18 (8) · 1574-1580

DOI: 10.1111/os.70335 ↗

At a glance

Journal article in Orthopaedic Surgery, published online on 2026-06-16. Written by Kejia Zhu and 5 co-authors. Institution: Department of Orthopedic Surgery and Orthopedic Research Institute, West China Hospital, Sichuan University, Chengdu, China. Major MEDLINE topics: Arthroplasty, Replacement, Knee, Venous Thrombosis, Deep Learning, Postoperative Complications. 16 references. The full text is freely available (open access, CC BY-NC-ND 4.0). Funded by: National Key Research and Development Program of China, Sichuan Provincial International Science and Technology Innovation Cooperation, 1·3·5 Projects for Artificial Intelligence West China Hospital.

AbstraktSubjectsBibliografické údajeCiteRelated studies

Abstract (original language)

Background

Deep vein thrombosis (DVT) following total knee arthroplasty (TKA) remains a major postoperative complication. Conventional risk assessment tools such as the Caprini score show limited predictive performance (AUC 0.65–0.72). We developed a multimodal deep learning framework that integrates early postoperative radiographs with clinical data to improve DVT risk stratification.

Methods

This single‐center retrospective cohort study analyzed 1200 patients undergoing primary unilateral TKA (DVT incidence: 18.0%) from January 2018 to June 2023. A dual‐branch architecture was implemented: ViT‐B/16 or ResNet50 processed standardized radiographs (metal artifact suppression; 512 × 512 resolution), while Clinical‐BERT encoded structured electronic medical record (EMR) representations. A dynamic attention fusion module generated sample‐specific modality weights (α) for feature fusion (α·v_img + (1 − α)·v_txt). Performance was evaluated on a stratified test set ( n = 180) against the Caprini score, unimodal models, and a nonattention concatenation fusion baseline. AUCs were compared using the DeLong test, and sensitivity/specificity differences were assessed using the McNemar test.

Results

The multimodal model achieved an AUC of 0.89 (95% CI: 0.85–0.92; DeLong p < 0.001 vs. all comparators), sensitivity 83% (76%–89%), specificity 88% (82%–93%), and F1‐score 0.79. It identified 91.7% (11/12) of DVT cases labeled low‐risk by Caprini. The learned attention coefficient α suggested modality reliance patterns associated with specific findings (e.g., higher α in cases with prosthesis–bone lucency > 2 mm; higher text contribution in cases with D‐dimer > 1.0 mg/L), but these associations do not establish causality. Anatomical heatmaps consistently localized high‐risk regions, particularly the popliteal vein projection area and tibial interface.

Conclusion

This attention‐based multimodal model integrates biomechanical (radiographic) and biochemical/clinical (EMR‐derived) signals to improve early postoperative DVT risk prediction while providing interpretable anatomical cues.

Subjects

MeSH terms (MEDLINE)

Arthroplasty, Replacement, KneeVenous ThrombosisDeep LearningPostoperative ComplicationsHumansRetrospective StudiesRisk AssessmentAgedFemaleMaleMiddle Aged

Author keywords

attention mechanismdeep vein thrombosismultimodal deep learningrisk predictiontotal knee arthroplasty

Authors and institutions

  • Department of Orthopedic Surgery and Orthopedic Research Institute, West China Hospital, Sichuan University, Chengdu, China.
  • Rehabilitation Medicine Center and Institute of Rehabilitation Medicine, West China Hospital, Sichuan University, Chengdu, China.
  • West China Biomedical Big Data Center, West China Hospital, Sichuan University, Chengdu, China.

ORCID profiles: Bin Shen, Yong Nie

Funding

  • National Key Research and Development Program of China
  • Sichuan Provincial International Science and Technology Innovation Cooperation
  • 1·3·5 Projects for Artificial Intelligence West China Hospital
  • 1·3·5 Project for Disciplines of Excellence-Clinical Research Fund, West China Hospital

Cite

APA

Zhu, K., Li, H., Zhang, H., Wang, B., Shen, B., & Nie, Y. (2026). Multimodal Deep Learning for Predicting Deep Vein Thrombosis Risk After Total Knee Arthroplasty: A Clinical Study. Orthopaedic Surgery, 18(8), 1574-1580. https://doi.org/10.1111/os.70335

Vancouver

Zhu K, Li H, Zhang H, Wang B, Shen B, Nie Y. Multimodal Deep Learning for Predicting Deep Vein Thrombosis Risk After Total Knee Arthroplasty: A Clinical Study. Orthopaedic Surgery. 2026;18(8):1574-1580. doi:10.1111/os.70335

BibTeX
@article{Zhu2026,
  author = {Zhu, Kejia and Li, Hang and Zhang, Hui and Wang, Biao and Shen, Bin and Nie, Yong},
  title = {Multimodal Deep Learning for Predicting Deep Vein Thrombosis Risk After Total Knee Arthroplasty: A Clinical Study},
  journal = {Orthopaedic Surgery},
  year = {2026},
  volume = {18},
  number = {8},
  pages = {1574--1580},
  doi = {10.1111/os.70335}
}
Start for free →