EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Automated transperineal ultrasound analysis using deep learning for pelvic floor dysfunction assessment after total hysterectomy

Yanqing Xu, Fan Yang, Fan Zhao, Runyan Ji

Frontiers in Oncology · 18.06.2026 · 16

DOI: 10.3389/fonc.2026.1730276 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Oncology, online erschienen am 18.06.2026. Verfasst von Yanqing Xu und 3 weiteren Autor:innen. Beteiligte Einrichtung: Ultrasonography Department, Affiliated Nantong Hospital 3 of Nantong University, Nantong Third People's Hospital, Nantong, China. Die Arbeit nennt 37 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Pelvic floor dysfunction (PFD) is a common functional disorder following total hysterectomy. Currently, the assessment mainly relies on manual measurements of transperineal ultrasound (TPUS) images, which suffers from low efficiency, subjectivity, and limited reproducibility.

Objective

To address these limitations, this work proposes an automated transperineal ultrasound analysis using deep learning for pelvic floor dysfunction assessment after total hysterectomy.

Methods

A labeled dataset including key anatomical landmarks such as the symphysis pubis, bladder neck, urethra, and puborectalis muscle was established, and a multi-scale shifted window Transformer was developed to achieve automatic segmentation and key point detection. Additionally, a geometric reasoning module was further designed to compute seven clinically relevant functional parameters, including bladder neck–symphysis distance, posterior urethrovesical angle, and urethral rotation angle.

Results

Experimental results demonstrated that the model achieved an average Dice coefficient of 88.67 ± 1.96% in segmentation, with key point localization errors controlled within 2 mm. The automatic measurement results are highly consistent with manual annotations, with Pearson correlation coefficients up to 0.92, and effectively distinguished functional differences among patients undergoing different surgical approaches.

Conclusion

The proposed method enables structured, automated, and objective TPUS image analysis, significantly reducing manual intervention. Although validated in patients with benign diseases, the approach is directly transferable to gynecologic oncology patients, who are at even higher risk of PFD due to more extensive surgery and adjuvant therapies. It provides a reliable tool for postoperative functional monitoring and therapeutic evaluation after total hysterectomy, and holds great potential for functional imaging-based rehabilitation assessment in post-hysterectomy patients.

Schlagwörter

Stichwörter der Autor:innen

deep learningpelvic floor dysfunction (PFD)total hysterectomytransformertransperineal ultrasound (TPUS)

Autor:innen und Institutionen

  • Ultrasonography Department, Affiliated Nantong Hospital 3 of Nantong University, Nantong Third People's Hospital, Nantong, China.
  • School of Electrical Engineering and Automation, Nantong University, Nantong, China.
  • Medical School of Nantong University, Nantong University, Nantong, China.

Zitieren

APA

Xu, Y., Yang, F., Zhao, F., & Ji, R. (2026). Automated transperineal ultrasound analysis using deep learning for pelvic floor dysfunction assessment after total hysterectomy. Frontiers in Oncology, 16. https://doi.org/10.3389/fonc.2026.1730276

Vancouver

Xu Y, Yang F, Zhao F, Ji R. Automated transperineal ultrasound analysis using deep learning for pelvic floor dysfunction assessment after total hysterectomy. Frontiers in Oncology. 2026;16. doi:10.3389/fonc.2026.1730276

BibTeX
@article{Xu2026,
  author = {Xu, Yanqing and Yang, Fan and Zhao, Fan and Ji, Runyan},
  title = {Automated transperineal ultrasound analysis using deep learning for pelvic floor dysfunction assessment after total hysterectomy},
  journal = {Frontiers in Oncology},
  year = {2026},
  volume = {16},
  doi = {10.3389/fonc.2026.1730276}
}
Kostenlos starten →