Development and validation of a machine learning-based predictive model for early outcomes following combined suction-assisted lipectomy and lymphovenous anastomosis in breast cancer-related lymphedema: a retrospective cohort study - Report - MDSpire

Development and validation of a machine learning-based predictive model for early outcomes following combined suction-assisted lipectomy and lymphovenous anastomosis in breast cancer-related lymphedema: a retrospective cohort study

  • By

  • Yonghao Cui

  • Hao Dong

  • Zixuan Yao

  • Shuai Pang

  • Yuguang Sun

  • Song Xia

  • Wenbin Shen

  • May 7, 2026

Share

Clinical Report: Machine Learning for Predicting Postoperative Outcomes in BCRL

Overview

This study developed a machine learning model to predict early postoperative outcomes in patients with breast cancer-related lymphedema (BCRL) undergoing combined suction-assisted lipectomy and lymphovenous anastomosis. The support vector machine (SVM) model demonstrated high sensitivity and specificity, identifying key predictors for patient outcomes.

Background

Breast cancer-related lymphedema (BCRL) is a prevalent complication that significantly affects patients' quality of life. Current postoperative risk stratification tools are inadequate, leading to variability in patient outcomes. This study addresses the need for predictive models to enhance early identification of high-risk patients following surgical interventions.

Data Highlights

VariableValue
Patients Enrolled300
Satisfactory Outcomes at 6 Months72.3%
Optimal AUC (SVM)0.891
Sensitivity (SVM)90.8%
Specificity (SVM)62.5%

Key Findings

  • The SVM model achieved an AUC of 0.891, indicating strong predictive performance.
  • Postoperative excess limb volume was identified as the strongest predictor of outcomes.
  • Three stable predictors were identified: excess limb volume, disease duration, and disease severity grade.
  • The model demonstrated superior generalization stability compared to artificial neural networks.
  • Decision curve analysis indicated net clinical benefit for the SVM model within a threshold range of 0.1–0.6.

Clinical Implications

The SVM model can assist clinicians in identifying patients at high risk for poor postoperative outcomes, allowing for tailored monitoring and interventions. This predictive framework may enhance rehabilitation protocols and improve patient care in BCRL management.

Conclusion

The development of a machine learning-driven predictive model represents a significant advancement in postoperative care for BCRL patients. By integrating SHAP analysis and a web-based tool, this model facilitates early risk stratification and targeted management strategies.

References

  1. Techniques in Coloproctology, 2025 -- Utilizing Deep Learning Neural Networks to Forecast Postoperative Complications
  2. Obesity Surgery, 2022 -- The Role of Artificial Intelligence in Bariatric Surgery
  3. Computational Framework, 2025 -- Differentiating Vascular Ligation and Lymphadenectomy in Oncological Sigmoidectomy
  4. Frontiers in Surgery, 2026 -- Perioperative machine learning models with SHAP interpretation
  5. Lymphology, 2023 -- The Diagnosis and Treatment of Peripheral Lymphedema: 2023 Consensus Document
  6. ScienceDirect, 2024 -- The efficacy of immediate lymphatic reconstruction after axillary lymph node dissection
  7. Frontiers, 2025 -- Machine learning-based personalized risk prediction model for breast cancer-related lymphedema
  8. THE DIAGNOSIS AND TREATMENT OF PERIPHERAL LYMPHEDEMA: 2023 CONSENSUS DOCUMENT OF THE INTERNATIONAL SOCIETY OF LYMPHOLOGY | Lymphology
  9. The efficacy of immediate lymphatic reconstruction after axillary lymph node dissection – A meta-analysis - ScienceDirect
  10. Frontiers | Machine learning-based personalized risk prediction model for breast cancer-related lymphedema after surgery

Original Source(s)

Development and validation of a machine learning-based predictive model for early outcomes following combined suction-assisted lipectomy and lymphovenous anastomosis in breast cancer-related lymphedema: a retrospective cohort study

  • by Yonghao Cui, Hao Dong, Zixuan Yao, Shuai Pang, Yuguang Sun, Song Xia, Wenbin Shen

  • May 7, 2026

Related Content

One Pathway for Esophageal Cancer?

Molecular profiling reveals precursor signals in BE-negative esophageal cancer

  • May 7, 2026

  • 2 min
Plasma miR-371a-3p Predicts Recurrence in Stage I Testicular Cancer: Interim Results From the CLIMATE Study

In the management of stage I testicular cancer, a persistent clinical dilemma is the identification of patients who truly need adjuvant treatment after orchiectomy rather than active surveillance alon...

‘Promising’ Signal Only: ctDNA in Early Breast Cancer Not Yet Ready for Clinical Use

Despite compelling prognostic associations across multiple retrospective data sets, no interventional trial has demonstrated that acting on circulating tumor DNA testing results in early breast cancer...