American Association of Plastic Surgeons

AAPS Home AAPS Home Past & Future Meetings Past & Future Meetings
Facebook   Instagram   Twitter   YouTube   LinkedIn

Back to 2025 Abstracts


Machine Learning Approach To Predict Pain Outcomes Following Primary And Secondary Targeted Muscle Reinnervation In Amputees
Floris V. Raasveld, MD1, Zihe Zhang, PhD2, Johnston Johnston, MD, PhD3, Anna Luan, MD1, Arya Rao, BSc1, William Renthal, MD, PhD3, Clifford Woolf, PhD1, Ian L. Valerio, MD, MS, MBA, FACS1, Kyle Eberlin, MD1.
1Massachusetts General Hospital, Cambridge, MA, USA, 2Boston Children's Hospital, Boston, MA, USA, 3Brigham and Women's Hospital, Cambridge, MA, USA.

Purpose: Targeted Muscle Reinnervation (TMR) can effectively treat and prevent neuropathic pain (NP) in amputees, but its success varies among patients. This study developed a Machine Learning (ML) model to predict NP mitigation outcomes from Primary and Secondary TMR based on patient characteristics.
Methods: Patients undergoing Primary or Secondary TMR at a tertiary care center from 2018 to 2024 were included if they had over six months of follow-up. Exclusion criteria included being under 18, undergoing minor or bilateral amputation, or lacking pain data. Sustained NP mitigation was defined as a pain score of ≤3 for ≥3 months. Data on demographics, comorbidities, and surgical factors were collected through chart review. Bayesian and nonparametric modeling with SHAP (Shapley Additive exPlanations) feature ablation built the prediction model, assessed via Relevance Vector Machine (RVM) accuracy.
Results: A total of 77 Primary and 101 Secondary TMR patients were included (median follow-up: 2.0 years). The model predicted sustained NP mitigation for 55.8% of Primary and 63.4% of Secondary TMR patients. RVM training scores were 0.90 (Primary) and 0.83 (Secondary), with test scores of 0.77 and 0.75, indicating strong predictive accuracy. Smoking, psychiatric comorbidities, level of amputation, preoperative pain and pain medication use had high impacts on model output.
Discussion: This RVM model accurately predicts NP mitigation in TMR patients, aiding preoperative decision-making. Future research aims to enhance its predictive capabilities.

Back to 2025 Abstracts