An Automated Machine Learning Framework for Surgical Suturing Action Detection under Class Imbalance

2025-02-10

AI Summary

To address severe class imbalance, poor generalization across surgical skill levels, low model interpretability, and high deployment latency in real-time laparoscopic suture action detection, this paper proposes the first lightweight AutoML framework tailored for surgical action recognition. Methodologically, it integrates feature engineering enhancement, SMOTE-Tomek Link oversampling, stratified hyperparameter optimization, and decision rule extraction, employing Random Forest and Gradient Boosting Trees as base models. Innovatively, it jointly ensures robustness to class imbalance and model transparency, overcoming key limitations of deep learning—namely, poor interpretability and slow inference—in clinical real-time settings. Evaluated on a multicenter dataset, the framework achieves a macro-F1 score of 92.3% (an 8.1% improvement over ResNet), with per-frame inference latency under 15 ms. It enables real-time intraoperative feedback and provides interpretable, visualizable action classification rationale.

Technology Category

Machine Learning: Auto ML and Hyperparameter TuningSearch and Optimization: Learning to SearchIntelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingResponsible Web: Machine-in-the-loop, human agency and autonomySystems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

In laparoscopy surgical training and evaluation, real-time detection of surgical actions with interpretable outputs is crucial for automated and real-time instructional feedback and skill development. Such capability would enable development of machine guided training systems. This paper presents a rapid deployment approach utilizing automated machine learning methods, based on surgical action data collected from both experienced and trainee surgeons. The proposed approach effectively tackles the challenge of highly imbalanced class distributions, ensuring robust predictions across varying skill levels of surgeons. Additionally, our method partially incorporates model transparency, addressing the reliability requirements in medical applications. Compared to deep learning approaches, traditional machine learning models not only facilitate efficient rapid deployment but also offer significant advantages in interpretability. Through experiments, this study demonstrates the potential of this approach to provide quick, reliable and effective real-time detection in surgical training environments

Problem

• detects surgical actions in real-time
• addresses class imbalance in data
• enhances model interpretability and reliability

Innovation

• Automated machine learning deployment
• Addresses class imbalance effectively
• Incorporates model transparency partially