Automated Microsurgical Tool Categorization using a Surface-Based EMG System.

Area of Science:

• Biomedical Engineering
• Surgical Technology
• Machine Learning in Medicine

Background:

• Surgical tool selection and usage critically impact surgical outcomes.
• Current video-based detection methods for surgical tools have limitations, such as masking information and failing to capture haptic feedback.
• Extracting localized information beyond visual data is valuable for surgical assessment.

Purpose of the Study:

• To evaluate the adoption of surface-based electromyography (sEMG) signals for detecting micro-surgical tools.
• To develop and test a machine learning model for classifying micro-surgical tools based on sEMG signals.
• To explore sEMG as a method for characterizing tool handling and assessing surgeon skills.

Main Methods:

• A two-channel sEMG acquisition system was designed and utilized.
• A staged, supervised dataset of micro-surgical tool usage was created.
• Five hand-crafted features from each sEMG channel were extracted and used to train an Artificial Neural Network (ANN) model.

Main Results:

• The ANN model achieved 97.43% accuracy in classifying five distinct micro-surgical tools.
• Classification was based on the press-and-release action detected through sEMG signals.
• The study demonstrated the feasibility of using sEMG for tool detection.

Conclusions:

• sEMG signals offer a viable method for detecting and classifying micro-surgical tools.
• This approach provides a novel way to assess subtle tool handling characteristics, complementing video-based assessments.
• The developed system represents a step towards advanced tool characterization and objective surgeon skill evaluation.