A Novel Application of Deep Learning (Convolutional Neural Network) for Traumatic Spinal Cord Injury Classification...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Traumatic spinal cord injury (TSCI) research requires reliable assessment tools.
• Electromyography (EMG) signals offer a potential method for monitoring TSCI effects.
• Non-human primate (NHP) models are crucial for pre-clinical studies of TSCI treatments.

Purpose of the Study:

• To develop and evaluate a novel classification system for TSCI using EMG signals in an NHP model.
• To compare the performance of a deep learning model (CNN) against a classical machine learning model (kNN).
• To establish an effective assessment tool for detecting TSCI and evaluating treatment efficacy.

Main Methods:

• Collected intramuscular EMG data from tail muscles of five Macaca fasicularis monkeys before and after induced spinal cord lesions.
• Developed a classification system based on a convolutional neural network (CNN) utilizing filtered, segmented EMG signals.
• Implemented a k-nearest neighbors (kNN) classifier using four hand-crafted EMG features for comparison.

Main Results:

• The CNN classifier achieved higher F-measures (89.8% left, 96.9% right) compared to the kNN classifier (89.7% left, 92.7% right).
• The CNN demonstrated superior performance in classifying pre- and post-lesion EMG data.
• CNN's ability to automatically learn high-level features from EMG segments proved advantageous.

Conclusions:

• The proposed CNN-based classification system shows high potential as an effective tool for TSCI assessment in NHPs.
• Deep learning techniques, specifically CNNs, offer a promising advancement over conventional methods for analyzing EMG signals in TSCI research.
• Further validation with larger subject numbers is recommended to confirm these findings.