Motor Evoked Potential-Based Spinal Cord Function Assessment in Intraoperative Monitoring Using Machine Learning
Background:
Manual muscle test (MMT) scores have been used to evaluate the muscle strength of patients with spinal surgery, which is performed both preoperatively and postoperatively because surgery sometimes may lead to muscle weakness. Recently, intraoperative spinal cord monitoring (IOM) has been used to assess spinal cord functions during surgery based on motor evoked potentials (MEPs); however, its standardized evaluation method has not been established. Thus, prompt and objective assessment of spinal cord functions during surgery is still challenging. This study aimed to construct a machine learning (ML)-based method for identifying abnormalities in MEP waveforms and to predict postoperative motor function.
Methods:
This study proposes a method for automatically classifying surgical outcomes into normal or abnormal from MEPs during surgery. The proposed method adopts an ML-based model based on long short-term memory (LSTM). We collected MEP waveforms during surgeries and their corresponding preoperative and postoperative MMT scores from multiple muscle groups in 797 patients.
Results:
The trained LSTM-AE model had a recall of 57%, a sensitivity of 80%, and a false positive rate of 4%.
Conclusion:
The proposed MEP classification method can potentially detect spinal cord compression or injury during surgical procedures.Clinical relevanceThis study demonstrates the potential of ML to improve the accuracy of intraoperative spinal cord function assessment, which could lead to better outcomes and reduced surgical complications.
