Supervised machine learning methods to identify muscles from MEP traces - a proof of concept design

Objective: Even for an experienced neurophysiologist, it is challenging to look at a single graph of an unlabeled motor evoked potential (MEP) and identify the corresponding muscle. We demonstrate that supervised machine learning (ML) can successfully perform this task and surpass trained neurophysiologists. Methods: Intraoperative MEP data from surgery on 36 patients was included for the classification task with 4 muscles: Extensor digitorum (EXT), abductor pollicis brevis (APB), tibialis anterior (TA) and abductor hallucis (AH). Three different supervised ML classifiers (random forest (RF), k-nearest neighbors (kNN) and logistic regression (LogReg)) were trained and tested on either raw or compressed data (PCA or feature extracted). Patient data was classified considering either all 4 muscles simultaneously, 2 muscles within the same extremity (EXT versus APB), or 2 muscles from different extremities (EXT versus TA). For comparison, we asked 30 experienced neurophysiologists to carry out a similar 4-muscle classifying task with MEP data from one patient. Results: In all cases, RF classifiers performed best and kNN second best. The highest performances were achieved on raw data. In the 4-muscle comparison, the RF classifier achieved an accuracy of 83%. Across limbs (EXT versus TA) it reached 97% accuracy, while in the within limb comparison (EXT versus APB) the dropped to 89%. On the other hand, human performance reached 64% accuracy on the 4-muscle comparison. Conclusion: Standard ML methods show surprisingly high performance on a classification task with minimally processed intraoperative MEP signals. This study illustrates the power and challenges of standard ML algorithms when handling intraoperative signals. Ultimately, ML might help improve warning criteria in IOM and safety in the operating room.