Prediction of Postoperative Speech Dysfunction Based on Cortico-Cortical Evoked Potentials and Machine Learning

Area of Science:

• Neurosurgery
• Machine Learning
• Neurophysiology

Background:

• Postoperative speech dysfunction is a significant concern following neurosurgery.
• Intraoperative cortico-cortical evoked potentials (CCEP) offer a potential biomarker for predicting neurological deficits.
• Refining tumor resection criteria can improve patient quality of life.

Purpose of the Study:

• To evaluate the predictive performance of machine learning models using initial intraoperative CCEP for postoperative speech dysfunction.
• To assess the feasibility of predicting speech deterioration before glial tumor removal.

Main Methods:

• Utilized CCEP data from 26 neurosurgery patients.
• Applied machine learning models: random forest, logistic regression, and support vector machines (linear, radial, polynomial kernels).
• Evaluated model performance using F1-score, sensitivity, and specificity.

Main Results:

• The best prediction model, a support vector machine with a polynomial kernel, achieved an F1-score of 0.638.
• Models generally exhibited high sensitivity (up to 0.993) but low specificity.
• Solely using pre-resection CCEP demonstrated insufficient predictive quality for speech dysfunction.

Conclusions:

• Intraoperative CCEP recorded before glial tumor resection alone is insufficient for accurate prediction of postoperative speech dysfunction.
• Further research into the postresectional dynamics of CCEP is warranted.
• This pilot study highlights the need for improved predictive models in neurosurgery.