An Automated Algorithm for the Identification of Somatosensory Cortex Using Magnetoencephalography

The localization of eloquent cortex is crucial for many neurosurgical applications, such as epilepsy and tumor resection. Non-invasive localization of these cortical areas using magnetoencephalography (MEG) is generally performed using equivalent current dipoles. While this method is clinically validated, source localization depends on several subjective parameters. This paper aimed to develop an automated algorithm for identifying the cortical area activated during a somatosensory task from MEG recordings. Our algorithm uses singular value decomposition to outline the cortical area involved in this task. For proof of concept, we evaluate our algorithm using data from 10 subjects with epilepsy. Our algorithm has a statistically significant overlap with the somatosensory cortex (the expected active area in healthy subjects) in 6 of 10 subjects. Having thus demonstrated proof of concept, we conclude that our algorithm is ready for further testing in a larger cohort of subjects.Clinical relevance-Our algorithm identifies the dominant cortical area and boundary of the cortical tissue involved in a task-related response.