Machine learning enables automatic brain mapping for safer neurosurgery

In brain diseases that require a surgical intervention (for example in brain tumors), it is crucial to make sure that all the affected tissue is removed while preserving as much healthy tissue as possible to save important brain functions. Among those, hand movements are especially vital for the quality of life that patients will enjoy after surgery. Therefore, surgeons go to great lengths to identify the responsible brain areas. Unfortunately, the hand representation has slightly different shapes and locations within the gyri and sulci of each individual brain and cannot be determined reliably before the surgery. Therefore, each surgical intervention starts with a dedicated mapping procedure to identify the relations between brain locations and their functions.

 

Up to now, the procedure for this brain mapping involved placing standardized, relative course electrode grids (with a spacing of 1 cm between electrode contacts) onto the brain’s surface during surgery. The surgeon would then manually test and analyze each and every electrode contact separately, looking for specific electrophysiological signs (short latency SSEP phase reversal) that indicate the border between somatosensory cortex (related to sensation) and motor cortex (needed for actively producing movements). This manual procedure is error prone, not always conclusive and takes a lot of time and effort.

 

A Houston clinical research group of neurosurgeons, anesthesiologists and biomedical engineers have now devised a completely new automatized mapping method – using up-to-date unsupervised machine learning methods and custom made electrocorticography (ECoG) grids produced by CorTec.

Compared to standard strip electrodes, the individually tailored ECoG grids allowed mapping of the whole brain surface in a single procedure, rather than repeating the mapping several times after repositioning a conventional electrode from one location to the other to cover the whole area. Also, higher spatial resolution could be achieved by using high-resolution grids with smaller inter-electrode spacing.

 

With this setup and their smart, unsupervised automatic mapping, the authors could achieve reliable mapping in real time and with better accuracy than using the manual procedure. The new method has the potential to make brain surgery more accurate and relieve surgeons from a substantial mental burden during surgery – helping them to focus on their main task – removing the right brain tissue safely and precisely to ensure patient safety.

 

Citation:

Priscella Asman, Sujit Prabhu, Dhiego Bastos, Sudhakar Tummala, Shreyas Bhavsar, Thomas Michael McHugh, Nuri Firat Ince

Unsupervised machine learning can delineate central sulcus by using the spatiotemporal characteristic of somatosensory evoked potentials.

2021 J. Neural Eng. 18 046038. doi: 10.1088/1741-2552/abf68a.

 

 

Learn more about the CorTec °AirRay Cortical Electrodes for human use.

Contact

Do you have any questions regarding our company or our technology?
We appreciate your feedback.