Optimized preoperative motor cortex mapping in brain tumors using advanced processing of transcranial magnetic stimulation data.

BACKGROUND AND OBJECTIVE: Transcranial magnetic stimulation (TMS) is a useful technique to help localize motor function prior to neurosurgical procedures. Adequate modelling of the effect of TMS on the brain is a prerequisite to obtain reliable data.

METHODS: Twelve patients were included with perirolandic tumors to undergo TMS-based motor mapping. Several models were developed to analyze the mapping data, from a projection to the nearest brain surface to motor evoked potential (MEP) amplitude informed weighted average of the induced electric fields over a multilayer detailed individual head model. The probability maps were compared with direct cortical stimulation (DCS) data in all patients for the hand and in three for the foot. The gold standard was defined as the results of the DCS sampling (with on average 8 DCS-points per surgery) extrapolated over the exposed cortex (of the tailored craniotomy), and the outcome parameters were based on the similarity of the probability maps with this gold standard.

RESULTS: All models accurately gauge the location of the motor cortex, with point-cloud based mapping algorithms having an accuracy of 83-86%, with similarly high specificity. To delineate the whole area of the motor cortex representation, the model based on the weighted average of the induced electric fields calculated with a realistic head model performs best. The optimal single threshold to visualize the field based maps is 40% of the maximal value for the anisotropic model and 50% for the isotropic model, but dynamic thresholding adds information for clinical practice.

CONCLUSIONS: The method with which TMS mapping data are analyzed clearly affects the predicted area of the primary motor cortex representation. Realistic electric field based modelling is feasible in clinical practice and improves delineation of the motor cortex representation compared to more simple point-cloud based methods.