OBJECTIVE: By looking at how the accuracy of preoperative brain mapping methods vary according to differences in the distance from the activation clusters used for the analysis, the present study aimed to elucidate how preoperative functional neuroimaging may be used in such a way that maximizes the mapping accuracy. METHODS: The eloquent function of 19 patients with a brain tumor or cavernoma was mapped prior to resection with both functional MRI (fMRI) and magnetoencephalography (MEG). The mapping results were then validated using direct cortical stimulation mapping performed immediately after craniotomy and prior to resection. The subset of patients with equivalent MEG and fMRI tasks performed for motor (n = 14) and language (n = 12) were evaluated as both individual and combined predictions. Furthermore, the distance resulting in the maximum accuracy, as evaluated by the J statistic, was determined by plotting the sensitivities and specificities against a linearly increasing distance threshold. RESULTS: fMRI showed a maximum mapping accuracy at 5 mm for both motor and language mapping. MEG showed a maximum mapping accuracy at 40 mm for motor and 15 mm for language mapping. At the standard 10-mm distance used in the literature, MEG showed a greater specificity than fMRI for both motor and language mapping but a lower sensitivity for motor mapping. Combining MEG and fMRI showed a maximum accuracy at 15 mm and 5 mm—MEG and fMRI distances, respectively—for motor mapping and at a 10-mm distance for both MEG and fMRI for language mapping. For motor mapping, combining MEG and fMRI at the optimal distances resulted in a greater accuracy than the maximum accuracy of the individual predictions. CONCLUSIONS: This study demonstrates that the accuracy of language and motor mapping for both fMRI and MEG is heavily dependent on the distance threshold used in the analysis. Furthermore, combining MEG and fMRI showed the potential for increased motor mapping accuracy compared to when using the modalities separately. Clinical trial registration no.: NCT01535430 (clinicaltrials.gov)

Accurate individual functional mapping of task activations is a potential tool for biomarker discovery and is critically important for clinical care. While structural imaging does not directly map task activation, we hypothesized that structural imaging contains information that can accurately predict variations in task activation between individuals. To this end, we trained a convolutional neural network to use structural imaging (T1-weighted, T2-weighted, and diffusion tensor imaging) to predict 47 different functional MRI task activation volumes across seven task domains. The U-Net model was trained on 591 subjects and then subsequently tested on 122 unrelated subjects. The predicted activation maps correlated more strongly with their actual maps than with the maps of the other test subjects. An ablation study revealed that a model using the shape of the cortex alone or the shape of the subcortical matter alone was sufficient to predict individual-level differences in task activation maps, but a model using the shape of the whole brain resulted in markedly decreased performance. The ablation study also showed that the additional information provided by the T2-weighted and diffusion tensor imaging strengthened the predictions as compared to using the T1-weighted imaging alone. These results indicate that structural imaging contains information that is predictive of inter-subject variability in task activation mapping and that cortical folding patterns, as well as microstructural features, could be a key component to linking brain structure to brain function.