gms | German Medical Science

Objective: Tumour growth next to functional areas requires precise delineation for maximal resection while minimizing the risk for functional deficits. Here, a novel approach is presented to analyse and visualize eloquent areas of the exposed human cortex during neurosurgical interventions using intraoperative thermography.

Methods: All patients provided informed consent. Focal activations of the primary somatosensory cortex were provoked by contralateral median nerve stimulation (SEP). Thermographic data was acquired during five resections of tumour tissue beneath the central sulcus. Thermography detects small changes 0.03°C in cortical temperature distribution. We designed a semiparametric regression framework to analyse the acquired thermographic data of the exposed human cortex. This framework compensates certain background signals while it retains the elicited response to the electrical stimulation. We validated the recognized eloquent areas with anatomical localization as well as electrophysiological measurements.

Results: In a simulation study on augmented data, our approach achieved a true-positive rate of 90.43% and a false-positive rate of 2.18%. Furthermore, we were able to visualize dense clusters of active pixels on the postcentral gyrus in seven of the nine tumour cases (77.8%) with respect to the significance level p<0.001 and corresponding z-score |z|≥3.2905. The average cluster size was 1848.3±1578.2 pixels with an (per cluster) average z-score of z=4.17±0.63. The standard deviation was affected by one case having two clusters of size 129 and 780 pixels. The spatial position of the extracted activation signal correlated in all cases with the electrophysiologic phase reversal measurements identifying the primary sensory cortex.

Conclusion: Temporal changes of temperature mainly correlate with changes in regional cerebral blood flow. This implies that we can visualize local neural activity due to neurovascular coupling. Based on this hypothesis, we designed a semiparametric regression framework which recognizes patterns in thermographic data even at low signal-to-noise ratios. The detection and visualization of neural activity on the exposed human cortex provides additional neurotopological information during neurosurgical interventions which correlates with anatomy and electrophysiologic responses.