Towards prediction-based connectivity-informed personalized awake surgery

The challenge faced by glioma neurosurgeons is to perform the maximal extent of resection while preserving a high level of cognitive functions (thus allowing patients to resume a normal socio-professional life). It has been demonstrated that awake surgery is currently the best methodology to achieve this goal. However, translating this approach from motor and language functions to higher-order cognitive functions (HOCF, for eg cognitive flexibility) raises specific issues. First, it is not always possible to test HOCF within the constrained intraoperative settings. Second, it has been observed that, in absence of intraoperative testing of HOCF, while many patients are impaired in the acute postoperative period, a large majority of them do recover in the long term, thanks to the plasticity-mediated reorganization of brain networks. In order to predict which patient would not recover from an intended glioma resection, we combined network-level disconnectomics lesional models with machine learning algorithms, and we applied this methodology to the specific case of cognitive flexibility. From this proof-of-concept study, we envision the development of a new software that would help neurosurgeon to preoperatively delineate the optimal extent of resection.