術後腫瘍周囲機能的結合性がグリオーマ患者の神経学的予後を予測する

The study investigated glioma patients after surgical resection of tumor tissue using postoperative functional magnetic resonance imaging (fMRI) to assess cavity-adjacent (perilesional) functional connectivity as a predictor of overall survival and functional recovery.

We developed an analytic method to quantify the postoperative whole-brain functional connectivity. Resting-state whole-brain fMRI scans acquired from 12 glioma patients following surgical resection were analyzed as part of a proof-of-concept study. In particular, connectivity of the resected perilesional area was compared to that of the corresponding contralateral homologue region, and the difference between perilesional and contralateral connectivity was calculated. To test whether the functional connectivity metric could predict recovery of neurological outcomes, we compared patients' connectivity metrics from postoperative scans with changes in Karnofsky Performance Status (KPS) score between preoperative assessment and 6-month follow-up. Additionally, we examined whether the connectivity metric could predict overall survival by dividing the patients into subgroups based on their median survival time and comparing connectivity metrics.

Our analysis showed altered functional connectivity between perilesional and corresponding contralateral regions following surgical resection of glioma. The connectivity metric from postoperative scans was significantly correlated with recovery of neurological outcomes, as reflected by changes in KPS from preoperative to 6 months postoperative period (ρ = 0.97, p < 0.001). Moreover, individuals with survival times greater than 15 months showed significantly higher connectivity than those with shorter survival times (p = 0.0016 and Cohen's d = 2.74 in all subjects, p = 0.02 and Cohen's d = 1.90 in the subset of subjects with Grade IV gliomas). Furthermore, we developed machine learning models based on functional connectivity features, and they were able to predict the survival time with an accuracy of 92% and predict the KPS changes with an absolute error of 5.84 ± 6.08.

Overall, our study showed that resting-state fMRI from patients after glioma resection is relevant to their long-term neurological outcomes: decreased connectivity in the perilesional regions compared to the contralateral regions indicates less survival time and worsened functional outcomes. The reported analytics from postsurgical fMRI scans, combined with the machine learning model, could provide important prognostic information for postsurgical recovery management.