Shared and malignancy-specific functional plasticity of dynamic brain properties for patients with left frontal glioma

• Published in: Cerebral cortex (New York, N.Y. 1991) Vol. 34; no. 1
• Main Authors: Cai, Siqi, Liang, Yuchao, Wang, Yinyan, Fan, Zhen, Qi, Zengxin, Liu, Yufei, Chen, Fanfan, Jiang, Chunxiang, Shi, Zhifeng, Wang, Lei, Zhang, Lijuan
• Format: Journal Article
• Language: English
• Published: United States Oxford University Press 14-01-2024
• Subjects: Brain; Brain Mapping; Brain Neoplasms - diagnostic imaging; Glioma - diagnostic imaging; Humans; Magnetic Resonance Imaging; network configuration; functional connectivity; support vector machine; dynamic network activity; left frontal glioma
• ISSN: 1047-3211; 1460-2199
• DOI: 10.1093/cercor/bhad445

Abstract
The time-varying brain activity may parallel the disease progression of cerebral glioma. Assessment of brain dynamics would better characterize the pathological profile of glioma and the relevant functional remodeling. This study aims to investigate the dynamic properties of functional networks based on sliding-window approach for patients with left frontal glioma. The generalized functional plasticity due to glioma was characterized by reduced dynamic amplitude of low-frequency fluctuation of somatosensory networks, reduced dynamic functional connectivity between homotopic regions mainly involving dorsal attention network and subcortical nuclei, and enhanced subcortical dynamic functional connectivity. Malignancy-specific functional remodeling featured a chaotic modification of dynamic amplitude of low-frequency fluctuation and dynamic functional connectivity for low-grade gliomas, and attenuated dynamic functional connectivity of the intrahemispheric cortico-subcortical connections and reduced dynamic amplitude of low-frequency fluctuation of the bilateral caudate for high-grade gliomas. Network dynamic activity was clustered into four distinct configuration states. The occurrence and dwell time of the weakly connected state were reduced in patients’ brains. Support vector machine model combined with predictive dynamic features achieved an averaged accuracy of 87.9% in distinguishing low- and high-grade gliomas. In conclusion, dynamic network properties are highly predictive of the malignant grade of gliomas, thus could serve as new biomarkers for disease characterization.