Evaluation of neuroinflammation in refractory MRI-negative epilepsy patients by feature selection algorithms

Background and Aims: The detection of neuroinflammation in MRI-negative epilepsy patients with drug resistance could be a promising clue for finding an effective treatment. The feature selection algorithms could therefore be helpful in introducing the best MRI parameter to describe an inflammation of the epileptic brain. Methods: A total of ۸۰ ROIs were drawn in both normal and epileptic white and gray matter in the hemisphere from seven drug-resistant epilepsy patients without abnormalities in the brain images to obtain brain parameters such as the magnetization transfer ratio, perfusion factors (Ktrans, Kep, Vb, CBV, CBF) and oxygen extraction fraction in the main veins (SS, Galen, ICVL, ICVR, SSS) in ۳-Tesla T۱ MRI images. The best parameter for detecting neuroinflammation compared to the normal brain hemisphere was determined by two feature selection algorithms (Relief and NCA). Results: Based on similar results which obtained from two feature selection methods were used to highlight the differences in the gray and white matter of the epileptic and normal brain regions, the highest score was achieved for the perfusion parameter Ktrans in gray matter, and magnetization transfer in white matter and gray matter was rated next. Conclusion: MRI parameters such as the perfusion constant Ktrans and the magnetization transfer ratio, which used to describe high blood supply, high macromolecular content and edema, could be remarkable signs of inflammation in the epileptic region of the epileptogenic brain. The detection of neuroinflammation in the brain could therefore open up a new approach for treating patients with refractory MRI-negative epilepsy.

کلیدواژه ها:

Neuroinflammation ، Refractory MRI-negative Epilepsy ، Feature Selection