Mapping Activity Area Localization in Functional MRI Imaging with Deep Learning based Automatic Segmented Brain Tumor for Presurgical Tumor Resection Planning

Abstract

Functional Magnetic Resonance Imaging (fMRI) determines small blood flow variations that arise due to brain activity. fMRI major study is about functional anatomy which determines the area of the brain controlling vital functions such as hand and foot motor movements for both left and right, speech mantra, and speech word activities. For this instinctive localization of activity areas for specific tasks is very important. This paper appropriately describes the fMRI paradigm timeline with a modified fMRI paradigm timeline due to the hemodynamic response function (HRF).   Efficient activity area localization of thirty-three patients for fMRI data acquired from the hospital is achieved with dynamic thresholding. Dynamic thresholding is also effective in removing excess highlighted areas which helps in the reduction in expert efforts and time required to generate the patient report.  The localize activity area is further mapped with deep learning-based automatic segmented brain tumor regions to find overlapping regions. The exact location of the overlapping region is recovered which helps with preoperative counseling and tumor resection planning. All the results are verified and validated by two expert radiologists from the Hospital.