Deep neural network based tumor tissue classification in hyperspectral brain images

Abstract

Brain tumor is the most severe type of tumors and it is an newlineaccumulation of aberrant brain cells. Approximately 30% of brain tumors and newlineCentral Nervous System tumors are high-grade malignant gliomas. newlineGlioblastoma (GBM), the most intrusive and rapidly dividing malignant tumor, newlineaccounts for around 55% of cases. The statistics on brain tumors show that newlinemore than 12,000 individuals are diagnosed with a primary brain tumor every newlineyear, including 500 children and young people, which is equivalent to 33 people newlineevery day. Glioblastomas, which are aggressive brain tumors, have a newlinesignificantly lower 5-year survival rate of around 5%. GBMs are highly newlineinvasive and can lead to complications and side effects for patients. newlineTreatments for brain tumors often involve surgery, radiation therapy, newlineand chemotherapy, either in isolation or in combination. Surgery is typically newlinethe initial treatment for brain tumors, involving the extraction of as much of the newlinetumor as possible. Surgery improves the survival and patients quality of life. newlineWhile several imaging modalities are utilised as navigational aids in brain newlinetumor procedures, they have different limitations. Neurosurgeons currently newlinerely on multiple intraoperative guidance tools for tumor resection assistance, newlineincluding intraoperative Image Guided Stereotactic (IGS) neuronavigation and newlineintraoperative Magnetic Resonance Imaging (MRI). MRI solves the brain shift newlinephenomenon but increases the surgery duration and requires MRI compatible newlinesurgical equipment and has low spatial clarity. newline