Tumors are defined as masses of cells that abnormally grow in a part of the body where it is not necessary. They are further classified as benign (non-cancerous, like moles) and malignant (cancerous, like sarcomas or leukemias) depending on the fact if they can spread to other parts of the body. This work focuses on Intracranial Brain Tumors, which refer to the tumors found within the intracranial space, i.e., the brain. Any kind of tumor that grows in a limited space like within the cranium, has a larger probability of causing a stroke. This stroke itself could be an ischemic one, where a clot in an artery causes decreased blood flow to some other part of the brain, or a hemorrhagic one, that causes actual bleeding within the brain due to bursting of an artery. However, both cases would lead to the death of some of the brain cells, (also called stroke lesions) that implies serious impairment of brain’s activities ranging from paralysis to death of the individual.
Hence, considering the amount of damage caused by a stroke, with the help of this work, an algorithm using deep learning (LSTM Networks) will be put forward with encoders and decoders that would help statistically determine the locations in the brain that are more vulnerable to stroke lesions in case the individual has a tumor. This work will be using dataset of 3D MRI images of the brain in order to diagnose the tumor as well as the presence of stroke lesions. And finally with the result of this algorithm, doctors and diagnosed patients would be more aware of the possibility of a stroke to be able to either prevent them or diagnose and treat them at an early stage.