Enhancing Accuracy in Detection of Meningioma Tumour Using RestNet 50 Deep CNN Model
Retinderdeep Singh, Chander Prabha
Abstract
Brain tumor diagnosis is critical as the treatment hinges on the location and size of the tumor in the medical field. It is the vital organ in the human body of all living beings as it controls motor skills, decision-making, and all the voluntary or involuntary actions of the body. Brain tumors are extremely fatal as the last stage leads to cancer; however, they can be cured if diagnosed early enough. A meningioma tumor develops via the meninges membranes that cover the brain and spinal cord. Meningioma is one of the uttermost types of deadly tumors and makes up roughly 30% of all cases related to tumors. The death rate is 35% for adults above 40 and 21% for children. MRI (Magnetic Resonance Imaging) is a vital tool for detecting tumors, offering high-resolution insights into brain structures. The paper proposes a ResNet50-based model to detect meningioma tumors from the MRI of the patient. The model is trained using an open-source dataset of images containing the MRIs of patients with meningioma tumors and some images of healthy individuals. The model uses the Adam optimizer and activation function (SoftMax) to optimize results and to normalize the output of the neural network respectively. The model achieved an outstanding accuracy of 96.63% typically greater than other traditional CNN models.