Transfer Learning Approach to Detect and Classify Glioma Tumors

Glioma brain tumors are malignant and are commonly found in adults. To provide suitable treatment for anindividual patient the doctors have to find the tumor location, size, growth, consider the symptoms and possible cure than the risks of the other treatment or modalities. Due to the advancing technologies like Deep Learning and Computer vision it has become easier to tackle the issues in the healthcare industry. Medical imaging plays a vital role in the diagnosis of tumors. The objective of this paper is to present an ideal and efficient model to accurately segment and classify the Low-Graded Gliomatumor using Transfer Learning to assist the doctors in detecting and localizing the tumors. The parameters of the classifier and segmentation models are tuned to produce better results compared to existing models. EfficientNet is used for classification and obtained 97% accuracy and the SegNet segmentation model obtained a 0.99 IOU score and Dice Coefficient of 0.81.