ABSTRACT
The Central Nervous System (CNS) is the most important part of the human body. Brain tumor is one of the deadliest diseases that affect the CNS. The survival rate of this is very low when compared to other types of cancer. Thus early detection will help to increase the survival rate and diagnose it properly. In this paper, we have proposed an effective way for detecting and classifying brain tumors using different versions of DarkNet like DarkNet-19 and DarkNet-53. We have customized these networks via the method of transfer learning for implementing different networks. Two detectors are proposed for detecting brain tumors from MRI data and two classifiers are proposed for classifying the MRI data into 4 different classes. The detector using DarkNet19 is providing an accuracy of 90.8% and that using DarkNet53 is providing an accuracy of 94.6%. The classifier using DarkNet19 is providing an accuracy of 90.6% that using DarkNet53 is providing an accuracy of 94.1%. The performance parameters indicate that the networks are performing well when compared to the state-of-the-art methods and these can be used for developing computer aided diagnosis systems.
