• Nandkishore Nangre
• Shubham Kumbhar
• Shubham Lingawar
• Abhishek Limkar
• Dr. Dwarkoba Gaikwad

A brain tumor is caused when brain cells divide and grow in an uncontrolled way. There are various types of brain tumor, each fatal in its own way, which is imperative to detect as early as possible. Magnetic Resonance Imaging (MRI) is a broadly used imaging technique to evaluate these tumors, limiting the use of explicit quantitative estimations in the clinical practice. The existing systems for classifying these tumors implemented direct usage of the fully connected network, and use of large filters which eventually increased the computation time and generated unsatisfactory results. Thus, these limitations of the existing systems motivated us to develop an automated and reliable classification system for brain tumor. We propose an automatic classification method based on Convolutional Neural Networks (CNN), by using comparatively small-sized kernels. The use of small kernels allows us to design a profound architecture, besides having a positive effect against over-fitting, given the fewer number of weights in the network. We are trying to implement a better activation function for improving the accuracy of the proposed model. We are investigating the use of intensity normalization as a pre-processing step, which along with data augmentation to be very effective for brain tumor segmentation in MRI images.

CNN, median filter, LeakyRelu, MaxPooling

"Brain Tumor Classification using CNN", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.6, Issue 5, page no.270-276, May-2019, Available :http://www.jetir.org/papers/JETIR1905M45.pdf

"Brain Tumor Classification using CNN", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 5, page no. pp270-276, May-2019, Available at : http://www.jetir.org/papers/JETIR1905M45.pdf