Highly Accurate Brain Tumor Segmentation and Classification Using Multiple Feature Sets

Megha Sunil  Borse; Murali Prasad  R; Tummala Ranga  Babu

doi:10.18502/fbt.v12i3.19183

Megha Sunil Borse Department of Electronics and Telecommunication Engineering, Cummins College of Engineering for Women, Karvenagar, Pune, India
Murali Prasad R Department of Electronics and Communication Engineering, Marri Laxman Reddy Institute of Technology and Management, Dundigal, Hyderabad, India
Tummala Ranga Babu Electronics & Communication Engineering, Rayapati Venkata Rangarao & Jagarlamudi Chandramouli College of Engineering, Guntur, India

DOI: https://doi.org/10.18502/fbt.v12i3.19183

Keywords: Brain Tumor; Adaptive Histogram Equalization; Pulse Coupled Neural Networks; Deep Convolutional Network; Local Binary Patterns.

Abstract

Purpose: Nowadays, detecting brain tumors is a crucial application. If a tumor is discovered later on, the medical issues are significant. Therefore, early diagnosis is essential. Magnetic Resonance Imaging (MRI) is the most recent detection, diagnosis, and assessment technology.

Materials and Methods: In this study, MRI images are segmented before input to a pulse-coupled neural network model to identify the existence of a tumor in the brain picture. The doctor may turn to this model for assistance if there are more input MRI brain pictures. This work preprocesses the images using normalization smoothing with linear filter and adaptive histogram. Statistical and Local Binary Patterns (LBP) features are extracted from the preprocessed images to perform the classification process. The Deep Convolutional Network (DCNN) is used to segment the image. The Pulse Coupled Neural Networks (PCNN) categorize the input images as normal and tumor.

Results: Accuracy, sensitivity, specificity, and precision are the various metrics evaluated. This work achieves 99.35 accuracies, 99.78 sensitivity, 98.45 specificities, and 97.61 precision. This work is compared with previous implementations to measure performance.

Conclusion: The comparison analysis improves tumor segmentation and classification accuracy. The suggested method yields great outcomes.