Categorization of Brain Tumor on MR Images Using Deep Learning Models

Convolution Neural Network (CNN) algorithms have demonstrated notable capability in effectively analyzing human body image datasets obtained from MRI or CT with sufficient efficiency to direct medical operations. Although clinicians and radiologists can adeptly detect the presence of tumors and gene...

Full description

Saved in:
Bibliographic Details
Published in2024 2nd International Conference on Device Intelligence, Computing and Communication Technologies (DICCT) pp. 430 - 435
Main Authors Sree, Mundru Navya, Singamaneni, Rajesh, Sri, Modugula Kavya
Format Conference Proceeding
LanguageEnglish
Published IEEE 15.03.2024
Subjects
Brain modeling
Brain tumor
CNN
Computational modeling
Deep learning
Deep learning models image preprocessing
Histograms
InceptionV3
Magnetic resonance imaging
Noise reduction
ResNet-50
Streaming media
Tumor classification
Online AccessGet full text

Cover

More Information
Summary:Convolution Neural Network (CNN) algorithms have demonstrated notable capability in effectively analyzing human body image datasets obtained from MRI or CT with sufficient efficiency to direct medical operations. Although clinicians and radiologists can adeptly detect the presence of tumors and general abnormalities in brain scans, differentiating between specific tumor types typically requires significant time and effort. This study introduces an approach employing CNN algorithms for precise brain tumor classification into four distinct classes, leveraging deep learning models such as ResNet-50 and InceptionV3.Prior to classification, the dataset undergoes preprocessing using Adaptive Median Filter (AMF) and Contrast Limited Adaptive Histogram Equalization (CLAHE) for denoising and contrast enhancement. The classification task involves categorizing images into four classes namely Glioma, Meningioma, Pituitary and Tumorless. The proposed methodology aims to streamline the analysis process by minimizing complexity while maximizing accuracy rates when compared to previous developments.
DOI:10.1109/DICCT61038.2024.10533087