Brain Tumor Detection and Classification Using Improved Unet

Machine learning algorithms have become popular in various fields, including health informatics, pandemic forecasting, user experience evaluation, and predicting shear strength. In medical imaging, machine learning algorithms are widely used to classify brain tumors. Pre-processing steps include aug...

Full description

Saved in:
Bibliographic Details
Published in2024 Asia Pacific Conference on Innovation in Technology (APCIT) pp. 1 - 6
Main Authors Palleti, Venkata Kusuma, Sivappagari, Chandra Mohan Reddy
Format Conference Proceeding
LanguageEnglish
Published IEEE 26.07.2024
Subjects
Accuracy
Brain cancer
Brain modeling
Brain Tumor
CNN
Deep learning
Image segmentation
Improved U-Net
Machine learning algorithms
Magnetic resonance imaging
MRI
U-Net
Online AccessGet full text

Cover

More Information
Summary:Machine learning algorithms have become popular in various fields, including health informatics, pandemic forecasting, user experience evaluation, and predicting shear strength. In medical imaging, machine learning algorithms are widely used to classify brain tumors. Pre-processing steps include augmentation, applying filters, segmenting data, and choosing features. Post-processing steps include identification and classification. Both conventional and deep learning approaches can be used to implement these steps, with hand-crafted features used in the conventional approach and models tuned in the deep learning approach.It is also possible to use metaheuristic algorithms to improve classification accuracy. Several studies have been done on this subject, and The focus of this research encompasses traditional and deep learning methods for detecting brain cancers images. Although the conventional approach is faster, the deep learning method has been reported better. A streamlined rendition of the U-Net++ deep learning architecture is presented for the segmentation of cerebral neoplasms in MRI scans. This implementation provides instantaneous segmentation capabilities without the necessity for extensive training data or supplementary data augmentation procedures. The U-Net++ model demonstrates encouraging outcomes on the dataset and outperforms conventional benchmark algorithms in terms of precision, while also applying moderate and precise noise filtration to enhance the quality of MRI images..
DOI:10.1109/APCIT62007.2024.10673512