Please use this identifier to cite or link to this item:
http://hdl.handle.net/10603/427616
Title: | A deep learning approach for brain Tumor classification based on Regional heterogeneity using pixel Level feature descriptors |
Researcher: | Gunasundari, C |
Guide(s): | |
Keywords: | Engineering and Technology Computer Science Computer Science Information Systems brain Tumor Regional heterogeneity |
University: | Anna University |
Completed Date: | 2022 |
Abstract: | Brain tumor is considered to be widely analyzed disease for newlineeffective diagnosis and treatment planning. Several approaches were framed newlineto detect and diagnose tumor at early stage. Medical imaging techniques play newlinea crucial role in the diagnosis of the tumors. These techniques are used to newlinelocate and assess the progression of the tumor before and after treatment. MRI newlineis usually the modality of choice for diagnosis and treatment planning for newlinebrain tumors because of its high resolution, delicate tissue contrast, and newlinenon-obtrusive qualities. More than one MRI slice is required to view different newlineregions of the brain, e.g., T1, T2, T1 contrast and FLAIR images. newlineThe proposed research focuses on Heterogeneity-aware Local newlineBinary Patterns for brain tumor detection. This approach tries to put solution newlineon the different problems of the arrangement of the Magnetic resonance newlineimaging. Texture analysis is performed in the input brain MRI to identify the newlinenature of tumor and categorize it. Prominent Feature descriptors are used for newlinethe extraction of the features from the MRI for the grouping of the tumor. newlineLocal Binary Pattern and Principal Component Analysis were used for feature newlineextraction and reduction and SVM is used for classification. These methods newlinecombine the intensity and the components of the shapes and the different newlineorders with the textures of the tumor from the MRI images. It achieves newline98.81% of accuracy; 01.54% of error rate; 97.99% of TPR; 96.12% of TNR; newline97.63% of precision and 97.82% of F1-score. newlineBrain tumor segmentation methods supported traditional image newlineprocessing and machine learning aren t ideal among the currently proposed newlinebrain segmentation methods newline |
Pagination: | xvii, 139p. |
URI: | http://hdl.handle.net/10603/427616 |
Appears in Departments: | Faculty of Information and Communication Engineering |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
01_title.pdf | Attached File | 22.2 kB | Adobe PDF | View/Open |
02_prelim pages.pdf | 1.36 MB | Adobe PDF | View/Open | |
03_content.pdf | 8.74 kB | Adobe PDF | View/Open | |
04_abstract.pdf | 7.65 kB | Adobe PDF | View/Open | |
05_chapter 1.pdf | 590.74 kB | Adobe PDF | View/Open | |
06_chapter 2.pdf | 37.63 kB | Adobe PDF | View/Open | |
07_chapter 3.pdf | 90.86 kB | Adobe PDF | View/Open | |
08_chapter 4.pdf | 241.25 kB | Adobe PDF | View/Open | |
09_chapter 5.pdf | 502.92 kB | Adobe PDF | View/Open | |
10_chapter 6.pdf | 1.58 MB | Adobe PDF | View/Open | |
11_annexures.pdf | 62.87 kB | Adobe PDF | View/Open | |
80_recommendation.pdf | 60.39 kB | Adobe PDF | View/Open |
Items in Shodhganga are licensed under Creative Commons Licence Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).
Altmetric Badge: