A Hybrid Filter Based Feature Selection, Segementation and Ensemble Learning Model for Brain Stroke Prediction

  • Vijayadeep Gummadi, N. Naga Malleswara Rao

Abstract

Brain stroke detection (SD) is one of the major chronic diseases which can be detected using the MRI images. Digital image denosing is one of the major problem in most of the real-time systems due to high noisy level and low level resolution. Inter- and intra-variance between the signals that lead to the noisy factor in digital images during image acquisition. Images are generated with various kinds of noises such as Gaussian, speckle, impulsive and combined noise in the brain stroke images. Most of the compressed or noisy images are difficult to analyze due to the existence of noise on the edges using traditional denosing techniques such as non-linear median filter, Bayesian filter, wavelet-based shearlet transform etc. In order to remove the noise in the speckle noise, traditional denoising approaches such as Bayesian denoise, non-local filter, wavelet based shearlet transformation, autoencoders etc. are used. Because of the existence of multiple additive, multiplicative and Gaussian noise, such denoising techniques are difficult to process ultrasound images and medical images. The problem of sparsity in the low SNR images can also not be resolved by these models. To overcome these issues, a hybrid non-linear filter and segmentation based ensemble learning framework is implemented in order to improve the denosing level on different types of imaging systems. Experimental results are simulated on different realtime noisy images in order to check the effiency of denosing approach compared to the conventional approaches.

Published
2022-01-13
How to Cite
N. Naga Malleswara Rao, V. G. (2022). A Hybrid Filter Based Feature Selection, Segementation and Ensemble Learning Model for Brain Stroke Prediction. Design Engineering, (1), 4823- 4838. Retrieved from http://www.thedesignengineering.com/index.php/DE/article/view/9885
Section
Articles