Detection and Classification of Automated Brain Stroke Lesion with Optimized Dual Stage Deep Stacked Auto-Encoder

Sunil Babu  Melingi; C.  Tamizhselvan; R.  Surender; Vanga Karunakar  Reddy; Ramesh Kumar  Mojjada

doi:10.18502/fbt.v13i1.20782

Sunil Babu Melingi Department of Computer Science and Engineering, Vignan’s Foundation for Science Technology & Research (Deemed to be University), Vadlamudi, Guntur District, A.P, India
C. Tamizhselvan Department of Data Science & Business Systems, SRM Institute of Science & Technology, Kattankulathur, Chengalpattu District, Tamilnadu, India
R. Surender Department of Electronics and Communication Engineering, Rajyalakshmi Engineering College, Tandalam, Chennai, Tamilnadu, India
Vanga Karunakar Reddy Department of Electronics and Communication Engineering, Matrusri Engineering College, Saidabad, Hyderabad, India
Ramesh Kumar Mojjada Latent View Analytics Ltd. Neville Towar(Block-3), Rajiv Gandhi Salai, Chennai, Tamil Nadu- 603002

DOI: https://doi.org/10.18502/fbt.v13i1.20782

Keywords: Dual Stage Deep Stacked Auto-Encoder; Evolved Gradient Descent Optimization scheme; Wavelet Domain; Non-Contrast Computed Tomography Images; Brain Stroke Lesions.

Abstract

Purpose: The brain is the main organ of the human body. Stroke, also known as cerebral thrombosis, is a medical condition in which a rupture occurs in the blood vessels in the brain, resulting in brain damage. Symptoms may appear when the brain's blood flow and other nutrients are disrupted. A stroke is a medical emergency that can lead to long-term neurological impairment, resulting in complications and, in some cases, fatalities.

Materials and Methods: According to the World Health Organization, stroke is the primary determinant of mortality and disability globally. The early identification of various cardiovascular warning signs can reduce the impact of a stroke. The brain stroke dataset is used in existing methods. Delimitation of classifying stroke is difficult because the complication of lesion shapes and acquiring a ground truth is problematic, as it requires high clinical expertise and anatomical knowledge. In this manuscript, a Detection and Classification of Automated Brain Stroke Lesion with optimized Dual Stage Deep Stacked Auto-Encoder is proposed to detect brain stroke at an early stage with great accuracy.

Results: The input image is taken from slice-level Non-Contrast CT pictures dataset. The collected images are pre-processed and enhanced by removing skull regions, and then the rotations are performed by midline symmetry process. The preprocessed ROI region is fed to feature extraction, and the features are extracted using the wavelet domain. Next, the extracted features are given to classification and classified using Dual stage deep stacked auto encoder (DS-DSAE) optimized with Evolved Gradient Descent Optimization (EGDO) to effectively classify acute infarct, chronic infarct and ischemic stroke, haemorrhagic stroke, and normal.

Conclusion: The goal is to reduce computing complexity and enhance accuracy. The performance of the proposed wavelet-DS-DSAE-EGDO method achieves High accuracy 30.56%, 12.32%, 15.6%, 16.6%, 25.6%, 32.2%; High Precision 28.74%, 32.2%, 14.5%, 16.55%, 17.8%, 23.4% is comparing with the existing methods