Hermite transform and support vector machine based analysis of schizophrenia disorder in magnetic resonance brain images. (2018)

• Record Type: Journal Article
• Authors: Latha, M.; Kavitha, G.
• Is Part Of: International journal of biomedical engineering and technology. Volume 27:Number 3(2018)
• Page Start: 203
• Page End: 220
• Publication Date: 2018
• Subjects: schizophrenia -- steered Hermite transform -- non-parametric region based active contour -- feature selection -- classifier -- SVM -- magnetic resonance images; Biomedical engineering -- Periodicals; 610.28
• Languages: English
• ISSNs: 1752-6418

Abstract

Schizophrenia (SZ) is a brain disorder characterised by disturbances in cognition and emotional responsiveness. In this work, steered Hermite Transform (HT) and SVM are used to analyse SZ. The non-parametric region-based active contour method is used to skull strip the MR images obtained from NAMIC database. These images are subjected to steered HT, and features such as mean, energy (E0 -E3 ), entropy and homogeneity are obtained. The significant features are selected using maximum relevance and subjected to classification. Results show that the proposed method is able to segment the brain region with higher accuracy (0.98), sensitivity (0.96) and F-score (0.95) compared to conventional methods. The prominent features mean and energy (E0 ) obtained from HT along with SVM could classify the normal and SZ better with an accuracy of 93.33% compared to Naïve Bayes and K-nearest neighbour classifiers. Hence, this framework could be used for better diagnosis of Schizophrenia.