Multi fractal features along with empirical wavelet transform for detecting glaucoma

• Published in: AIP conference proceedings 2022-11, Vol.2483 (1)
• Main Authors: Abdulwahed, Hussam Y., Mohammed, Arkan Jassim
• Format: Article
• Language: English
• Subjects: Classifiers; Datasets; Empirical analysis; Feature extraction; Fractals; Glaucoma; Image acquisition; Image analysis; Image classification; Image enhancement; Image processing; Image segmentation; Least squares; Retinal images; Support vector machines; Wavelet transforms
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0117137

The main objective of this paper is to look further into the prospect of retinal image analysis aiming for the detection and tracing of glaucoma. The computer-based component analysis entails the use of image processing techniques in order to pre-process, identify and segment the region of interest, feature extraction as well as classification from the data-set images. The structural deformation of the optic disc and cup parts, one of the primary signs of glaucoma, necessitates high specificity in the segmentation procedure. As a solution to this issue, multi-fractal features as a non-morphological aspect are obtained in the proposed method from the enhanced optic disc area. Furthermore, the extracted image is decomposed using the Empirical Wavelet Transform to acquire different bands sub images. Subsequently, multifractal features where extracted. The resultant features are classified using a variety of classifiers. Among (Least Squares Support Vector Machine, Naive Bayes, Logistic Regression classifiers) Least Square Support Vector Machine outperforms all classifiers in terms of output. The proposed method has been investigated using different data-sets which includes regular and glaucoma images. Using Least Squares Support Vector, this proposed approach achieves a sensitivity of 86.71 percent in classifying glaucoma images and total accuracy of 89.94 percent.