Machine learning techniques for medical diagnosis of diabetes using iris images. (April 2018)

• Record Type: Journal Article
• Authors: Samant, Piyush; Agarwal, Ravinder
• Is Part Of: Computer methods and programs in biomedicine. Volume 157(2018)
• Page Start: 121
• Page End: 128
• Publication Date: 2018-04
• Subjects: Iris -- Diabetes -- Segmentation -- Feature extraction -- Classification -- Disease diagnosis; Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2018.01.004
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlights: Diabetes detection through computer machine vision technique using iris. Statistical, textural and discrete wavelet transformation features extracted from region of interest. Comparison of 6 different classifier was presented. Abstract: Background and Objective: Complementary and alternative medicine techniques have shown their potential for the treatment and diagnosis of chronical diseases like diabetes, arthritis etc . On the same time digital image processing techniques for disease diagnosis is reliable and fastest growing field in biomedical. Proposed model is an attempt to evaluate diagnostic validity of an old complementary and alternative medicine technique, iridology for diagnosis of type-2 diabetes using soft computing methods. Methods: Investigation was performed over a close group of total 338 subjects (180 diabetic and 158 non-diabetic). Infra-red images of both the eyes were captured simultaneously. The region of interest from the iris image was cropped as zone corresponds to the position of pancreas organ according to the iridology chart. Statistical, texture and discrete wavelength transformation features were extracted from the region of interest. Results: The results show best classification accuracy of 89.63% calculated from RF classifier. Maximum specificity and sensitivity were absorbed as 0.9687 and 0.988, respectively. Conclusion: Results have revealed the effectiveness and diagnostic significance of proposed model for non-invasive and automatic diabetes diagnosis.