Translated from Turkish language
Abstract
Iridology is a form of complementary medicine that relies on examining the pattern, color and other properties of the iris to determine information about the patient’s systemic health. Today, many doctors use this form of analysis in conjunction with other healthcare techniques to facilitate a better understanding of patients’ health care needs. However, this examination and description is very subjective and depends on the experience of the doctors. It is also a time consuming and tiring process for physicians. In this context, in order to eliminate subjectivity in examination and identification and to reveal a more objective definition, a deep learning and image processing-based method is proposed in this study for the diagnosis of diabetes by using iridology card from iris images. In the proposed method, the iris boundaries were found and the pancreas region shown on the iridology card was removed from the iris fully automatically. With the image processing steps applied, an area related to the Pancreas (diabetes) was found on the iris and automatic segmentation was made from the eye image. Afterwards, these images were applied to convolutional neural networks to diagnose diabetes and compared with different convolutional neural network architectures. As a result, it was observed that the proposed method with VGG-16 architecture and automatic segmentation of the area of the pancreatic region was more successful with 80% Accuracy, 100% Sensitivity, 71.42% Precision, 60% Specificity and 83.33% F1 Score performance metrics.
Author: Yüksek Lisans Tezleri
Source: Afyon Kocatepe University, Turkey – http://acikerisim.aku.edu.tr/xmlui/handle/11630/8496
Download Full Abstract: https://iridology-research.com/pdf/10192372.pdf
