Hybrid auto encoder network for iris nevus diagnosis considering potential malignancy

Iris nevus can be described as a growth commonly found on the iris, or regions surrounding the pupil. This growth is usually pigmented and non-cancerous, and therefore harmless; often requiring little medical attention. However, it has been established that there exists a relatively high risk of transformation of such growths into iris melanoma, which is cancerous or malignant. Furthermore, it has been shown that iris nevus infected patients also run risk of developing secondary glaucoma which requires very crucial medical intervention. Considering the above mentioned severe medical conditions that are associated with iris nevus, its diagnosis hence becomes very important. Generally, the diagnosis of iris nevus is achieved by examining eye images of patients taken by a medical expert. However, diagnosis is not an easily achievable task considering how racial and environmental factors affect the colour of patients´ irises and pupils; hence pigmented growths may be concealed from a medical examiner. Also, factors such as stress and fatigue from examiners can lead to erroneous diagnosis. This research presents the use of trained hybrid auto encoders in the intelligent diagnosis of iris nevus. It is suggested that the use of the designed system as described in this work can significantly raise the confidence of medical diagnosis.