Title

Iris Based Medical Diagnosis

Document Type

Oral Presentation

Campus where you would like to present

Ellensburg

Event Website

https://digitalcommons.cwu.edu/source

Start Date

16-5-2019

End Date

16-5-2019

Abstract

Automatic medical diagnosis without the involvement of trained medical personnel is a game changer in modern medicine. However, the implementation of these “computer doctors” is still in an early stage and there is a long process ahead until it will be accepted by the medical community. Our research is focusing on building a completely automatic tool to identify different eye pathologies by analyzing the irises of humans. The application we built can analyze, recognize, and identify different ocular diseases using as input human iris images such as geometry issues, occlusions, tissue and last but not least healthy eyes. First, the software segments the iris part of the eye from the input image which is later involved in a recognition phase considering several machine learning strategies. For our research experiments we considered different complex convolusional neural networks, namely the VGG19, the pretrained Xception, and the pretrained InceptionV3 model. To evaluate the accuracy of our method the Warsaw BioBase Disease Iris Database v2.1 was considered. The obtained results are very promising and an overall accuracy of 76% was achieved to separate healthy iris images from the non healthy ones.

Faculty Mentor(s)

Szilard Vajda

Department/Program

Computer Science

Share

COinS
 
May 16th, 11:00 AM May 16th, 12:00 PM

Iris Based Medical Diagnosis

Ellensburg

Automatic medical diagnosis without the involvement of trained medical personnel is a game changer in modern medicine. However, the implementation of these “computer doctors” is still in an early stage and there is a long process ahead until it will be accepted by the medical community. Our research is focusing on building a completely automatic tool to identify different eye pathologies by analyzing the irises of humans. The application we built can analyze, recognize, and identify different ocular diseases using as input human iris images such as geometry issues, occlusions, tissue and last but not least healthy eyes. First, the software segments the iris part of the eye from the input image which is later involved in a recognition phase considering several machine learning strategies. For our research experiments we considered different complex convolusional neural networks, namely the VGG19, the pretrained Xception, and the pretrained InceptionV3 model. To evaluate the accuracy of our method the Warsaw BioBase Disease Iris Database v2.1 was considered. The obtained results are very promising and an overall accuracy of 76% was achieved to separate healthy iris images from the non healthy ones.

https://digitalcommons.cwu.edu/source/2019/Oralpres/176