A Deep Feature Fusion of Improved Suspected Keratoconus Detection with Deep Learning

Ali H Al-Timemy; Laith Alzubaidi; Zahraa M Mosa; Hazem Abdelmotaal; Nebras H Ghaeb; Alexandru Lavric; Rossen M Hazarbassanov; Hidenori Takahashi; Yuantong Gu; Siamak Yousefi

doi:10.3390/diagnostics13101689

A Deep Feature Fusion of Improved Suspected Keratoconus Detection with Deep Learning

Diagnostics (Basel). 2023 May 10;13(10):1689. doi: 10.3390/diagnostics13101689.

Authors

Ali H Al-Timemy¹, Laith Alzubaidi^{2

3}, Zahraa M Mosa⁴, Hazem Abdelmotaal⁵, Nebras H Ghaeb¹, Alexandru Lavric⁶, Rossen M Hazarbassanov^{7

8}, Hidenori Takahashi⁹, Yuantong Gu^{2

3}, Siamak Yousefi^{10

11}

Affiliations

¹ Biomedical Engineering Department, Al-Khwarizmi College of Engineering, University of Baghdad, Baghdad 10011, Iraq.
² School of Mechanical, Medical, and Process Engineering, Queensland University of Technology, Brisbane, QLD 4000, Australia.
³ ARC Industrial Transformation Training Centre-Joint Biomechanics, Queensland University of Technology, Brisbane, QLD 4000, Australia.
⁴ Department of Physics, College of Science, Al-Nahrain University, Baghdad 64021, Iraq.
⁵ Department of Ophthalmology, Assiut University, Assiut 71526, Egypt.
⁶ Computers, Electronics and Automation Department, Stefan cel Mare University of Suceava, 720229 Suceava, Romania.
⁷ Medical School, Universidade Anhembi Morumbi, São Paulo 03101-001, Brazil.
⁸ Department of Ophthalmology and Visual Sciences, Paulista Medical School, Federal University of São Paulo, São Paulo 04021-001, Brazil.
⁹ Department of Ophthalmology, Jichi Medical University, Tochigi 329-0431, Japan.
¹⁰ Department of Ophthalmology, University of Tennessee Health Science Center, Memphis, TN 38163, USA.
¹¹ Department of Genetics, Genomics, and Informatics, University of Tennessee Health Science Center, Memphis, TN 38163, USA.

Abstract

Detection of early clinical keratoconus (KCN) is a challenging task, even for expert clinicians. In this study, we propose a deep learning (DL) model to address this challenge. We first used Xception and InceptionResNetV2 DL architectures to extract features from three different corneal maps collected from 1371 eyes examined in an eye clinic in Egypt. We then fused features using Xception and InceptionResNetV2 to detect subclinical forms of KCN more accurately and robustly. We obtained an area under the receiver operating characteristic curves (AUC) of 0.99 and an accuracy range of 97-100% to distinguish normal eyes from eyes with subclinical and established KCN. We further validated the model based on an independent dataset with 213 eyes examined in Iraq and obtained AUCs of 0.91-0.92 and an accuracy range of 88-92%. The proposed model is a step toward improving the detection of clinical and subclinical forms of KCN.

Keywords: convolutional neural networks; deep learning; feature fusion; keratoconus; machine learning; transfer learning.

Grants and funding

ARVO Roche 2022/ARVO