An AMDOCT-NET for Automated AMD Detection under Evaluations of Different Image Size, Denoising and Cropping

• Main Authors: You, H.-Y., Wei, H.-T., Lin, C.-H., Ji, J.-Y., Liu, Y.-H., Lu, C.-K., Wang, J.-K., Huang, T.-L.
• Format: Conference or Workshop Item
• Published: Institute of Electrical and Electronics Engineers Inc. 2021
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85124889225&doi=10.1109%2fECBIOS51820.2021.9510570&partnerID=40&md5=f389d9d943d6b71fe6cad8b443328a64; http://eprints.utp.edu.my/29160/

This paper proposed a novel deep learning architecture, called the AMDOCT-NET architecture, to accurately detect age-related macular degeneration (AMD) on optical coherence tomography (OCT) images. Using the AMDOCT-NET architecture, the performance of various image processing, such as resizing, denoising, and cropping has been evaluated. The simulation results show that the AMDOCT-NET architecture with an input size of 224�224 pixels, no cropping, and no denoising achieves the accuracy of 99.09 to automatically detect the AMD. Compared with the well-known deep learning architecture, VGG16, the AMDOCT-NET improves accuracy by 2.09 and reduces the model size by 53.7. © 2021 ECBIOS 2021. All rights reserved.