Conference Paper2025

Assessing the Impact of Architectural Design in Deep Learning Models for Fundus-Based Ocular Disease Detection

Abdullah Al Alam Shanto, Mumtahina Esha, Sadia Alam Kotha, Riyadul Islam, Sefatul Wasi, Siam Tahsin Bhuiyan, Rashedur Rahman, Saadia Binte Alam

2025 IEEE 2nd International Conference on Computing, Applications and Systems (COMPAS)

IEEE, pp. 1-6

  • Good Health and Well-being — Relevance: 94%
DOI: 10.1109/compas67506.2025.11381612URL

Abstract

Ocular diseases such as Cataract, Age-Related Macular Degeneration, Diabetic Retinopathy, and uncorrected refractive errors like Myopia, constitute the leading causes of visual impairment globally, predominantly affecting the elderly populations. Early diagnosis is crucial to avert irreversible vision loss. Conventional diagnostic techniques like fundus fluorescein angiography (FFA), optical coherence tomography (OCT), and fundus image analysis can be resource-intensive and reliant on the expertise of specialists. In recent times, different automated methods for diagnosis have been developed to help specialists in their work. This study presents a comparative analysis of three deep learning models: EfficientNetB3 (CNN), ViT-B-16 (Vision-Transformer), and ConvNeXtXLarge (CNN inspired by Transformers), for multiclass ocular disease classification utilizing the ODIR-5K dataset. Among the three evaluated architectures, ConvNeXtXLarge demonstrated the highest performance, achieving an area under the receiver operating characteristic curve (AUROC) of 0.892 and EfficientNetB3 achieved an AUROC of 0.870, and ViT-B-16 achieved an AUROC of 0.864. Regarding classification accuracy, ConvNeXt-XLarge reached 73.6%, exceeding EfficientNetB3 by 4.0% and outperforming ViT-B-16 by 2.0%. Grad-CAM visualizations provided additional insight since the heatmaps generated by EfficientNetB3 appeared to align more closely with clinically relevant regions than those produced by ConvNeXtXLarge or ViT-B-16. This suggests that, while ConvNeXtXLarge and ViT-B-16 show better scores, EfficientNet gives a better clinically aligned decision. Architectural design plays a pivotal role in shaping both diagnostic accuracy and the transparency of models in ophthalmic imaging, as demonstrated by this comprehensive evaluation that combines quantitative metrics with interpretability insights.

Keywords

  • diabetic retinopathy
  • myopia
  • age-related macular degeneration
  • EfficientNetB3
  • eye disease diagnosis
  • fundus images
  • cataract
  • model interpretability
  • Grad-CAM
  • vision transformer

CCDS Authors

Riyadul Islam
Riyadul Islam

Riyadul Islam

  • Intern, CCDS

Bachelors

Daffodil International University

Published at

Scientific Reports

Siam Tahsin Bhuiyan
Siam Tahsin Bhuiyan

Siam Tahsin Bhuiyan

  • Intern, CCDS

Bachelors

Independent University, Bangladesh

Published at

Scientific Reports

Md. Rashedur Rahman, D. Eng.
Md. Rashedur Rahman, D. Eng.

Md. Rashedur Rahman, D. Eng.

  • Co-Director, MIRA & NEST Wings

Assistant Professor

Department of Computer Science and Engineering

Independent University, Bangladesh

Published at

CHI, HCII, Scientific Reports, Cognitive Computation, Neural Computing and Applications

Saadia Binte Alam, PhD
Saadia Binte Alam, PhD

Saadia Binte Alam, PhD

  • Director, MIRA & WiSE Wings

Associate Professor

Department of Computer Science and Engineering

Independent University, Bangladesh

Published at

HCII, Scientific Reports, Cognitive Computation, Neural Computing and Applications

References

  1. 1.Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, Li Fei-Fei. (2009). ImageNet: A large-scale hierarchical image database. 2009 IEEE Conference on Computer Vision and Pattern Recognition, 248–255[10.1109/cvpr.2009.5206848]
  2. 2.Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie. (2022). A ConvNet for the 2020s. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 11966–11976[10.1109/cvpr52688.2022.01167]
  3. 3.Zewen Li, Fan Liu, Wenjie Yang, Shouheng Peng, Jun Zhou. (2021). A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects. IEEE Transactions on Neural Networks and Learning Systems, 33(12), 6999–7019[10.1109/tnnls.2021.3084827]
  4. 4.Jaimie D Steinmetz, Rupert R A Bourne, Paul Svitil Briant, Seth R Flaxman, Hugh R B Taylor, Jost B Jonas, Amir Aberhe Abdoli, Woldu Aberhe Abrha, Ahmed Abualhasan, Eman Girum Abu-Gharbieh, Tadele Girum Adal, Ashkan Afshin, Hamid Ahmadieh, Wondu Alemayehu, Sayyed Amirpooya Samir Alemzadeh, Ahmed Samir Alfaar, Vahid Alipour, Sofia Androudi, Jalal Arabloo, Aries Berhe Arditi, Brhane Berhe Aregawi, Alessandro Arrigo, Charlie Ashbaugh, Elham Debalkie Ashrafi, Desta Debalkie Atnafu, Eleni Amin Bagli, Atif Amin Winfried Baig, Till Winfried Bärnighausen, Maurizio Battaglia Parodi, Mahya Srikanth Beheshti, Akshaya Srikanth Bhagavathula, Nikha Bhardwaj, Pankaj Bhardwaj, Krittika Bhattacharyya, Ali Bijani, Mukharram Bikbov, Michele Bottone, Tasanee M Braithwaite, Alain M Bron, Sharath A Burugina Nagaraja, Zahid A Butt, Florentino Luciano L Caetano dos Santos, Vera L James Carneiro, Robert James Casson, Ching-Yu Jasmine Cheng, Jee-Young Jasmine Choi, Dinh-Toi Chu, Maria Vittoria M Cicinelli, João M G Coelho, Nathan G A Congdon, Rosa A A Couto, Elizabeth A M Cromwell, Saad M Dahlawi, Xiaochen Dai, Reza Dana, Lalit Dandona, Rakhi A Dandona, Monte A Del Monte, Meseret Derbew Molla, Nikolaos Alemayehu Dervenis, Abebaw Alemayehu P Desta, Jenny P Deva, Daniel Diaz, Shirin E Djalalinia, Joshua R Ehrlich, Rajesh Rashad Elayedath, Hala Rashad B Elhabashy, Leon B Ellwein, Mohammad Hassan Emamian, Sharareh Eskandarieh, Farshad G Farzadfar, Arthur G Fernandes, Florian S Fischer, David S M Friedman, João M Furtado, Shilpa Gaidhane, Gus Gazzard, Berhe Gebremichael, Ronnie George, Ahmad Ghashghaee, Syed Amir Gilani, Mahaveer Golechha, Samer Randall Hamidi, Billy Randall R Hammond, Mary Elizabeth R Kusuma Hartnett, Risky Kusuma Hartono, Abdiwahab I Hashi, Simon I Hay, Khezar Hayat, Golnaz Heidari, Hung Chak Ho, Ramesh Holla, Mowafa J Househ, John J Emmanuel Huang, Segun Emmanuel M Ibitoye, Irena M D Ilic, Milena D D Ilic, April D Naghibi Ingram, Seyed Sina Naghibi Irvani, Sheikh Mohammed Shariful Islam. (2020). Causes of blindness and vision impairment in 2020 and trends over 30 years, and prevalence of avoidable blindness in relation to VISION 2020: the Right to Sight: an analysis for the Global Burden of Disease Study. The Lancet Global Health, 9(2), e144–e160[10.1016/s2214-109x(20)30489-7]
  5. 5.Matthew J Burton, Jacqueline Ramke, Ana Patricia Marques, Rupert R A Bourne, Nathan Congdon, Iain Jones, Brandon A M Ah Tong, Simon Arunga, Damodar Bachani, Covadonga Bascaran, Andrew Bastawrous, Karl Blanchet, Tasanee Braithwaite, John C Buchan, John Cairns, Anasaini Cama, Margarida Chagunda, Chimgee Chuluunkhuu, Andrew Cooper, Jessica Crofts-Lawrence, William H Dean, Alastair K Denniston, Joshua R Ehrlich, Paul M Emerson, Jennifer R Evans, Kevin D Frick, David S Friedman, João M Furtado, Michael M Gichangi, Stephen Gichuhi, Suzanne S Gilbert, Reeta Gurung, Esmael Habtamu, Peter Holland, Jost B Jonas, Pearse A Keane, Lisa Keay, Rohit C Khanna, Peng Tee Khaw, Hannah Kuper, Fatima Kyari, Van C Lansingh, Islay Mactaggart, Milka M Mafwiri, Wanjiku Mathenge, Ian McCormick, Priya Morjaria, Lizette Mowatt, Debbie Muirhead, Gudlavalleti V S Murthy, Nyawira Mwangi, Daksha B Patel, Tunde Peto, Babar M Qureshi, Solange R Salomão, Virginia Sarah, Bernadetha R Shilio, Anthony W Solomon, Bonnielin K Swenor, Hugh R Taylor, Ningli Wang, Aubrey Webson, Sheila K West, Tien Yin Wong, Richard Wormald, Sumrana Yasmin, Mayinuer Yusufu, Juan Carlos Silva, Serge Resnikoff, Thulasiraj Ravilla, Clare E Gilbert, Allen Foster, Hannah B Faal. (2021). The Lancet Global Health Commission on Global Eye Health: vision beyond 2020. The Lancet Global Health, 9(4), e489–e551[10.1016/s2214-109x(20)30488-5]
  6. 6.Meindert Niemeijer, Bram van Ginneken, Stephen R. Russell, Maria S. A. Suttorp-Schulten, Michael D. Abra`moff. (2007). Automated Detection and Differentiation of Drusen, Exudates, and Cotton-Wool Spots in Digital Color Fundus Photographs for Diabetic Retinopathy Diagnosis. Investigative Ophthalmology & Visual Science, 48(5), 2260[10.1167/iovs.06-0996]
  7. 7.Ning Li, Tao Li, Chunyu Hu, Kai Wang, Hong Kang. (2021). A Benchmark of Ocular Disease Intelligent Recognition: One Shot for Multi-disease Detection. Lecture notes in computer science, 177–193[10.1007/978-3-030-71058-3_11]
  8. 8.Qi Bi, Xu Sun, Shuang Yu, Kai Ma, Cheng Bian, Munan Ning, Nanjun He, Yawen Huang, Yuexiang Li, Hanruo Liu, Yefeng Zheng. (2023). MIL-ViT: A multiple instance vision transformer for fundus image classification. Journal of Visual Communication and Image Representation, 97, 103956[10.1016/j.jvcir.2023.103956]
  9. 9.Kyoung Min Kim, Tae-Young Heo, Aesul Kim, Joohee Kim, Kyu Jin Han, Jaesuk Yun, Jung Kee Min. (2021). Development of a Fundus Image-Based Deep Learning Diagnostic Tool for Various Retinal Diseases. Journal of Personalized Medicine, 11(5), 321[10.3390/jpm11050321]
  10. 10.Lamiaa Abdel-Hamid. (2019). Glaucoma Detection from Retinal Images Using Statistical and Textural Wavelet Features. Journal of Imaging Informatics in Medicine, 33(1), 151–158[10.1007/s10278-019-00189-0]
  11. 11.Yusaku Katada, Nobuhiro Ozawa, Kanato Masayoshi, Yoshiko Ofuji, Kazuo Tsubota, Toshihide Kurihara. (2020). Automatic screening for diabetic retinopathy in interracial fundus images using artificial intelligence. Intelligence-Based Medicine, 3-4, 100024[10.1016/j.ibmed.2020.100024]
  12. 12.Usharani Bhimavarapu, Nalini Chintalapudi, Gopi Battineni. (2024). Automatic Detection and Classification of Hypertensive Retinopathy with Improved Convolution Neural Network and Improved SVM. Bioengineering, 11(1), 56[10.3390/bioengineering11010056]
  13. 13.Amrit Das, Rohan Giri, Gunjan Chourasia, A. Anilet Bala. (2019). Classification of Retinal Diseases Using Transfer Learning Approach. , 2080–2084[10.1109/icces45898.2019.9002415]
  14. 14.Maonian Wu, Ying Lu, Xiangqian Hong, Jie Zhang, Bo Zheng, Shaojun Zhu, Naimei Chen, Zhentao Zhu, Weihua Yang. (2022). Classification of dry and wet macular degeneration based on the ConvNeXT model. Frontiers in Computational Neuroscience, 16, 1079155[10.3389/fncom.2022.1079155]
  15. 15.Abdullah Ali Jabber, Ahmed Hadi, Salim Wadi, Ghada Ali. (2024). Cataract Detection and Classification Using Deep Learning Techniques. International Journal of Computing and Digital Systems, 17(1), 1–10[10.12785/ijcds/1571037888]
  16. 16.Taiseer Rakiin Ahad, Saadia Binte Alam, Farhan Ha-Mim, Ashad Nur, Sefatul Wasi, Rashedur Rahman. (2024). Comparative Analysis of Deep Learning Models for Intracerebral Hemorrhage Detection from Computed Tomography. , 328–333[10.1109/iccit64611.2024.11022361]
  17. 17.Joyita Faruk, Saadia Binte Alam, Sk Sadia Tasnim Elma, Sefatul Wasi, Rashedur Rahman, Syoji Kobashi. (2024). Kidney Abnormality Detection Using Segmentation-Guided Classification on Computed Tomography Images. , 414–419[10.1109/icmlc63072.2024.10935113]
← All publications