EUVIMEDEUVIMED entdecken
Crossref

CNN-LSTM Model for Effective Glaucoma Detection from Fundus Images

Raghda M. Nasr, Manal A. Abdel-Fattah, Ahmed Samir Montasser

Journal of Innovative Image Processing · 11.07.2026 · 8 (3) · 894-915

DOI: 10.36548/jiip.2026.3.008 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Innovative Image Processing, online erschienen am 11.07.2026. Verfasst von Raghda M. Nasr und 2 weiteren Autor:innen. Die Arbeit nennt 23 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Glaucoma is one of the main causes of irreversible blindness, where delayed detection may increase the risk of permanent damage to the optic nerve. Developing accurate and computationally efficient computer-aided diagnostic systems using low-cost retinal fundus images remains challenging, especially with multi-disease datasets and limited clinical metadata. This paper proposes a hybrid CNN-LSTM multi-model for detecting glaucoma and six others ophthalmic diseases by merging retinal fundus images with clinically annotated patient symptoms. The model was trained and evaluated using the ODIR-5K dataset and expert-guided symptom annotation. A lightweight convolutional neural network was employed to extract visual features, while a Long Short-Term Memory network processed textual symptom descriptions. The extracted features were fused for classification across seven classes. The model achieved an accuracy of 92%, a macro-average F1-score of 74.7%, a weighted-average F1-score of 92%, and a macro-averaged AUC of 96.8% across seven predicted classes. Then, Explainable AI (XAI) techniques were applied to verify that diagnostic decisions were mainly driven by clinically relevant retinal regions. These results reveal that integrating visual and symptom-based information can improve retinal disease classification while maintaining computational efficiency, making the proposed framework suitable for deployment in limited-resource environments.

Zitieren

APA

Nasr, R. M., Abdel-Fattah, M. A., & Montasser, A. S. (2026). CNN-LSTM Model for Effective Glaucoma Detection from Fundus Images. Journal of Innovative Image Processing, 8(3), 894-915. https://doi.org/10.36548/jiip.2026.3.008

Vancouver

Nasr RM, Abdel-Fattah MA, Montasser AS. CNN-LSTM Model for Effective Glaucoma Detection from Fundus Images. Journal of Innovative Image Processing. 2026;8(3):894-915. doi:10.36548/jiip.2026.3.008

BibTeX
@article{Nasr2026,
  author = {Nasr, Raghda M. and Abdel-Fattah, Manal A. and Montasser, Ahmed Samir},
  title = {CNN-LSTM Model for Effective Glaucoma Detection from Fundus Images},
  journal = {Journal of Innovative Image Processing},
  year = {2026},
  volume = {8},
  number = {3},
  pages = {894--915},
  doi = {10.36548/jiip.2026.3.008}
}
Kostenlos starten →