A Comprehensive Study on Multimodel Imaging for Early Detection of Ocular Disease

EasyChair Preprint 15585

Abstract

Age-related macular degeneration (ARMD), caused by age-specific retinal damage, is one of the primary causes of blindness. As individuals age, ARMD can impair daily activities and lead to severe vision loss if untreated. Researchers use deep learning (DL) models with OCT and fundus images to diagnose ARMD. This study explores the benefits of combining these imaging modalities with DL approaches, using the Project Macula dataset to develop models with transfer learning. This technique allows the models to adapt to new inputs and learn from diverse data. The study focuses on precision medicine and digital twin concepts, using a dataset of OCT and color fundus images. The attributes were carefully chosen to represent a variety of ARMD cases. A network architecture ensemble was proposed to integrate OCT and fundus data, enhancing prediction accuracy. The dataset was divided into subgroups for testing, validation, and training, with preprocessing to standardize inputs. Experiments compared combined data with individual modalities and other algorithms. Metrics like recall, accuracy, and precision demonstrated the model's effectiveness. This combined approach consistently outperformed individual modalities, highlighting the advantages of data fusion in DL techniques. The findings support automating ARMD screening and advancing computer-aided diagnosis tools, improving patient outcomes.

Keyphrases: AI, ARMD, Color Fundus, Multimoda, OCT Fundus, Ocular Disease

@booklet{EasyChair:15585,
  author    = {Hira Farooq and M Azam Zia and Muhammad Asif},
  title     = {A Comprehensive Study on Multimodel Imaging for Early Detection of Ocular Disease},
  howpublished = {EasyChair Preprint 15585},
  year      = {EasyChair, 2024}}