Table 3 Summary of studies utilizing multimodal AI approaches in diabetic retinopathy
Year | Author | Task | Multimodal data types | Dataset scale | Dataset availability |
|---|---|---|---|---|---|
2020 | Li X et al. [58] | Fundus disease classification | Fundus images, synthesized FFA images | 1200 images in [55]; 1200 images in [56]; 88,702 images in [57] | Free available after registration in https://ichallenges.grand-challenge.org/, https://www.kaggle.com/competitions/diabetic-retinopathy-detection/data |
2021 | He X et al. [59] | Fundus disease classification | Fundus images, OCT images | 933 eyes of 498 patients | Private dataset |
2021 | Li X et al. [60] | Retinal disease Recognition | Fundus images, OCT images | 1,193 eyes of 836 subjects | Private dataset |
2022 | Hervella Á et al. [61] | Diabetic retinopathy classification | Fluorescein angiography and color retinography | 59 multimodal image pairs | Freely available from http://misp.mui.ac.ir/data/eye-images.html |
2023 | El Habib Daho M et al. [65] | Diabetic retinopathy classification | Ultra-widefield color fundus images and OCTA images | 875 eyes from 444 patients | Upon request |
2023 | Li Y et al. [66] | Detection of proliferative diabetic retinopathy | OCT and OCTA images, fundus images | 64 patients with diabetes | Private dataset |
2024 | Bidwai P et al. [67] | Diabetic classification | OCTA images and fundus images | 222 images of 76 people | Upon request |