Deep learning approach may help expedite screening of late-stage AMD

Due to the large amount of data generated from optical coherence tomography angiography (OCTA) used in the diagnosis of age-related macular degeneration (AMD), utilizing artificial intelligence to analyze the data is needed.

A deep learning (DL) approach utilizing a combination of OCTA, OCT structure, 2D b-scan flow images, and high definition (HD) 5-line b-scan cubes can be used to detect AMD class as well as determine non-AMD vs. non-neovascular (NNV) AMD vs. NV AMD.

Multimodal data were used as input to 2D-3D Convolutional Neural Networks (CNNs).

Choroidal neovascularization and geographic atrophy were identified as important biomarkers for AMD for CNNs and experts. CNNs predict biomarkers with accuracy up to 90.2% (positive-predictive-value up to 75.8%).

“Detection of AMD and its biomarkers from OCTA data via CNNs has tremendous potential to expedite screening of early and late-stage AMD patients,” the authors concluded.

Reference
Thakoor KA, Yao J, Bordbar D, et al. A multimodal deep learning system to distinguish late stages of AMD and to compare expert vs. AI ocular biomarkers. Sci Rep. 2022;12(1):2585. doi: 10.1038/s41598-022-06273-w. PMID: 35173191; PMCID: PMC8850456.