Multidimensional imaging model for diagnosing primary open-angle glaucoma using fundus photographs and optical...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Primary open-angle glaucoma (POAG) is a leading cause of irreversible blindness worldwide.
• Early detection and accurate diagnosis are crucial for effective management and prevention of vision loss.
• Current diagnostic methods may have limitations in detecting early structural damage, especially in complex cases like myopia.

Purpose of the Study:

• To develop and evaluate a novel multidimensional imaging model for enhanced POAG detection.
• To integrate multiple imaging modalities including fundus photographs, RNFL thickness maps, GCIPL thickness maps, and deviation maps.
• To assess the model's diagnostic performance against unimodal models and human experts.

Main Methods:

• A convolutional neural network-based multidimensional imaging model was developed using a dataset of 1054 eyes.
• The model combined fundus photographs, RNFL, and GCIPL thickness and deviation maps.
• Performance was evaluated using area under the receiver operating characteristic curve (AUC) on a held-out test dataset.

Main Results:

• The multidimensional imaging model achieved a superior AUC of 0.970, outperforming models using only fundus photographs (AUC 0.945) or RNFL/GCIPL maps (AUC 0.931-0.958).
• The model's performance was significantly better than that of three glaucoma ophthalmologists (AUCs ranging from 0.755 to 0.847).
• The model demonstrated robust performance across different myopia levels, showing promise for diagnosing myopia combined with POAG.

Conclusions:

• The developed multidimensional imaging model offers improved diagnostic accuracy for POAG compared to unimodal approaches.
• This AI-driven tool has the potential to significantly aid clinicians in diagnosing POAG, particularly in cases with co-existing myopia.
• The findings suggest a valuable advancement in ophthalmic diagnostics, potentially leading to better patient outcomes.