Automated Detection of Central Retinal Artery Occlusion Using OCT Imaging via Explainable Deep Learning

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Central retinal artery occlusion (CRAO) is a critical condition causing sudden vision loss.
• Accurate and rapid diagnosis of CRAO is essential for timely intervention and visual preservation.
• Optical Coherence Tomography (OCT) provides detailed retinal imaging crucial for diagnosis.

Purpose of the Study:

• To evaluate the capability of a deep learning model in detecting CRAO using OCT data.
• To assess the model's performance in differentiating CRAO from other causes of acute visual impairment.

Main Methods:

• A retrospective external validation study utilized OCT and clinical data from two German university medical centers.
• A deep learning classification model was developed and trained on expertly graded OCT data.
• The model employed a multiclass, five-fold cross-validation scheme to classify control, CRAO, and differential diagnosis groups.

Main Results:

• The deep learning model achieved high performance metrics across all classes.
• Area Under the Curve (AUC) values were 0.96 for CRAO, 0.99 for controls, and 0.90 for differential diagnoses.
• The model demonstrated strong sensitivity and specificity in identifying CRAO and related conditions.

Conclusions:

• The developed machine learning algorithm (MLA) shows significant potential for accurate CRAO detection.
• This technology can aid in the rapid identification of less common etiologies in emergency settings.
• Deployment of MLAs could enhance diagnostic capabilities for urgent retinal pathologies.