Development of a cloud-based estimator for analysing the disc-fovea angle in fundus images using a stacking ensemble...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• The disc-fovea angle (DFA) is a critical parameter in fundus image analysis.
• Accurate and efficient DFA measurement is essential for diagnosing and managing various ocular conditions.

Purpose of the Study:

• To develop a cloud-based software for automated disc-fovea angle (DFA) analysis in fundus images.
• To evaluate the accuracy and efficiency of this automated DFA assessment compared to manual analysis.

Main Methods:

• A retrospective study included 682 fundus images from healthy individuals and patients with cyclotropia.
• A stacking ensemble model, integrating object detection and machine learning algorithms, was developed for automatic DFA analysis.
• Manual DFA analysis was performed by two independent examiners using web-based software, with the average value considered the ground truth.

Main Results:

• The automatic DFA measurements showed no significant difference compared to manual analysis (p=0.52).
• Automated analysis demonstrated a significantly shorter processing time per image (0.430±0.240 s) compared to manual analysis (15.794±1.558 s) (p<0.001).

Conclusions:

• The developed cloud-based software utilizing a stacking ensemble model provides accurate DFA measurements.
• Automated DFA analysis offers a significant time-saving advantage, making it a valuable tool for clinical applications.