Glaucoma detection using non-perfused areas in OCTA.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computational Biology

Background:

• Ophthalmic diseases cause reduced capillary perfusion, often visualized with optical coherence tomography angiography (OCTA).
• Traditional vessel density analysis on OCTA images lacks sensitivity for early pathological changes.
• Current methods quantifying non-perfused areas depend heavily on accurate vessel segmentation, a significant challenge.

Purpose of the Study:

• To develop a novel, robust method for quantifying capillary perfusion changes in ophthalmic diseases.
• To improve the accuracy and reliability of distinguishing between glaucoma patients and healthy controls using OCTA.
• To offer a computationally efficient and explainable alternative to deep learning for OCTA analysis.

Main Methods:

• Utilized perfusion-distance measures to compute intercapillary areas, reducing sensitivity to vessel segmentation errors.
• Developed a novel classification method based on features from the probability density function of perfusion-distance areas.
• Evaluated the method on various capillary plexuses and compared its performance against existing techniques and deep learning models.

Main Results:

• The proposed method demonstrated superior performance in classifying glaucoma patients compared to previous handcrafted feature methods.
• Achieved classification accuracy comparable to deep learning models trained on raw OCTA images.
• Highlighted the robustness of perfusion-distance measures against segmentation inaccuracies.

Conclusions:

• The novel perfusion-distance based method provides an accurate, efficient, and explainable approach for analyzing OCTA images in ophthalmic diseases.
• This technique offers a reliable alternative to complex deep learning models for clinical applications.
• The findings suggest a promising direction for early detection and monitoring of glaucoma and other related conditions.