Blind Source Separation of Retinal Pulsatile Patterns in Optic Nerve Head Video-Recordings

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Medical Imaging

Background:

• Dynamic optical imaging of retinal hemodynamics is crucial for vision and eye-disease research.
• Current methods for analyzing retinal hemodynamic patterns are manual or semi-automated.
• Video-ophthalmoscopy (VO) recordings capture distinct functional phenomena like spontaneous venous pulsations (SVP).

Purpose of the Study:

• To propose and validate a novel concept using blind source separation (BSS) for automated localization of distinct hemodynamically synchronized areas in retinal video-imaging.
• To assess the feasibility of BSS techniques combined with K-means post-processing for analyzing optic nerve head (ONH) hemodynamics.

Main Methods:

• Applied spatial principal component analysis (sPCA) and spatial independent component analysis (sICA) as BSS techniques.
• Utilized a K-means based post-processing method on monocular and binocular VO recordings of the ONH in healthy subjects.
• Analyzed dynamic patterns, including heart rate and low-frequency oscillations.

Main Results:

• BSS automatically identified three reproducible areas: SVP, optic cup pulsations (OCP), and other pulsations (OP).
• K-means post-processing effectively reduced spike noise while preserving signal integrity.
• Significant phase shifts were observed between SVP, OCP, and OP, with heart rate-related dynamics and potential respiratory effects detected.

Conclusions:

• BSS offers a pioneering, automated approach for localizing distinct hemodynamically active regions in retinal video-imaging.
• The developed method accurately identifies and characterizes hemodynamic patterns in the ONH, providing insights into ocular blood flow.
• This technique holds promise for advancing research in vision and eye diseases by enabling objective analysis of retinal hemodynamics.