Navigational outcomes with a depth-based vision processing algorithm in a second-generation suprachoroidal retinal...

Area of Science:

• Biomedical Engineering
• Ophthalmology
• Computer Vision

Background:

• Retinal prosthesis recipients often face challenges with navigation due to impaired vision.
• Existing vision processing (VP) methods may not fully restore environmental perception.
• Novel VP algorithms are needed to enhance spatial awareness and mobility for implant users.

Purpose of the Study:

• To compare the effectiveness of a new depth-based VP method, local background enclosure (LBE), against a standard method, Lanczos2 (L2).
• To evaluate performance in both controlled laboratory settings and complex real-world environments.
• To assess the impact of VP methods on obstacle detection and navigation for suprachoroidal retinal prosthesis users.

Main Methods:

• Four participants with suprachoroidal retinal prostheses were tested using both LBE and L2 VP methods.
• Laboratory trials involved navigating a white corridor with randomized obstacles.
• Real-world trials required participants to navigate complex urban scenes, identifying and avoiding obstacles.

Main Results:

• The LBE method showed significantly higher obstacle detection rates (63.6%) compared to L2 (48.5%) in the laboratory setting (p<0.001).
• In real-world scenarios, object identification accuracy was higher with LBE (50.2%) versus L2 (41.7%), with a risk difference of 8 percentage points (p=0.081).

Conclusions:

• The LBE vision processing algorithm offers improved performance for retinal prosthesis users.
• Depth-based visual cues provided by enhanced VP algorithms can significantly improve real-world navigation outcomes.
• Further development of advanced VP methods holds promise for restoring functional vision and mobility.