AI tools demonstrate effectiveness at helping people with vision loss complete tasks
Key takeaways:
- AI apps and glasses had varying effectiveness in 14 daily living tasks but were well received by participants.
- AI technologies were most useful for text-related tasks.
People with vision loss, or PVL, were satisfied with the use of AI technologies to complete vision-related activities of daily living, according to a cross-sectional, counterbalanced, crossover study.
Such findings could ultimately help “to guide technology recommendations based on PVL clinical characteristics and functional needs,” the study authors noted.
“With advancements in computing power and digital imaging, the integration of AI into portable tools such as smart glasses and smartphone apps is shaping the future of assistive technology,” study author William H. Seiple, PhD, chief research officer at Lighthouse Guild and a research professor at NYU Grossman School of Medicine, said in a Lighthouse Guild press release.
According to Seiple and colleagues, there are issues regarding the use of AI in people with vision loss, such as types of information and sensors needed, the nature of the user interface and what algorithm should be used, and “proposed solutions have been appearing rapidly.”
Yet, “the functionality of AI in the hands of PVL must be objectively quantified,” they wrote in Translational Vision Science & Technology.
In the analysis, Seiple and colleagues evaluated two smart glasses, the OrCam and Envision Glasses, and two AI apps, Google Lookout and Seeing AI, based on 14 daily living activities in 25 people with vision loss. These technologies were referred to as assistive artificial intelligence implementations (AAIIs).
The categories of tasks evaluated included:
- searching and identifying, such as identifying a color, describing a room or finding a person;
- reading text, such as reading an article, street sign, medicine bottle or handwriting; and
- reading text in columns, such as text in a TV guide or table of contents.
The data on AAII effectiveness were collected across five 90-minute sessions, including a baseline session, in which usual refraction and no AI was used for tasks, and four sessions in which one of the four AAIIs was used.
Seiple and colleagues reported that the odds ratios for completing text tasks were substantially higher than at baseline in five of the six tasks with OrCam and all six tasks with Envision, Seeing AI and Lookout.
The impact of the AAIIs on the six searching and identifying tasks were more varied. The odds ratios for completing a task were better than at baseline for five tasks with Seeing AI, four tasks with Envision, three tasks with OrCam and two tasks with Lookout.
Participants expressed satisfaction with all the AAIIs, with a median Likert score of 5 out of 5 for Seeing AI, 4 out of 5 for OrCam, and 3 out of 5 for Envision and Lookout.
The researchers acknowledged some study limitations, such as potential volunteer bias. They also did not know if outcomes would have been different if participants used the AAIIs for an extended amount of time at home.
“These technologies hold tremendous promise for creating a more inclusive, accessible world for PVL,” Seiple said in the release.
Reference:
- Lighthouse Guild study reveals AI’s transformative benefits for individuals with vision loss. https://lighthouseguild.org/news/lighthouse-guild-study-reveals-ais-transformative-benefits-for-individuals-with-vision-loss/. Published Jan. 27, 2025. Accessed March 20, 2025.
Perspective
Back to Top
As a doctor of optometry who serves patients with a variety of functional limitations because of their vision loss, I find this study by Seiple and colleagues quite encouraging for the future possibilities in low vision rehabilitation. This study compared four types of AI integration in assistive technology. Systematic testing revealed that all four technologies significantly improved reading ability, with the free Seeing AI app showing the best overall performance for both reading and object recognition tasks. What makes these findings particularly useful for clinicians is how the study addressed both subjective and objective outcomes in using the AI tools; this data allow for a better “real-world” understanding of the technology utilization.
Also, I am especially interested in the cost implications of this study’s findings. Free or low-cost AI apps seem to outperform expensive low vision devices across a majority of the tasks. While the study shows that current AI technologies still struggle with complex visual tasks beyond reading, patients reported high satisfaction with all the tools tested. This research comes at an exciting time as AI capabilities are advancing rapidly, suggesting that AI-powered solutions may fundamentally improve independence for people with vision loss.
Low vision care has the potential for exponential growth with AI and AI integration into assistance devices. I hope further research continues to investigate ease of use and quality of life improvement with other technologies such as the Ray-Ban Meta glasses, and I would like to see continued comparisons with more traditionally branded low vision devices. Low vision devices generally have a heavy cost burden; improved price points with the types of technologies studied in this article will likely allow for more people with vision loss better access to our world.
Collapse
Read more about