NETRA: Interactive Display for Estimating Refractive Errors and Focal Range

Abstract:
We introduce an interactive, portable, and inexpensive solution for estimating refractive errors in the human eye. While expensive optical devices for automatic estimation of refractive correction exist, our goal is to greatly simplify the mechanism by putting the human subject in the loop. Our solution is based on a high-resolution programmable display and combines inexpensive optical elements, interactive GUI, and computational reconstruction. The key idea is to interface a lenticular view-dependent display with the human eye at close range - a few millimeters apart. Via this platform, we create a new range of interactivity that is extremely sensitive to parameters of the human eye, such as the refractive errors, focal range, focusing speed, lens opacity, etc. We propose several simple optical setups, verify their accuracy, precision, and validate them in a user study.

Paper

* High Resolution (7MB)
* Low Resolution (600K)

Citation: Vitor F. Pamplona, Ankit Mohan, Manuel M. Oliveira, Ramesh Raskar. NETRA: Interactive Display for Estimating Refractive Errors and Focal Range. Proc. of SIGGRAPH 2010 (ACM Transactions on Graphics 29, 4), 2010.

Bibtex:

@article{Pamplona_et_al_2010,
author = {Vitor F. Pamplona and Ankit Mohan and Manuel M. Oliveira and Ramesh Raskar},
title = {NETRA: Interactive Display for Estimating Refractive Errors and Focal Range },
journal = {ACM Transactions on Graphics (proceedings of ACM SIGGRAPH)},
year = {2010},
volume = {29},
number = {4},
}

Videos
MIT Media Lab LabCast: Credit, Paula Aguilera and Jonathan Williams

IDG News Service Video: here

SIGGRAPH paper technical video: here.
Brief Technical Description
The NETRA system uses the dual of a Shack-Hartman sensor, and replaces the laser with simple user interaction. Todays methods using the Shack-Hartmann sensor shine a laser into the eye of the patient, and measure the reflected light with a wavefront sensor. Hence they are quite expensive, and require a trained professional operator. A cell phone based solution significantly reduces the cost of the device and makes it appropriate for self-evaluation, while still providing comparable data.

The subject looks into this display at a very close range and aligns (overlaps) displayed patterns (Figure 1). Since the light rays from these patterns pass through different regions of the visual system, the alignment task gives a measure of the optical distortions of those regions. The subject repeats this procedure for a few meridians with appropriate variation in the patterns. The system computes the corresponding refractive error for myopia, hyperopia and astigmatism.

Prototype Nexus One Prototype
Figure 2: Current prototypes using the Samsung Behold II and the Nexus One. We place an optical phase plate to create virtual images and achieve 0.6 and 0.4 diopter resolution respectively.

Evaluation: We tested accuracy and precision of the technique in two experiments: (i) using lenses and a SLR camera and (ii) comparing our device against actual prescriptions in a user study. The resolution is 0.4 diopters using the Nexus One device (focal length 30mm). The Apple iPhone 4G, with the new Retina Display should achieve a resolution of approximately 0.28 diopters (focal length 30mm). For measuring eye correction, the average absolute errors from the known prescriptions were under 0.5 diopter (s = 0.2) for both cylindrical and spherical powers. The average absolute error of our estimates of the cylindrical axis was under 6 degrees. Optometrists typically prescribe in multiples of 0.25 diopter, and 10 degrees axis.

In controlled user experiments with 16 subjects, the average absolute errors from the known prescriptions were under 0.5 diopter, with a standard deviation of 0.2 diopter for both cylindrical (astigmatism) and spherical powers (myopia and hyperopia). The average absolute error of the cylindrical axis is less than 6 degrees. We are able to achieve this without the use of cycloplegic eye drops for relaxing accommodation.

Limitations: Since our solution relies on subjective feedback, it cannot be used by individuals who cannot reliably perform the user-required tasks, such as very young children.
Existing Techniques
Existing systems to diagnose refractive eye conditions include Snellen charts (with a set of trial lenses), auto-refractometers, and wavefront aberrometers. The NETRA solution offers some unique benefits over these existing techniques, which make it specially suited for deployment in developing countries:

* Cost: NETRA relies on innovative use of existing hardware (a cell-phone), and custom software. We augment an existing cell-phone display with a device that costs as little as $2. This is significantly cheaper than sophisticated techniques such as aberrometers, and even a set of trial lenses which costs $100 or more.
* Safety: Since our device does not use lasers, has no need for cycloplegic drugs, and includes no moving parts, there are reduced concerns about danger in improper use, specialized training, or damage in transit.
* Speed: Conventional methods of diagnosis usually require two steps (step 1 objectively measures the refraction error, and step 2 verifies it based on patient feedback). Our hybrid approach combines these in a single user-driven step to obtain a measurement in less than 3 minutes.
* Accuracy: Unlike the Snellen chart, the NETRA system relies on a simple alignment task rather than the patient's ability to discern blur. This gives comparable accuracy, and a simpler user interaction task.
* Mobility: Based on a cell-phone, the NETRA system easily fits in a backpack for remote deployment.
* Self-evaluation: The ease of use, cheap cost, and the inherent safety of the NETRA system allows for (possibly at-home) self-evaluation.

Current solutions for analyzing refractive errors.
Figure 3: Current solutions for analyzing refractive errors. Subjective Methods (far left and center) rely upon the user's judgment of sharpness or blurriness of a test object. Objective Methods (far right) require a mechanically moving lens, a camera, a trained technician, and a large investment.

Refraction Services Requirement on Developing Countries [Vision 2020 Report]: The following table provides a comprehensive list of techniques and equipment for assessing refractive conditions of an eye.

Technique


Objectivity


Speed


Accuracy/ Reliability


Electricity Requirements


Mobility


Training


Equipment Requirements (Cost bracket)*


Cost Efficiency rank


Suitability for Children

Retinoscopy (Slit Lamp)


Objective = does not rely on patient responses


Fast


+/- 0.50D unless affected by media opacities or accommodation


Batteries


Good


High


Retinoscope, plus trial lens set and trial frame ($2000), OR phoropter ($1600), OR variable focus specs ($1600)


Economical � low up-front cost, high durability, low maintenance


Sometimes

Subjective refraction (Eye Charts)


Subjective = does rely on patient responses


Slow


+/- 0.25D but dependent on patient reliability


None


Good


High


Trial lens set and trial frame ($1400), OR phoropter ($1000), OR variable focus specs ($600)


Economical


Sometimes (only with experienced practitioners)

Auto Refraction


Objective


Fast


Relies on both equipment and patient factors


Mains


Low


Basic


Auto-refractor ($15K)


Expensive


No

Portable Auto Refraction


Objective


Fast


Relies on both equipment and patient factors


Mains or batteries


Good


Basic


Portable auto-refractor ($20K)


Expensive


No

NETRA


Subjective


Fast


Relies on both equipment and patient factors

< 0.50D unless affected by accommodation


Cell phone batteries


Excellent


Basic


Plastic Piece ($2) and a cell phone ($300).


Economical


No
* Costs were extracted from the Vision2020 report. Some cheaper options may exist - for example, we were able to acquire a set of trial lenses for $300. Note that simple reading charts can be expensive because they must be used under optimal lighting conditions and need a set of trial lenses.
Potential Impact
More than two billion people worldwide have refractive error. Very few have access to quality eye-care because existing solutions require a trained optometrist or expensive equipment [VISION2020 Report, Holden2007]. This impacts the developing world in a significant way:

* 517 million have uncorrected near-vision impairment affecting daily livelihood.
* 153 million have uncorrected far-vision impairment (affecting 2% of the world population).
* Uncorrected Refractive Errors are the 2nd leading cause of blindness* globally. 87% of those affected live in the developing world.
* For many children, hyperopia may remain undiagnosed, leading to undue stress and headaches.
* Access to a trained optometrist and equipment is extremely limited.

* WHO definition for blindness: vision worse that 3/60 in the better eye.

Uncorrected refractive problems may lead to a significant loss in productivity, with estimates ranging from USD 88.74 to USD 133 billion. To put things in perspective, this productivity loss exceeds the annual GDP of 46 of the 52 African countries. Our technology can address all types of refractive errors.

The PerfectSight initiative is a deployment platform which aims to bring NETRA to the hands of doctors and people worldwide. Our partners and two new team members - Margaret McKenna (IT consultant for developing contries) and Chika Ekeji (MIT Sloan) - will help us manufacture and deploy the device.

We have received excellent response, both to our technology and the deployment strategy:

* Accepted for publication at ACM SIGGRAPH 2010.
* Live demo at the 3rd workshop on Computer Vision Applications for the Visually Impaired (CVAVI 10) at CVPR 2010.
* Business model was a finalist in the development track of the MIT $100K Entrepreneurship Competition 2010.
* MIT IDEAS 2010 award, sponsored by the Lemelson-MIT program.
* Second place by popular vote in the MIT Global Challenge 2010.
* Citation by Susan Hockfield, president of MIT, in her Commencement speech.
* "An amazing innovation ... tremendous contibution to the challenges of the developing world", Frank Moss, Director of the Media Lab in his speech at the sponsor week.

Acknowledgments: Thanks to the volunteers who tested our device, Xiaoxi Wang and Andrew Song for prototype construction, Tiago Allen Oliveira for illustrations, Tyler Hutchison for the video voiceover, Taya Leary for her relentless support, the entire Camera Culture group for all the useful discussions; and the reviewers for their valuable feedback. Dr. James Kobler (Mass. General Hospital), Dr. Dr. Joseph Ciolino (Mass. Eye and Ear Infirmary), and Dr. Fuensanta Vera Diaz (Schepens Eye Research Institute) provided valuable resources and insightful discussions about optometry and ophthalmology. We thank Dick Lyon and Google's open source office for Google Nexus One mobile phone, and Samsung Electronics for Samsung Behold II mobile phone. Vitor and Manuel acknowledge CNPq-Brazil fellowships 142563/2008-0, 200763/2009-1, 200284/2009-6, 476954/2008-8. Ramesh Raskar is supported by an Alfred P. Sloan Research Fellowship.