Home   |   News   |   Join Us   |   People   |   Projects   |   Publications   |   Talks   |   Courses


CATRA: Cataract Maps with Snap-on
Eyepiece for Mobile Phones




Vitor Pamplona        Erick Passos       Jan Zizka        Manuel M. Oliveira
Everett Lawson          Esteban Clua       Ramesh Raskar

Camera Culture Group - MIT Media Lab
UFRGS Instituto de Informática
UFF Media Lab

See our latest work on tailored displays


SIGGRAPH Paper      Vitor's PhD. Thesis     Frequently Asked Questions     Hi-Res Free Images    Collaboration


CATRA

CATRA
CATRA
Figure 1: Can we create a device that makes people aware of their early cataract condition? Using a light-field display, our method projects time-dependent patterns onto the fovea. Interactive software measures the visibility and point spread function across subapertures of the crystallin lens. By repeating this procedure for several light-paths, the cataracts size, position, density, and scattering profile are estimated. Ilustration: Tiago Allen

Abstract

Cataracts are the leading cause of blindness worldwide.  Existing diagnostic methods require Slit-lamps. We propose (CATRA) a new solution to detect and quantify cataracts with a compact eyepiece attached to a cell phone. With no moving parts and built from off-the-shelf components, our solution is well suited for the developing world.

A cataract-affected eye scatters and refracts light before it reaches the retina, caused by a fogging or clouding of the lens. We measure this deformation or (clouding) by allowing one to compare a good light path with a light path blocked by the cataract. Current methods for cataract detection require costly equipment and highly trained clinicians. They utilize back-scattering which is observed and subjectively diagnosed. However, this does not address the early onset of cataract affected vision, as early opacities are difficult to detect. Back scattering can be misleading as it does not account for what the patient actually sees.  Existing techniques present a simple grading of severity, while our technique presents a full map of opacity and scattering.  Our technique allows for a coupling of quantifiable data with the users visual experience.

CATRA utilizes a forward scattering technique, which allows the user to respond to what they visually experience.  Our device scans the lens section by section. The user sees our projected patterns and presses a few buttons to map the light attenuation in each section of the eye.  This information is collected by the device creating an attenuation map of the entire lens.  This allows individuals to monitor the progression of the severity of the cataract.  Our maps capture a full point spread function of the lens, allowing us to simulate the visual perception of a cataract affected subject over time.  Early cataract onset is difficult to diagnose.  We present a device for measuring cataracts, which is highly portable and collects quantifiable data to help tackle a global health problem making it ideal for the developing worl

Please look at FAQ for more information. Collaboration opportunities and clinical study plans,
please introduce yourself at our google group.

Paper

Videos








Awards


Vodafone's 1st prize

PopScience's Best of What is New 2011

MIT IDEAS Winner

 MIT Ideas 2011 Winner Award
MIT Global Challenge 2011 Public Choice Award

Brief Technical Description

CATRA goes beyond traditional subjective cataract evaluation procedures (slit-lamps plus optometrists) by taking advantage of forward scattering to compute quantitative maps for opacity, attenuation, contrast, and point-spread function of a cataract-affected eye. Similar NETRA, our approach adds simple optical components on cell phones to create collimated beams of light to scan the crystallin lens. Placed close to the viewers’ eye, these beams scatter when the light path hits a cataract, changing the Subject's view. Subject marks scattering regions and performs subaperture visual acuity tests to generate the desired maps. Our current prototype uses the Samsung Behold II cell phone and reaches an accuracy of about 0.4mm in the position and 0.1mm2 in size of the opacity on early cataract condition of elderly patients. The cost of the external white optical piece is under US $2.

Figure 2 illustrates our interactive method. By measuring the pupil size, which defines the discretization of the pupil area and enables the computation of the cataract size in meaningful physical units, we sequentially scan the subject’s crystallin to identify the presence of cataracts. If this is found to be true, the subject identifies the position of opacities and in a posterior step measures the light attenuation for each sub-aperture of the eye, thus creating opacity and attenuation maps. Measured attenuation values estimate the intensity of the local PSF peak. The subject then performs perceptual pattern matching to measure the tail of the PSF. If the attenuation value is small, the tail may be bigger than the fovea, and its direct measurement is not reliable. Contrast-sensitivity tests, described later, give us an approximation for the PSF. For each step of the test, the user’s role is to match patterns which are projected onto the fovea.

 
Figure 2: Overall hierarchical method to efficiently measure cataracts. (a) Software automatically scans the lens to check for the presence of cataracts. (b) If a scattering spot is found, the scanning procedure is repeated with the subject’s feedback. (c) By matching the brightness of two alternating paths of light we compute an attenuation map. (d) For a high scattering spot, the local contrast-sensitivity test replaces the local PSF measurement. In this case the subject increases the contrast of the displayed pattern up to a point where the letter becomes discernible. (e) Local PSF matching is the most detailed mapping, where the peak and Gaussian spread are measured for each scattering spot. The four maps together summarize the forward scattering effects of cataracts.

Rendering Subject's View: We propose an image-based approach for simulating the vision of a specific individual affected by cataracts. A depth-accommodation-dependent convolution of sub-aperture PSFs simulates the view of a cataract-affected eye. We convolve depth-masked patches of the input image with their corresponding PSFs and combine the results into the final image. Each depth-based PSF is computed by combining the sub-aperture PSFs, which are given by the scattering profile of an eye lens. Figure 3 shows a simulated night-driving scene with the experimental data used to render it. Cataract shape (b) can be seen as a mask on the ”bokeh” effect of the PSF composition.

 
Figure 3: Simulation of a cataract-affected view and its progression over time.

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. 18 subjects tested our prototype based on a cell-phone display. Each subject took the test twice for training, and at least twice for data collection. Each map measures the observed attenuation for 24 testing points. A group of 5 early cataract-affected volunteers (ages 68 to 76 and one 30-year-old) took the initial scanning procedure. Estimated total size for their cataracts is 1.46, 0.86, 0.82, 0.46 and 0.34mm2 . Their average error in position is 0.18mm ± 0.01, 0.43mm ± 0.18, 0.40mm ± 0.15, 0.60mm ± 0.36 and 0.32mm±0.10, and in size is 0.08mm2 ±0.01, 0.11mm2 ±0.09, 0.17mm2 ± 0.10, 0.27mm2 ± 0.14 and 0 respectively. 2 early cataract-affected volunteers took the brightness matching test. Average error is 12.2917% ± 0.01% and 6.38% ± 3.64%. Additionally, 14 healthy eyes were scanned and no cataract was found. We notice that optometrists have no such information nowadays.

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.

Comparison with Existing Techniques

Cataracts are generally detected subjectively by locating a white reflex during a slit lamp examination. Cataracts can be assessed by backscattering or forward scattering analysis (Table 1).

Backscattering examination: A slit-lamp microscope is used to backscatter light from cataract spots. This technique, however, requires numerous focusing magnifications, angling and lighting possibilities and its reproducibility is very poor. The Scheimpflug slit-lamp photography tilts the camera’s depth of field to consistently get transversal sharp focused images of the lens. Cataract scatters light and appears as varied elevations in accordance to location and severity. Scheimpflug has the disadvantage of requiring many pictures, in different meridians, to reliably estimate the size of the opacity.

Forward scattering examination: Retro-illumination techniques flood the retina with light, whose reflex reaches the crystallin from behind, propagating the scattering to the camera. Mean gray level, best fitting polynomials, feature extraction, and other image processing techniques are used to automatically measure size and shape of the cataract. Since the position of the spot is unknown, focusing skills are essential.

Research alternatives such as femtosecond lasers, and optical coherence tomography may provide new high-quality tools to estimate the size and position of a cataract. Using a Shack-Hartmann device, the coherent light ray hits the crystallin from behind and reaches the sensor. Blur captured by each lenslet is a local PSF of the lens. Although these techniques have been successfully used for cataract surgery, their high costs limit the adoption for diagnostic purposes.

 

Slit-lamp +

Visual Acuity Tests

Scheimpflug

Photography

Retro-illumination Techniques

Shack-Hartmann Wavefront Aberrometer

Optical Coherence Tomography

Ours

Scattering

Backward

Backward

Forward

Forward

---

Forward

Training

High

High

Medium

Low

High

Low

Cost

 

 

 

 

 

 

Data Log

No

Computer

Computer

Computer

Computer

Cellphone

Mobility

> 1Kg

> 10Kg

> 10Kg

> 10Kg

> 10Kg

< 300g

Speed

Slow

Slow

Fast

Fast

Medium

Fast

Scalability

Hard

Hard

Hard

Hard

Hard

Easy

AC Need

Yes

Yes

Yes

Yes

Yes

No

Networked?

No

No

No

No

No

Yes

Self-Diagnose

No

No

No

No

No

Yes

Method

Subjective

Subjective

Objective

Objective

Objective

Subjective

Early Cataracts

Hard to find

Hard to find

Hard to find

Yes

Yes

Yes

Accuracy

Personal Skills

Personal Skills

Personal Skills

High

High

High

Cost

$5,000.00

$20,000.00

$5,000.00

$15,000.00

$10,000.00

$2.00

Table 1: Comparison of our technique against current available technologies and research tools.

We notice that many people will argue that a single picture of the cataract-affected eye is enough for a doctor to diagnose the disease. This is true for advanced stages of cataracts, when people are technically blind and are taken to surgery right away. Our device, however, operates best in early stages of the disease, being potentially helpful to alert patients, monitor, and to allow lifestyle adjustments to reduce further development.

Potential Impact

Cataracts are denatured crystallin proteins that are clamped together in the nucleus, on the cortex or under the capsule of the crystallin (Figure 2). With the continuous production and accumulation of lens fibers throughout life, the crystallin becomes thicker and more compact. This disease is the leading cause of avoidable blindness worldwide [WHO 2005] and this occurrence is highly correlated to the aging process. 250M people worldwide have undiagnosed cataract that affects their vision, poverty and illiteracy. 17% of the +40-year-old Americans have cataracts, 50% of +75-year-old have had cataracts, and its incidence is expected to grow with the increasing longevity [NIH-EDPRSG 2004; Li et al. 2010]. It is estimated that one third of Americans will undergo cataract surgery in their life time [Palanker et al. 2010]. There is currently no efficient method to prevent it or to completely stop its growth. The rate of this expansion, however, can be controlled if early diagnostics are obtained [Fostera et al. 2003]. Methods to detect early cataracts and assess its progression over time could be potentially helpful for the development and testing of new treatments [Asbell et al. 2005], to alert patients, and to allow lifestyle adjustments to reduce further development [Datiles et al. 2008].

Experiences shared on the user studies: Many of the test subjects were fascinated by their opacity map on the screen of a smart-phone. One of the cataract-affected subjects has reported difficulty in explaining the visual effects to his family. A simple rendering tool may address these communication issues between them. Response from the local community has been great. Our data shows a reasonable repeatability, but some users found the alignment task difficult to understand. The owner of a respectful company that provides health care in developing countries has demonstrated excitement about the technology: “Village health workers will be able to cheaply and quickly flag early stage cataracts and macular degeneration in order to refer individuals to hospitals, where their vision can be restored before they effectively become blind”.

Reactions from ophthalmologists: Several research and local practicing ophthalmologists have been in collaboration with this project, and are enthusiastic about its unique outcomes. Many of them have experimented with the device, and the general response has reinforced that reliable quantitative measurements for cataracts are already very helpful for screening purposes. One of them commented on their experience that the Shack-Hartmann wavefront sensor to measure high-order optical distortions of the human eye had no practical application twenty years ago. Today, the high accuracy of these devices provide the only reliable data for the LASIK surgery. Widespread availability of devices like ours, which generate quantitative data about cataracts, may benefit the future of diagnostic and surgical practice. Since cataracts are highly correlated with macular degeneration, many doctors have suggested the use of this device as a side screening tool for other visual impairments.

A few ophthalmologists we have been discussing with, reported strong concerns about the complete absence of a glare disability test in order to obtain a driver’s license. For instance, visual acuity tests, in general, do not assess for glare and night driving effects, while simple and cheap tests such as ours, would reveal currently unchecked impairments. Our overall goal is to create tools that empower self-awareness about commonly unscreened health condition of the eye. We stress that this device does not directly diagnose or treat for cataracts, but in the future, methods like this might be able to give a complete summary of visual performance. Our hope is that these results encourage more people to design and develop interactive tools which will augment the understanding of the human visual experience.


Acknowledgments:

We would like to thank doctors Bruce Moore, Rob Pocaro, Yvvone Tsai, Shrikant Bharadwaj, Fabiano Cade, and Caio Regatieri for sharing their thoughts on the clinical use for this device. Fredo Durand, Bill Freeman, Sam Hasinoff, Ankit Mohan, Leandro Fernandes, Mahdi Mohammad-Bagher, and Kartic Subr for their comments on an earlier draft of the paper. All the SIGGRAPH reviewers for their thorough and insightful feedback. Tiago Oliveira for the illustration on Figure 1, Tyler Hutchison for the voice-over and the entire Camera Culture group for their unrelenting support; Vitor and Manuel acknowledge CNPq-Brazil fellowships 142563/2008-0, 200763/2009-1, 200284/2009-6, 308936/2010-8, 480485/2010-0. Erick acknowledges CAPES-Brazil scholarship BEX 2529-10-6. Ramesh is supported by an Alfred P. Sloan Research Fellowship, and Esteban by a Young Scientist of Rio de Janeiro Fellowship.

Media Coverage


Pictures


© Mike Ritter / ritterbin.com
PerfectSight: Camera Evaluation
© Mike Ritter / ritterbin.com

© Mike Ritter / ritterbin.com
PerfectSight: Camera Evaluation
© Mike Ritter / ritterbin.com

© Mike Ritter / ritterbin.com
PerfectSight: Camera Evaluation
© Mike Ritter / ritterbin.com
MIT Ideas & Global Challenge Awards

DLP Projector - Final Prototype
PerfectSight: Camera Evaluation
Stack of LCDs - Final Prototype
CATRA
Cell phone Final Prototype

Slit-lamp comparison

Subject on the Slit Lamp

Subject Testing

Fundus Camera - Retroillumination

Picture on the Slit-Lamp

Simulation of Advanced Cataracts

Scratched Contact Lens

Controlled Evaluation

Simulation of Cataracts

LCD Projector Prototype

LCD Projector Prototype

LCD Projector Prototype

Photoframe Prototype

Vuzix HMD Prototype

Vuzix HMD Prototype

First Cell phone Prototype

DLP Projector Prototype

Stack of LCDs Prototype

Stack of LCDs Prototype

Testing DLP Prototype

Testing Stack of LCDs Prototype

Testing prototype for Reviewers

First Fake Cataracts

Prototype for Reviewers

Prototype for Reviewers

Darkening the Scene

Testing Diffusers

Recent projects in the Camera Culture group

Flutter Shutter Camera 6D Display Bokode BiDi Screen NETRA
Cameras 6D Display
Lighting and Viewpoint aware displays
Bokode
Long Distance Barcodes
BiDi Screen
Touch+3D Hover on Thin LCD
NETRA
Cellphone based Eye Test

MIT Media Lab - All rights reserved.