Automated Diabetic Retinopathy

Project Goal: Develop a tool that can analyze retinal mages and
automatically compute the level of diabetic retinopathy in a patient.
To detect diabetic retinopathy in an image, one has to use image
processing techniques to determine different characteristics in a
retinal image, namely, micro-aneurysms, haemoorhages, exudates,
cotton-wool spots and blood vessel growth. Based on these
characteristics, one has to characterize the patient into different
levels of diabetic retinopathy - normal, mild NPDR, moderate NPDR,
sever NPDR, PDR, CSME.

The inference mechanism used for this classification and analysis will
use a combination of image processing techniques and advanced machine
learning techniques. Given several images where the diagnosis result
is known, the plan is to use this dataset as a training dataset and
apply machine learning techniques to analyze and determine the level
of diabetic retinopathy in a new image.

The overall plan is to obtain many retinal images from Aravind Eye
Hospitals which have been already analyzed by experts there to form
the training as well as the testing data set. If the NYU opthalmology
dept can share retinal images, they can also be used in the analysis.

Contacts:
Lakshmi Subramanian, CS
Yann Lecun, CS
Jasmine Francis, NYU School of Medicine


There are various stages in Diabetic retinopathy:

1. Mild Nonproliferative Retinopathy. At this earliest stage,
microaneurysms occur. They are small areas of balloon-like swelling in
the retina's tiny blood vessels.

What to look for in the image: Microaneurysms are small red dots that appear 
as like micro blood clots in the image.

2. Moderate Nonproliferative Retinopathy. As the disease progresses, some
blood vessels that nourish the retina are blocked.

What to look for: You find haemmorhages (dark red spots), cotton wool
spots (hazy yellow spots) and darkish yellow spots.

3. Severe Nonproliferative Retinopathy. Many more blood vessels are
blocked, depriving several areas of the retina with their blood
supply.  These areas of the retina send signals to the body to grow
new blood vessels for nourishment.

What to look for: We have lots of haemmorhages and dark yellow spots
in the eye. You also see a large density of blood vessel growth around
a region because the main blood vessel is blocked. This appears in the
image as a region with a whole range of thin red curves around a
region.

4. Proliferative Retinopathy. At this advanced stage, the signals sent
by the retina for nourishment trigger the growth of new blood
vessels. This condition is called proliferative retinopathy. These new
blood vessels are abnormal and fragile. They grow along the retina and
along the surface of the clear, vitreous gel that fills the inside of
the eye. By themselves, these blood vessels do not cause symptoms or
vision loss. However, they have thin, fragile walls. If they leak
blood, severe vision loss and even blindness can result.

What to look for: Apart from haemorrhages and fat accumulation, we see
thinning of blood vessels and abnormal growth of blood vessels.

5. Clinically Severe Macular Edema (CSME):

Macular Edema is a different but very related problem.

What to look for: Here we would see haemorrhages and darkish yellow
spots near the macular region of the eye.


There are other conditions which can create false positives in the analysis:

(a) Drusen: Uniform yellowness throughout the eye due to old
age. Looks slightly similar to the yellow fat deposits. But this is a
normal but old eye. This is a false positive scenario.

(b) Cataract images: If someone has had cataract, then their images
are totally blurred and cannot be diagnosed.

Some sample images are available at:

http://www.cs.nyu.edu/~lakshmi/dr_images.tar.gz

These images have been classified into different categories too.