Automated Detection of Diabetic Retinopathy with Sartphone Based Photography

Project Title

Automated Detection of Diabetic Retinopathy with Sartphone Based Photography

Project Area of Specialization Artificial IntelligenceProject Summary

Diabetic Retinopathy (DR) is an eye condition that can cause permanent vision impairment or blindness in people who have diabetes. A diabetes patient needs to get a comprehensive eye exam at least once a year.

According to the National Diabetes Survey of Pakistan 2016-2017, 27.4 million people over the age of 20 are affected by diabetes.

Population-based studies suggest that about a third of all diabetics will be affected by diabetic retinopathy.

The eye screening for Diabetic Retinopathy is not carried out in rural areas of Pakistan due to expensive eye screening equipment. Hence, our project serves as a cheap solution to detect Diabetic Retinopathy via a flagship smartphone and a 20D Volk (20 Diopter) lens.

Project Objectives

1. Detection of preventable blindness

2. Analyzing acquired data to identify trends and improve diagnosis

3. Making cost-effective DR screening possible in low-income areas

Project Implementation Method

In order to acquire fundus (back of the eye) images of the retina following steps are taken:

1. Set the smartphone camera on video mode.

2. Turn on the camera's flashlight.

3. After darkening the room, press the camera record button on the smartphone to record a continuous video.

4. Hold the +20 diopter (D) or any other indirect ophthalmoscopy lens in front of the patient’s eye with your thumb and index finger. Use the middle and ring fingers to stabilize the hand and lens and to help keep eyelids open.

5. Hold the camera 10 - 35 cm from the lens. Direct the camera along the patient's pupillary axis. Aim the light to the pupil and find the retina glow. Direct the light through the lens onto the retina and continue video recording.

6. While recording, move the camera and the lens to find a good focus and an image free of light reflections. Adjust the handheld lens to and from the patient’s eye to see a clear retina image filling the entire lens area. NOTE: In our experience, the lens should be held 3 - 5 cm from the eye (this is slightly further away from the cornea surface compared to what is appropriate for an indirect ophthalmoscopy exam).

7. Try to eliminate unnecessary light reflections by adjusting the camera and lens positions. Losing the image while searching the retina means that the camera and its flashlight are not aligned with the handheld lens and pupil. It takes practice to keep the camera, handheld lens, and patient pupil aligned.

8. Continue to record the video until a good view of the area of interest without significant light reflections and aberrations is captured.

9. Stop the recording and ask the patient to sit back and be comfortable.

10. Re-play the recorded movie until there is a good retina view within the handheld lens area. Stop the movie and take a screenshot of the view.

Benefits of the Project

Smartphone fundus photography is a simple technique to obtain fundus pictures using a smartphone camera and a conventional handheld indirect ophthalmoscopy lens with the improvement in machine learning algorithms and Automated AI analysis, smartphone retinal imaging has a very high sensitivity for detecting DR and thus it can be an initial tool for mass retinal screening in people with diabetes. 

Technical Details of Final Deliverable

A smartphone application using android studio will be developed that will be integrated with the Deep Learning or Machine Learning Model to predict the disease.

Images of the eye will be taken as input and after processing these images, the output will be produced predicting the presence or absence of DR. The patient will be recommended to consult a retina specialist in case DR is detected.

A 3D printed scope will be attached to the smartphone camera that will hold the  20D lens and the smartphone in alignment. The scope will be about 15 cm long according to the focal length and optional distance recorded while practicing. This scope will help keep the smartphone and the lens stable during the image acquisition process.

Final Deliverable of the Project Software SystemCore Industry ITOther Industries Health Core Technology Artificial Intelligence(AI)Other Technologies 3D/4D PrintingSustainable Development Goals Good Health and Well-Being for People, Affordable and Clean Energy, Industry, Innovation and InfrastructureRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	80000
20D Lens	Equipment	1	70000	70000
3D printed scope	Miscellaneous	1	10000	10000