Sie sind auf Seite 1von 28

An Automatic

Diabetic Retinopathy
Telescreening system of Thailand

Jirach Suthammanas
Sirindhorn International Institute of Technology(SIIT)
Thammasat University

1
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

2
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

3
What is Diabetic Retinopathy ?
A damage to the retina that caused by
complications of diabetes.

How serious ?

4
Can this be successfully treated ?
Yes !
How to successfully treated ?
Early Detection !
How to detect ?
Using fundus image

5
• Approximately 6 million diabetic patients in Thailand

• They must have their eyes screen on a yearly basis

• Thailand have only 800 qualified ophthalmologists

• 400 of this ophthalmologists reside in Bangkok

• An automated system is need.

6
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

7
web-based application based on SOA architecture

8
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

9
Input :: Digital fundus image
Output :: Result Image + XML result

10
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

11
These are 2 type of user interface

• Command Line – for batch processing at server

• Graphical User Interface – for ophthalmologists

12
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

13
After processed each image the application produce
2 output

1. Result Image

2. XML Result

14
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

15
We can test the system using ground-truth image

16
Result

The System

Input Image

Ground Truth image

Ophthalmologists

17
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

18
There will be 2 engines in the system

19
Consists of 8 processing steps

20
• Improve ability of exudates detection in low contrast
non-dilated pupil images

• This engine used clustering techniques which can


increase accuracy of the screening process

• The accuracy of the engine is 92.52%.

21
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

22
• C++

• OpenCV

• Qt

• Eclipse

• GNU C++ Compiler

23
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

24
Benefits :
1) Thailand citizen
• More diabetic patient in rural area can have a chance to have their
eye screen.

2) Ophthalmologists
• Reduce the number of the patient that the ophthalmologists really
need to treat.

Output :
• A reliable telescreening system that produce zero false
negative result

25
• Introduction
• Overview
• Web Interface
• The Application
• Output
• Result Validation
• Engines
• Tools
• Benefits and Expected Outputs
• Summary

26
• An Automatic Diabetic Retinopathy Telescreening
System

• We will use SOA structure to provide this service via


the Internet

• Our system will consist of 2 screening Algorithms

27
28

Das könnte Ihnen auch gefallen