TOSCA-Imaging--developing Internet based image processing software for screening and diagnosis of diabetic...

Area of Science:

• Ophthalmology
• Medical Imaging
• Telemedicine

Background:

• Diabetic retinopathy poses a significant threat to vision.
• Early detection and regular screening are crucial for managing diabetic retinopathy.
• Existing diagnostic methods can be resource-intensive and geographically limited.

Purpose of the Study:

• To develop and implement internet-based software and image databases for diabetic retinopathy screening and diagnosis.
• To create a communication platform for transmitting and analyzing retinal images.
• To integrate European-based image processing research into a single accessible system.

Main Methods:

• Construction of an internet-based communication platform for retinal image analysis.
• Implementation of automated routines for detecting microaneurysms, venous beading, and hard exudates.
• Development of image alignment for serial analysis and a reference image database.
• Validation of detection algorithms and the reference database through expert grading.

Main Results:

• Detection of individual lesions (e.g., normality) achieved ~80% sensitivity and specificity.
• Detection of lesion patterns (e.g., macular edema) exceeded 95% sensitivity and specificity.
• Reference database validation showed <90% sensitivity/specificity for any lesion and >95% for overall retinopathy grade.

Conclusions:

• TOSCA-Imaging successfully developed and implemented internet-based software for diabetic retinopathy screening.
• The system integrates international image processing efforts for accessible diagnosis.
• The developed tools demonstrate high accuracy in identifying diabetic retinopathy features and grading severity.