Image Processing in the Tosca Project – Developing Internet Based Software for Screening and Diagnosis of Diabetic Retinopathy Ole K Hejlesen 1, Bernhard.

Image Processing in the Tosca Project – Developing Internet Based Software for

Screening and Diagnosis of Diabetic Retinopathy

Ole K Hejlesen1, Bernhard M Ege1, Karl-Hans Englmeier2, Steve Aldington3, Leo McCanna4, Toke Bek5

1Department of Health Science and Technology, Aalborg University, Fredrik Bajersvej 7 D1, DK-9220 Aalborg, Denmark, 2GSF-MEDIS, Munich,

Germany, 3Imperial College, London, UK, 4OCUCO Ltd. Dublin, Ireland 5Department of Ophthalmology, Aarhus University Hospital, Denmark

Aim

• to develop Internet based software and image data bases

• for screening and diagnosis of diabetic retinopathy and

• implementing it into a real life situation

The work consisted of

• Construction of an Internet based communication platform for transmitting and analyzing retinal images

• Implementation of routines for – detecting the first microaneurysm (transition from normal

to pathologic) – detecting patients needing referral for treatment (presence

of venous beading or hard exudates near the fovea)– serial analysis (image alignment)

• Construction of a reference image data base

Communication structure

Internet based communication platform – example: image alignment

• Web Client – standard browser

• Web Server – image analysis broker

• Image Processing Servers – ex: image alignment

Image alignmentUnaligned

images

Alignedimages

Landmark extraction

Pointmatching

Polynomialtransf.

Imagewarping

Cross validation results

Deviation from “perfect”, i.e. the deviation of rotation, translation and mean distance from 0, and the deviation of scale from 1 (Cideciyan 1995).

Affine PolynomialScale µ 0.00052 0.00009

s2 0.000000240 0.000000009

Rotation µ 0.045 0.013

s2 0.00100 0.00007

Translation x µ 0.43 0.13

s2 0.190 0.007

Translation y µ 0.400 0.085

s2 0.1400 0.0054

µ 0.92 0.58

s2 0.12 0.03

Transformation

Mean dist. [pixel]

100 + 50 + 50 + 50 +50 + 50 patients

Results

Validation of the reference image data base by double grading by two expert graders suggested a sensitivity and a specificity of just below 90% for any lesion and of more than 95% for predicting the overall retinopathy grade.

Conclusion

TOSCA-Imaging has succeeded in its primary aim of developing Internet based software and implementing it into a real life situation, integrating work within image processing done in four different European countries (England, Germany, Ireland, and Denmark) to be accessed from one Internet web site.

Acknowledgements

• The TOSCA Project was financed by EU's Fifth Framework IST Programme