gms | German Medical Science

GMDS 2013: 58. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V. (GMDS)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie

01. - 05.09.2013, Lübeck

Quantitative Conjunctival Provocation Test

Meeting Abstract

  • István Sárándi - RWTH Aachen, Aachen, DE
  • Thomas Deserno - RWTH Aachen, Aachen, DE
  • Dan Classen - Center of Rhinology and Allergology, Wiesbaden, DE
  • Oliver Pfaar - Center of Rhinology and Allergology, Wiesbaden, DE
  • Anatoli Astvatsatourov - University Hospital Cologne, Köln, DE
  • Ralph Mösges - University Hospital Cologne, Köln, DE

GMDS 2013. 58. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e.V. (GMDS). Lübeck, 01.-05.09.2013. Düsseldorf: German Medical Science GMS Publishing House; 2013. DocAbstr.245

doi: 10.3205/13gmds074, urn:nbn:de:0183-13gmds0741

Published: August 27, 2013

© 2013 Sárándi et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.



Introduction: Allergic diseases are becoming increasingly widespread worldwide. For documentation and assessment of treatment’s efficiency, provocation tests such as the conjunctival provocation test (CPT) have been established [1]. In CPT, pollen solutions of different concentrations are applied into the patient’s eye. In case of a responding patient, the blood vessels of the conjunctiva and sclera expand increasing its redness, which is inspected visually by the physician. However, qualitative evaluation is observer-specific and hinders CPT to be used as a primary endpoint in controlled clinical trials. In this contribution, we apply image processing yielding automatic and quantitative CPT measurements.

Materials and Methods: Images of the eye before and after application of allergen solution are captured with a digital camera. To measure the redness of the eye in each image, we first extract the sclera. This is done using a combination of Hough circle transform to detect the iris, normalization and thresholding in YUV color space, morphological operations and connected component analysis. Registration of the extracted sclera images of the same patient is performed using the scale invariant feature transform (SIFT) [2]. These regions of interest (ROI) from the sequence of images then are intersected to quantify redness the same portion of each image. The redness of the sclera is determined by a thresholding and projection operation in HSV color space. Finally, the relative redness is calculated by dividing the redness after the application of the allergen by the redness before it. In contrast to our previous work [3], segmentation is performed fully automatically, and the regions of interest have been aligned before the color measure is computed. Implementation is based on Java and ImageJ and parts of the JavaSIFT plugin are used. The algorithm was applied to 46 experiments of 22 patients of an observational diagnostic study. According to the CPT protocol, 2 to 4 images per experiment were acquired, with 100 up to 100,000 SQ/ml of allergen solution applied. In total 166 standardized photographs have been processed, all of them acquired using a head rest controlling the distance, an Olympus MFT Camera with macro lenses, and a controlled illumination setting.

Results: The automatic segmentation performs robustly on all images. We counted in total 17 false positive areas, which are correctly removed in 11 of them when the sclera ROIs of multiple images are intersected, since the SIFT algorithm correctly registered all image sequences. For 100, 1000, 10000, 20000, 50000, 100000 SQ/ml concentration of allergen solution, a mean relative redness of 0.9447, 1.2333, 2.9567, 2.6471, 3.5742 and 3.6487, was obtained, respectively, based on the numbers of 14, 13, 26, 14, 14 and 6 images, respectively. Hence, the average relative redness plotted against the allergen concentration increases logarithmically according to the regression formula 0.428 ln(x) - 1.279, with a coefficient of determination R^2=0.93.

Conclusion: The presented method is suitable for computer-aided diagnostics in allergic rhinitis/rhinoconjunctivitis. CPT becomes suitable for as surrogate endpoint in controlled clinical trials with reduced number of subjects.


Riechelmann H, Epple B, Gropper G. Comparison of conjunctival and nasal provocation test in allergic rhinitis to house dust mite. Int Arc Allergy Immunol. 2003;130(1):51-59.
Lowe DG. Distinctive image features from scale-invariant keypoints. Int J Comput Vis. 2004; 60(2):91-110.
Bista SR, Sarandi I, Dogan S, Astvatsatourov A, Mösges R, Deserno TM. Automatic conjunctival provocation test combining Hough circle transform and self-calibrated color measurements. Proc SPIE. 2013; 8670.