Purpose
To design, validate and implement image grading software application for the purposes of uveitis clinical trials. Currently the activity of posterior uveitis is judged by the amount of the vitreous haze. Printed photographic scales are available for clinical assessment only. A new digital algorithmic scale for vitreous haze grading was described recently and is suitable for computerized evaluation of fundus digital images from patients with vitreous opacification. We aimed to develop software application in order to facilitate the image display and randomization as well as automatic recording of the gradings.
Methods
Fast programming languages were used to create facile graphical user interface for image display. Randomization algorithm was designed and incorporated in the application. Access for compressed and non-compressed image files was developed and made available for selection from the user. Input of image data was programmed via excel spreadsheet or image file name. The testing was performed on Pentium computer with quad processor running Windows 7 operating system.
Results
Small set of 10 digital images was initially used to test the principle. 165 color digital Images from uveitis patients were used to validate the application. The images were acquired 10 times by two independent users with 100% accuracy. Randomization was checked on a smaller subset of 95 images also with 100% accuracy. The results were reproduced 100 % when the test was repeated 10 times. The access time for each image was less than 1 sec. The application was validated against data set from participants in national uveitis clinical trial. Correlation coefficient of 0.86 was achieved between 2 independent graders. Grading scores were recorded by the software and exported as spreadsheet for later statistical analysis.
Conclusions
Robust reliable digital image grading software was designed validated and implemented for the purposes of clinical trials. It may be used to grade large set of images against color digital standards photographs or to evaluate for better, same, worse outcome in a Reading Center settings.
Keywords: 550 imaging/image analysis: clinical •
746 uveitis-clinical/animal model •
468 clinical research methodology