AUTOMATED DETECTION OF VITRITIS USING ULTRAWIDE-FIELD FUNDUS PHOTOGRAPHS AND DEEP LEARNING.

BACKGROUND/PURPOSE: Evaluate the performance of a deep learning algorithm for the automated detection and grading of vitritis on ultrawide-field imaging.

METHODS

Cross-sectional noninterventional study. Ultrawide-field fundus retinophotographs of uveitis patients were used. Vitreous haze was defined according to the six steps of the Standardization of Uveitis Nomenclature classification. The deep learning framework TensorFlow and the DenseNet121 convolutional neural network were used to perform the classification task. The best fitted model was tested in a validation study.

RESULTS

One thousand one hundred eighty-one images were included. The performance of the model for the detection of vitritis was good with a sensitivity of 91%, a specificity of 89%, an accuracy of 0.90, and an area under the receiver operating characteristics curve of 0.97. When used on an external set of images, the accuracy for the detection of vitritis was 0.78. The accuracy to classify vitritis in one of the six Standardization of Uveitis Nomenclature grades was limited (0.61) but improved to 0.75 when the grades were grouped into three categories. When accepting an error of one grade, the accuracy for the six-class classification increased to 0.90, suggesting the need for a larger sample to improve the model performances.

CONCLUSION

A new deep learning model based on ultrawide-field fundus imaging that produces an efficient tool for the detection of vitritis was described. The performance of the model for the grading into three categories of increasing vitritis severity was acceptable. The performance for the six-class grading of vitritis was limited but can probably be improved with a larger set of images.