[ad_1]
Automated detection of exudates from fundus photos performs an vital function in diabetic retinopathy (DR) screening and analysis, for which supervised or semi-supervised studying strategies are sometimes most well-liked. Nonetheless, a possible limitation of supervised and semi-supervised studying based mostly detection algorithms is that they rely considerably on the pattern dimension of coaching knowledge and the standard of annotations, which is the basic motivation of this work. On this research, we assemble a dataset containing 1219 fundus photos (from DR sufferers and wholesome controls) with annotations of exudate lesions. Along with exudate annotations, we additionally present 4 further labels for every picture: left-versus-right eye label, DR grade (severity scale) from three totally different grading protocols, the bounding field of the optic disc (OD), and fovea location. This dataset supplies a terrific alternative to research the accuracy and reliability of various exudate detection, OD detection, fovea localization, and DR classification algorithms. Furthermore, it should facilitate the event of such algorithms within the realm of supervised and semi-supervised studying.
References
PubMed
[ad_2]
Source link