All people with diabetes have the risk of developing diabetic retinopathy(DR), a vision-threatening complication. Early detection and timely treatmentcan reduce the occurrence of blindness due to DR. Computer-aided diagnosis hasthe potential benefit of improving the accuracy and speed in DR detection. Thisstudy is concerned with automatic classification of images with microaneurysm(MA) and neovascularization (NV), two important DR clinical findings. Togetherwith normal images, this presents a 3-class classification problem. We proposea modified color auto-correlogram feature (AutoCC) with low dimensionality thatis spectrally tuned towards DR images. Recognizing the fact that the imageswith or without MA or NV are generally different only in small, localizedregions, we propose to employ a multi-class, multiple-instance learningframework for performing the classification task using the proposed feature.Extensive experiments including comparison with a few state-of-art imageclassification approaches have been performed and the results suggest that theproposed approach is promising as it outperforms other methods by a largemargin.
展开▼
