Multi-Label Classification of Fundus Images With EfficientNet

Abstract

Convolutional neural network (CNN) has achieved remarkable success in the field of fundus images due to its powerful feature learning ability. Computer-aided diagnosis can obtain information with reference value for doctors in clinical diagnosis or screening through proper processing and analysis of fundus images. However, most of the previous studies have focused on the detection of a certain fundus disease, and the simultaneous diagnosis of multiple fundus diseases still faces great challenges. We propose a multi-label classification ensemble model of fundus images based on CNN to directly detect one or more fundus diseases in the retinal fundus images. Every single model consists of two parts. The first part is a feature extraction network based on EfficientNet, and the second part is a custom classification neural network for multi-label classification problems. Finally, the output probabilities of different models are fused as the final recognition result. And it was trained and tested on the data set provided by ODIR 2019 (Peking University International Competition on Ocular Disease Intelligent Recognition). The experimental results show that our model can be trained on fewer data sets and get good results.

Classification of Multiple Retinal Disorders from Enhanced Fundus Images Using Semi-supervised GAN

“…This measure is observed to be the highest for the method proposed in [14]. However, the proposed GAN model also gives a good level of agreement compared to the other methods [21,26]. The F1-score is considered as a weighted average of precision and recall terms.…”

“…Here, the authors proposed a novel CNN model and presented the visual results of classification on the test data set. To further highlight the quantitative results, Densenet, Inception, Resnet, MobileNet [23], Efficientnet [24], VGG, and Xception [25] models were engaged in training the classifier [26]. Here, each fundus image was distinctly passed to a grayscale histogram equalization module and a color histogram equalization module before feeding the classifier.…”

“…However, in this work, the finetuned model is trained on the augmented and enhanced larger sets of fundus images. The proposed semi-supervised GAN model is compared with the transfer learning methods as illustrated in [14,21,26] and it is compared with InceptionResnet model. Considering the classification accuracy of about 87%, the proposed GAN outperforms the other existing models.…”

“…Considering the classification accuracy of about 87%, the proposed GAN outperforms the other existing models. The least classification accuracy is obtained using the method in [26]. The Kappa score is a metric used to measure the inter-rater reliability.…”

Classificação Morfológica de Galáxias no S-PLUS por Combinação de Redes Convolucionais

“…Existem vários relatos de sucesso do uso desta técnica em deep learning, dentre eles a classificac ¸ão d e r etina humana [41], melanoma [42] e de anomalias na Via Láctea [43]. Existem várias técnicas de agrupamento (Ensemble), como Boosting [44,45], Bagging [46] e Stacking [47][48][49].…”