TY - GEN
T1 - Retinal image quality classification using fine-tuned CNN
AU - Sun, Jing
AU - Wan, Cheng
AU - Cheng, Jun
AU - Yu, Fengli
AU - Liu, Jiang
N1 - Publisher Copyright:
© Springer International Publishing AG 2017.
PY - 2017
Y1 - 2017
N2 - Retinal image quality classification makes a great difference in automated diabetic retinopathy screening systems. With the increase of application of portable fundus cameras, we can get a large number of retinal images, but there are quite a number of images in poor quality because of uneven illumination, occlusion and patients movements. Using the dataset with poor quality training networks for DR screening system will lead to the decrease of accuracy. In this paper, we first explore four CNN architectures (AlexNet, GoogLeNet, VGG-16, and ResNet-50) from ImageNet image classification task to our Retinal fundus images quality classification, then we pick top two networks out and jointly fine-tune the two networks. The total loss of the network we proposed is equal to the sum of the losses of all channels. We demonstrate the super performance of our proposed algorithm on a large retinal fundus image dataset and achieve an optimal accuracy of 97.12%, outperforming the current methods in this area.
AB - Retinal image quality classification makes a great difference in automated diabetic retinopathy screening systems. With the increase of application of portable fundus cameras, we can get a large number of retinal images, but there are quite a number of images in poor quality because of uneven illumination, occlusion and patients movements. Using the dataset with poor quality training networks for DR screening system will lead to the decrease of accuracy. In this paper, we first explore four CNN architectures (AlexNet, GoogLeNet, VGG-16, and ResNet-50) from ImageNet image classification task to our Retinal fundus images quality classification, then we pick top two networks out and jointly fine-tune the two networks. The total loss of the network we proposed is equal to the sum of the losses of all channels. We demonstrate the super performance of our proposed algorithm on a large retinal fundus image dataset and achieve an optimal accuracy of 97.12%, outperforming the current methods in this area.
KW - Convolutional neural networks (CNN)
KW - Fine-tuning
KW - No-reference image quality assessment (NR-IQA)
KW - Retinal image
UR - http://www.scopus.com/inward/record.url?scp=85029821249&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-67561-9_14
DO - 10.1007/978-3-319-67561-9_14
M3 - Conference contribution
AN - SCOPUS:85029821249
SN - 9783319675602
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 126
EP - 133
BT - Fetal, Infant and Ophthalmic Medical Image Analysis - International Workshop, FIFI 2017 and 4th International Workshop, OMIA 2017 Held in Conjunction with MICCAI 2017, Proceedings
A2 - Cardoso, M. Jorge
A2 - Arbel, Tal
PB - Springer Verlag
T2 - International Workshop on Fetal and Infant Image Analysis, FIFI 2017 and 4th International Workshop on Ophthalmic Medical Image Analysis, OMIA 2017 held in Conjunction with 20th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2017
Y2 - 14 September 2017 through 14 September 2017
ER -
