Abstract
Diabetic retinopathy (DR) is one of the most common causes of vision loss in people who have diabetes for a prolonged period. Convolutional neural networks (CNNs) have become increasingly popular for computer-aided DR diagnosis using retinal fundus images. While these CNNs are highly reliable, their lack of sufficient explainability prevents them from being widely used in medical practice. In this article, we propose a novel explainable deep learning ensemble model where weights from different models are fused into a single model to extract salient features from various retinal lesions found on fundus images. The extracted features are then fed to a custom classifier for the final diagnosis of DR severity level. The model is trained on an APTOS dataset containing retinal fundus images of various DR grades using a cyclical learning rates strategy with an automatic learning rate finder for decaying the learning rate to improve model accuracy. We develop an explainability approach by leveraging gradient-weighted class activation mapping and shapely adaptive explanations to highlight the areas of fundus images that are most indicative of different DR stages. This allows ophthalmologists to view our model's decision in a way that they can understand. Evaluation results using three different datasets (APTOS, MESSIDOR, IDRiD) show the effectiveness of our model, achieving superior classification rates with a high degree of precision (0.970), sensitivity (0.980), and AUC (0.978). We believe that the proposed model, which jointly offers state-of-the-art diagnosis performance and explainability, will address the black-box nature of deep CNN models in robust detection of DR grading.
- [1] 2020. Prevalence, incidence and future projection of diabetic eye disease in Europe: A systematic review and meta-analysis. European Journal of Epidemiology 35, 1 (2020), 11–23.Google Scholar
Cross Ref
- [2] . 2010. Diabetic retinopathy. Lancet 376, 9735 (2010), 124–136.Google Scholar
Cross Ref
- [3] . 2017. Automated detection of diabetic retinopathy using deep learning. In Proceedings of AMIA Joint Summits on Translational Science 2017, 147–155.Google Scholar
- [4] . 2020. Diabetic retinopathy detection through deep learning techniques: A review. Informatics in Medicine Unlocked 20 (2020), 100377.Google Scholar
Cross Ref
- [5] . 2017. Grader variability and the importance of reference standards for evaluating machine learning models for diabetic retinopathy, 2017, CoRR, abs/1710.01711.Google Scholar
- [6] . 2021. A multimodal, multimedia point-of-care deep learning framework for COVID-19 diagnosis. ACM Trans. Multimedia Comput. Commun. 17, 1s, Article 18 (2021), 24 pages.
DOI: DOI: https://doi.org/10.1145/3421725Google Scholar - [7] . 2016. Convolutional neural networks for diabetic retinopathy. Procedia Computer Science 90, 2016, 200–205.Google Scholar
Cross Ref
- [8] , G. Muhammad, and M. Kumar. 2021. eDiaPredict: An ensemble-based framework for diabetes prediction. ACM Trans. Multimedia Comput. Commun. Appl. 17, 2s, Article 66 (2021), 26 pages. Google Scholar
Digital Library
- [9] , and Y. Zhang. 2018. Deep convolutional neural networks for diabetic retinopathy detection by image classification. Computers & Electrical Engineering 72, (2018), 274–282. Google Scholar
Digital Library
- [10] . 1992. Acceleration of stochastic approximation by averaging. SIAM Journal on Control and Optimization 30, 4 (1992), 838–855. Google Scholar
Digital Library
- [11] . 2017. Cyclical learning rates for training neural networks. In Proceedings of the IEEE Winter Conference on Applications of Computer Vision 464–472.Google Scholar
Cross Ref
- [12] APTOS. 2019. APTOS 2019 blindness detection. https://www.kaggle.com/c/aptos2019-blindness-detectionLast. accessed October 25, 2020.Google Scholar
- [13] . 2014. Feedback on a publicly distributed database: The Messidor database. Image Analysis & Stereology 33, 3 (2014), 231–234.Google Scholar
Cross Ref
- [14] , Manesh Kokare, Girish Deshmukh, Jaemin Son, Woong Bae, Lihong Liu, Jianzong Wang, Xinhui Liu, Liangxin Gao, TianBo Wu, Jing Xiao, Fengyan Wang, Baocai Yin, Yunzhi Wang, Gopichandh Danala, Linsheng He, Yoon Ho Choi, Yeong Chan Lee, Sang-Hyuk Jung, Zhongyu Li, Xiaodan Sui, Junyan Wu, Xiaolong Li, Ting Zhou, Janos Toth, Agnes Baran, Avinash Kori, Sai Saketh Chennamsetty, Mohammed Safwan, Varghese Alex, Xingzheng Lyu, Li Cheng, Qinhao Chu, Pengcheng Li, Xin Ji, Sanyuan Zhang, Yaxin Shen, Ling Dai, Oindrila Saha, Rachana Sathish, Tânia Melo, Teresa Araújo, Balazs Harangi, Bin Sheng, Ruogu Fang, Debdoot Sheet, Andras Hajdu, Yuanjie Zheng, Ana Maria Mendonça, Shaoting Zhang, Aurélio Campilho, Bin Zheng, Dinggang Shen, Luca Giancardo, Gwenolé Quellec, and Fabrice Mériaudeau. 2020. IDRiD: diabetic retinopathy – Segmentation and grading challenge. Med. Image Anal. 59 (2020), 101561.Google Scholar
Cross Ref
- [15] . 2019. Applying deep learning for epilepsy seizure detection and brain mapping visualization. ACM Trans. Multimedia Comput. Commun. 15, 1s 10 (2019), 17 pages. Google Scholar
Digital Library
- [16] , and N. Kumar. 2021. EEG-Based pathology detection for home health monitoring. IEEE Journal on Selected Areas in Communications 39, 2 (2021), 603–610.
DOI: 10.1109/JSAC.2020.3020654Google Scholar - [17] . 2012. SVM and neural-network-based diagnosis of diabetic retinopathy. International Journal of Computer Applications 41, 1 (2012), 6–12.Google Scholar
Cross Ref
- [18] . 2014. Automatic screening and classification of diabetic retinopathy fundus images. In Proceedings of the International Conference on Communications in Computer and Information Science. 113–122.Google Scholar
Cross Ref
- [19] , J. de la Calleja, A. Benitez, and M. A. Medina. 2012. Image-based classification of diabetic retinopathy using machine learning. In Proceedings of the 12th International Conference on Intelligent Systems Design and Applications (ISDA). 826–830.Google Scholar
- [20] . 2020. Detection of early signs of diabetic retinopathy based on textural and morphological information in fundus images. Sensors 2020, 20, 1005.
DOI: DOI: 10.3390/s20041005Google ScholarCross Ref
- [21] . 2020. Explainable AI and mass surveillance system-based healthcare framework to combat COVID-I9 like pandemics. IEEE Network 34, 4 (2020), 126–132.Google Scholar
Cross Ref
- [22] , M. Al-Hammadi, and G. Muhammad. 2019. Automatic fruit classification using deep learning for industrial applications. IEEE Transactions on Industrial Informatics 15, 2 (2019), 1027–1034.Google Scholar
- [23] , S. Arulmalar, M. Usha, V. Prathiba, K. S. Kareemuddin, R. M. Anjana, and V. Mohan. 2015. Validation of smartphone based retinal photography for diabetic retinopathy screening. PLoS One 10, 9 (2015).Google Scholar
Cross Ref
- [24] . 2016. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA 316, 22 (2016), 2402–2410.Google Scholar
Cross Ref
- [25] 2016. Rethinking the inception architecture for computer vision. In Proc.of IEEE Conference on Computer Vision and Pattern Recognition 2016, 2818–2826.Google Scholar
Cross Ref
- [26] Diabetic Retinopathy Detection. https://www.kaggle.com/c/diabetic-retinopathy-detection/data, Last accessed October 30, 2020.Google Scholar
- [27] Messidor-2 DR Grades. https://www.kaggle.com/google-brain/messidor2-dr-grades, Last accessed October 30, 2020.Google Scholar
- [28] . 2017. Automated identification of diabetic retinopathy using deep learning. Ophthalmology 124, 7 (2017), 962–969.Google Scholar
Cross Ref
- [29] , N.-N. Yeh, S.-J. Chen, and Y.-C. Chung. 2019. Computer-assisted diagnosis for diabetic retinopathy based on fundus images using deep convolutional neural network. Mobile Information Systems 2019,
Article ID 6142839 . https://doi.org/10.1155/2019/6142839Google Scholar - [30] . 2017. Cloud-supported cyber–physical localization framework for patients monitoring. IEEE Systems Journal 11, 1 (2017), 118–127.Google Scholar
Cross Ref
- [31] . 2020. Deep learning approach to diabetic retinopathy detection. arXiv:2003.02261 [cs.LG], 2020.Google Scholar
- [32] . 2019. Transfer learning based detection of diabetic retinopathy from small dataset. CoRR, abs/1905.07203.Google Scholar
- [33] 2019. Convolutional neural networks for mild diabetic retinopathy detection: An experimental study. bioRxiv, 2019.Google Scholar
- [34] , C.-L. Chen, C.-M. Liang, M.-C. Tai, J.-T. Liu, P.-Y. Wu, M.-S. Deng, Y.-W. Lee, T.-Y. Huang, and Y.-H. Chen. 2020. Leveraging multimodal deep learning architecture with retina lesion information to detect diabetic retinopathy. Translational Vision Science & Technology (TVST) 9, 2 (2020), 41. https://doi.org/10.1167/tvst.9.2.41Google Scholar
Cross Ref
- [35] . 2020. Deep learning frameworks for diabetic retinopathy detection with smartphone-based retinal imaging systems. Pattern Recognition Letters 135, (2020), 409–417.Google Scholar
Cross Ref
- [36] . 2019. Severity analysis of diabetic retinopathy in retinal images using hybrid structure descriptor and modified CNNs. Measurement 135 (2019), 762–767.Google Scholar
Cross Ref
- [37] . 2020. Exudate detection for diabetic retinopathy using pretrained convolutional neural networks. Complexity 2020.Google Scholar
Digital Library
- [38] . 2020. Explainable diabetic retinopathy using efficientNET. In Proceedings of the 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC). 2020.
DOI: DOI: 10.1109/EMBC44109.2020.9175664Google ScholarCross Ref
- [39] . 2019. An explainable AI-based computer aided detection system for diabetic retinopathy using retinal fundus images. CAIP (1) 2019, 457–468.Google Scholar
- [40] . 2020. A survey on medical image analysis in diabetic retinopathy. Med Image Anal 64 101742.Google Scholar
Cross Ref
- [41] , W. Dong, R. Socher, L. Li, K. Li, and L. Fei-Fei. 2009. ImageNet: A large-scale hierarchical image database. International Conference on Computer Vision and Pattern Recognition (CVPR), 248–255.Google Scholar
Cross Ref
- [42] , Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. 2015. Going deeper with convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 1–9.Google Scholar
- [43] . 2016. Learning deep features for discriminative localization.In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2016, 2921–2929.
DOI: DOI: 10.1109/CVPR.2016.319Google ScholarCross Ref
- [44] , M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra. 2019. Grad-CAM: Visual explanations from deep networks via gradient-based localization. International Journal of Computer Vision 128, 2 (2019), 336–359.Google Scholar
Digital Library
- [45] . 2017. A unified approach to interpreting model predictions. In Proceedings of Neural Information Processing Systems (NIPS). 4768–4777. Google Scholar
Digital Library
- [46] . Cyclical Learning Rates Implementation. https://github.com/bckenstler/CLR.Google Scholar
Index Terms
An Explainable Deep Learning Ensemble Model for Robust Diagnosis of Diabetic Retinopathy Grading
Recommendations
Computer-aided diagnosis of diabetic retinopathy: A review
Diabetes mellitus may cause alterations in the retinal microvasculature leading to diabetic retinopathy. Unchecked, advanced diabetic retinopathy may lead to blindness. It can be tedious and time consuming to decipher subtle morphological changes in ...
DenseHyper: an automatic recognition system for detection of hypertensive retinopathy using dense features transform and deep-residual learning
AbstractHigh blood pressure and diabetes are associated with a retinal abnormality known as Hypertensive Retinopathy (HR). The severity-level and duration of hypertension are straightly related to the incidence of HR-eye disease. The HR damages the ...
An Automated Model using Deep Convolutional Neural Network for Retinal Image Classification to Detect Diabetic Retinopathy
ICCA 2020: Proceedings of the International Conference on Computing AdvancementsDiabetic Retinopathy is considered as one of the significant reasons for vision impairment. Its identification involves detecting the presence of some features in retinal fundus images by clinicians which is a time and resource consuming procedure and a ...






Comments