Convolutional neural network transfer for automated glaucoma identification

Most current systems for automated glaucoma detection in fundus images rely on segmentation-based features, which are known to be influenced by the underlying segmentation methods. Convolutional Neural Networks (CNNs) are powerful tools for solving image classification tasks as they are able to learn highly discriminative features from raw pixel intensities. However, their applicability to medical image analysis is limited by the non-availability of large sets of annotated data required for training. In this article we present results of analysis of the viability of using CNNs that are pre-trained from non-medical data for automated glaucoma detection. Two different CNNs, namely OverFeat and VGG-S, were applied to fundus images to generate feature vectors. Preprocessing techniques such as vessel inpainting, contrast-limited adaptive histogram equalization (CLAHE) or cropping around the optic nerve head (ONH) area were explored within this framework to evaluate the improvement in feature discrimination, combined with both ℓ1 and ℓ2 regularized logistic regression models. Results on the Drishti-GS1 dataset, evaluated in terms of area under the average ROC curve, suggests the viability of this approach and offer significant evidence of the importance of well-chosen image pre-processing for transfer learning when the amount of data is not sufficient for fine-tuning the network.

[1]  László G. Nyúl,et al.  Classifying Glaucoma with Image-Based Features from Fundus Photographs , 2007, DAGM-Symposium.

[2]  Stephen Lin,et al.  Efficient Optic Cup Detection from Intra-image Learning with Retinal Structure Priors , 2012, MICCAI.

[3]  Kotagiri Ramamohanarao,et al.  An effective retinal blood vessel segmentation method using multi-scale line detection , 2013, Pattern Recognit..

[4]  Tien Yin Wong,et al.  Glaucoma Detection by Learning from Multiple Informatics Domains , 2015 .

[5]  A. Ng Feature selection, L1 vs. L2 regularization, and rotational invariance , 2004, Twenty-first international conference on Machine learning - ICML '04.

[6]  George Azzopardi,et al.  Trainable COSFIRE filters for vessel delineation with application to retinal images , 2015, Medical Image Anal..

[7]  Seth R Flaxman,et al.  Prevalence and causes of vision loss in Latin America and the Caribbean: 1990–2010 , 2014, British Journal of Ophthalmology.

[8]  Ronald M. Summers,et al.  Deep Convolutional Neural Networks for Computer-Aided Detection: CNN Architectures, Dataset Characteristics and Transfer Learning , 2016, IEEE Transactions on Medical Imaging.

[9]  László G. Nyúl,et al.  Glaucoma risk index:  Automated glaucoma detection from color fundus images , 2010, Medical Image Anal..

[10]  Jayanthi Sivaswamy,et al.  Glaucoma classification with a fusion of segmentation and image-based features , 2016, 2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI).

[11]  Matthew B. Blaschko,et al.  Learning Fully-Connected CRFs for Blood Vessel Segmentation in Retinal Images , 2014, MICCAI.

[12]  H. Fujita,et al.  Detection of retinal nerve fiber layer defects on retinal fundus images for early diagnosis of glaucoma. , 2010, Journal of biomedical optics.

[13]  Tien Yin Wong,et al.  Relationship of retinal vascular tortuosity with the neuroretinal rim: the singapore malay eye study. , 2010, Investigative ophthalmology & visual science.

[14]  Andrew Zisserman,et al.  Speeding up Convolutional Neural Networks with Low Rank Expansions , 2014, BMVC.

[15]  Hiroshi Ishikawa,et al.  Imaging of the optic nerve and retinal nerve fiber layer: an essential part of glaucoma diagnosis and monitoring. , 2014, Survey of ophthalmology.

[16]  Shuicheng Yan,et al.  Automatic Feature Learning for Glaucoma Detection Based on Deep Learning , 2015, MICCAI.

[17]  Bram van Ginneken,et al.  Comparative study of retinal vessel segmentation methods on a new publicly available database , 2004, SPIE Medical Imaging.

[18]  Tom Fawcett,et al.  An introduction to ROC analysis , 2006, Pattern Recognit. Lett..

[19]  Yoshua Bengio,et al.  How transferable are features in deep neural networks? , 2014, NIPS.

[20]  Eric R. Ziegel,et al.  The Elements of Statistical Learning , 2003, Technometrics.

[21]  Suraya Mohammad,et al.  Texture analysis for glaucoma classification , 2015, 2015 International Conference on BioSignal Analysis, Processing and Systems (ICBAPS).

[22]  J. Jan,et al.  Analysis of retinal nerve fiber layer via Markov random fields in color fundus images , 2012, 2012 19th International Conference on Systems, Signals and Image Processing (IWSSIP).

[23]  László G. Nyúl,et al.  Effects of Preprocessing Eye Fundus Images on Appearance Based Glaucoma Classification , 2007, CAIP.

[24]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[25]  Stefan Carlsson,et al.  CNN Features Off-the-Shelf: An Astounding Baseline for Recognition , 2014, 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops.

[26]  T. Wong,et al.  Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta-analysis. , 2014, Ophthalmology.

[27]  Bram van Ginneken,et al.  Off-the-shelf convolutional neural network features for pulmonary nodule detection in computed tomography scans , 2015, 2015 IEEE 12th International Symposium on Biomedical Imaging (ISBI).

[28]  Hayit Greenspan,et al.  Deep learning with non-medical training used for chest pathology identification , 2015, Medical Imaging.

[29]  Jayanthi Sivaswamy,et al.  Optic Disk and Cup Segmentation From Monocular Color Retinal Images for Glaucoma Assessment , 2011, IEEE Transactions on Medical Imaging.

[30]  Matthew B. Blaschko,et al.  Discovering predictors of mental health service utilization with k-support regularized logistic regression , 2016, Inf. Sci..

[31]  Ralf-Peter Tornow,et al.  Analysis of Visual Appearance of Retinal Nerve Fibers in High Resolution Fundus Images: A Study on Normal Subjects , 2013, Comput. Math. Methods Medicine.

[32]  Jayanthi Sivaswamy,et al.  A Comprehensive Retinal Image Dataset for the Assessment of Glaucoma from the Optic Nerve Head Analysis , 2015 .

[33]  Bram van Ginneken,et al.  Automatic classification of pulmonary peri-fissural nodules in computed tomography using an ensemble of 2D views and a convolutional neural network out-of-the-box , 2015, Medical Image Anal..

[34]  Attila Budai,et al.  Thickness related textural properties of retinal nerve fiber layer in color fundus images , 2014, Comput. Medical Imaging Graph..

[35]  Joachim Hornegger,et al.  Computer-Aided Diagnostics and Pattern Recognition: Automated Glaucoma Detection , 2015 .

[36]  Andrew Zisserman,et al.  Return of the Devil in the Details: Delving Deep into Convolutional Nets , 2014, BMVC.

[37]  Nathan Srebro,et al.  Sparse Prediction with the $k$-Support Norm , 2012, NIPS.

[38]  J. Crowston,et al.  Optic disk size and glaucoma. , 2007, Survey of ophthalmology.

[39]  Baihua Li,et al.  Automatic extraction of retinal features from colour retinal images for glaucoma diagnosis: A review , 2013, Comput. Medical Imaging Graph..

[40]  Iasonas Kokkinos,et al.  Sub-cortical brain structure segmentation using F-CNN'S , 2016, 2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI).

[41]  Elena Prokofyeva,et al.  Epidemiology of Major Eye Diseases Leading to Blindness in Europe: A Literature Review , 2011, Ophthalmic Research.

[42]  M. Sonka,et al.  Retinal Imaging and Image Analysis. , 2010, IEEE transactions on medical imaging.

[43]  Frédéric Zana,et al.  Segmentation of vessel-like patterns using mathematical morphology and curvature evaluation , 2001, IEEE Trans. Image Process..

[44]  Matthew B. Blaschko,et al.  A Discriminatively Trained Fully Connected Conditional Random Field Model for Blood Vessel Segmentation in Fundus Images , 2017, IEEE Transactions on Biomedical Engineering.

[45]  Tien Yin Wong,et al.  Retinal vascular geometry and glaucoma: the Singapore Malay Eye Study. , 2013, Ophthalmology.

[46]  Mariano Alcañiz Raya,et al.  Automatic Detection of Optic Disc Based on PCA and Mathematical Morphology , 2013, IEEE Transactions on Medical Imaging.

[47]  B. H. Romeny,et al.  Curvature Based Biomarkers for Diabetic Retinopathy via Exponential Curve Fits in SE(2) , 2015 .

[48]  Xiang Zhang,et al.  OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks , 2013, ICLR.