Digitale Bildverarbeitung und Tiefe Neuronale Netze in der Augenheilkunde – aktuelle Trends

Zusammenfassung

Der Einsatz von Tiefen Neuronalen Netzen (Deep Learning) eröffnet neue Möglichkeiten in der digitalen Bildverarbeitung. Auch für die Auswertung von Bilddaten in der Ophthalmologie wird diese Methode erfolgreich eingesetzt und findet weite Verbreitung. In diesem Artikel wird die methodische Vorgehensweise beim Deep Learning betrachtet und der klassischen Vorgehensweise für die Entwicklung von Methoden für die digitale Bildverarbeitung gegenübergestellt. Dabei wird auf Unterschiede eingegangen und die wichtiger werdende Rolle von Trainingsdaten für die Modellbildung erklärt. Weiterhin wird die Vorgehensweise des Transfer-Lernens (Transfer Learning) für Deep Learning am Beispiel eines Datensatzes aus der kornealen Konfokalmikroskopie vorgestellt. Dabei wird auf die Vorteile der Methode und auf Besonderheiten beim Umgang mit medizinischen Mikroskopdaten eingegangen.

Abstract

The use of deep neural networks (“deep learning”) creates new possibilities in digital image processing. This approach has been widely applied and successfully used for the evaluation of image data in ophthalmology. In this article, the methodological approach of deep learning is examined and compared to the classical approach for digital image processing. The differences between the approaches are discussed and the increasingly important role of training data for model generation is explained. Furthermore, the approach of transfer learning for deep learning is presented with a representative data set from the field of corneal confocal microscopy. In this context, the advantages of the method and the specific problems when dealing with medical microscope data will be discussed.

Schlüsselwörter

digitale Bildverarbeitung - maschinelles Lernen - Tiefe Neuronale Netze - Transfer-Lernen

Key words

digital image processing - machine learning - deep learning - transfer learning

• Literatur

• 1 Bourne RRA, Flaxman SR, Braithwaite T. et al. Magnitude, temporal trends, and projections of the global prevalence of blindness and distance and near vision impairment: a systematic review and meta-analysis. Lancet Glob Health 2017; 5: e888-e897 doi:10.1016/S2214-109X(17)30293-0
• 2 Foot B, MacEwen C. Surveillance of sight loss due to delay in ophthalmic treatment or review: frequency, cause and outcome. Eye (Lond) 2017; 31: 771-775 doi:10.1038/eye.2017.1
• 3 De Fauw J, Ledsam JR, Romera-Paredes B. et al. Clinically applicable deep learning for diagnosis and referral in retinal disease. Nat Med 2018; 24: 1342-1350 doi:10.1038/s41591-018-0107-6
• 4 Ting DSW, Pasquale LR, Peng L. et al. Artificial intelligence and deep learning in ophthalmology. Br J Ophthalmol 2019; 103: 167-175 doi:10.1136/bjophthalmol-2018-313173
• 5 Gad H, Khan A, Akhtar N. et al. Corneal nerve and endothelial cell damage in patients with transient ischemic attack and minor ischemic stroke. PLoS One 2019; 14: e0213319 doi:10.1371/journal.pone.0213319
• 6 Petropoulos IN, Ponirakis G, Khan A. et al. Corneal confocal microscopy: ready for prime time. Clin Exp Optom 2019;
• 7 Bohn S, Sperlich K, Allgeier S. et al. Cellular in vivo 3D imaging of the cornea by confocal laser scanning microscopy. Biomed Opt Express 2018; 9: 2511-2525 doi:10.1364/BOE.9.002511
• 8 Allgeier S, Maier S, Mikut R. et al. Mosaicking the subbasal nerve plexus by guided eye movements. Invest Ophthalmol Vis Sci 2014; 55: 6082-6089 doi:10.1167/iovs.14-14698
• 9 Tavakoli M, Ferdousi M, Petropoulos IN. et al. Normative values for corneal nerve morphology assessed using corneal confocal microscopy: a multinational normative data set. Diabetes Care 2015; 38: 838-843 doi:10.2337/dc14-2311
• 10 Chen X, Graham J, Dabbah MA. et al. An automatic tool for quantification of nerve fibers in corneal confocal microscopy images. IEEE Trans Biomed Eng 2017; 64: 786-794 doi:10.1109/TBME.2016.2573642
• 11 Khan A, Kamran S, Akhtar N. et al. Corneal confocal microscopy detects a reduction in corneal endothelial cells and nerve fibres in patients with acute ischemic stroke. Sci Rep 2018; 8: 17333 doi:10.1038/s41598-018-35298-3
• 12 Gulshan V, Peng L, Coram M. et al. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA 2016; 316: 2402-2410 doi:10.1001/jama.2016.17216
• 13 Sedai S, Antony B, Mahapatra D, Garnavi R. Joint Segmentation and uncertainty Visualization of retinal Layers in optical Coherence Tomography Images using Bayesian deep Learning. In: Stoyanov D, Taylor Z, Ciompi F, Xu Y. eds. Computational Pathology and ophthalmic medical Image Analysis. Cham (Schweiz): Springer Nature; 2018: 219-227
• 14 Poplin R, Varadarajan AV, Blumer K. et al. Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning. Nat Biomed Eng 2018; 2: 158-164 doi:10.1038/s41551-018-0195-0
• 15 Allgeier S, Bartschat A, Bohn S. et al. 3D confocal laser-scanning microscopy for large-area imaging of the corneal subbasal nerve plexus. Sci Rep 2018; 8: 7468 doi:10.1038/s41598-018-25915-6
• 16 Feurer M, Hutter F. Towards further automation in AutoML. In: ICML AutoML Workshop 2018. Im Internet: https://ml.informatik.uni-freiburg.de/papers/18-AUTOML-AutoAutoML.pdf Stand: 17.09.2019
• 17 Thornton C, Hutter F, Hoos HH. et al. Auto-WEKA: combined selection and hyperparameter optimization of classification algorithms. In: Proceedings of the 19th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining 2013. 847-855 Im Internet: http://doi.acm.org/10.1145/2487575.2487629 Stand: 17.09.2019
• 18 Jähne B. Digitale Bildverarbeitung. Berlin, Heidelberg: Springer;; 2012
• 19 Lu L, Zheng Y, Carneiro G, Lin Y. eds. Deep Learning and convolutional neural Networks for medical Image Computing. Advances in Computer Vision and Pattern Recognition. Berlin, Heidelberg: Springer; 2017
• 20 Litjens G, Kooi T, Bejnordi BE. et al. A survey on deep learning in medical image analysis. Med Image Anal 2017; 42: 60-88 doi:10.1016/j.media.2017.07.005
• 21 Angermueller C, Pärnamaa T, Parts L. et al. Deep learning for computational biology. Mol Syst Biol 2016; 12: 878 doi:10.15252/msb.20156651
• 22 Maier A, Syben C, Lasser T. et al. A gentle introduction to deep learning in medical image processing. Z Med Phys 2019; 29: 86-101 doi:10.1016/j.zemedi.2018.12.003
• 23 He K, Zhang X, Ren S. et al. Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition 2016. 770-778 Im Internet: https://ieeexplore.ieee.org/document/7780459 Stand: 17.09.2019
• 24 Szegedy C, Liu W, Jia Y. et al. Going deeper with convolutions. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition 2015. 1-9 Im Internet: https://ieeexplore.ieee.org/document/7298594 Stand: 17.09.2019
• 25 Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. In: International Conference on Learning Representations 2015. Im Internet: https://ieeexplore.ieee.org/document/7486599 Stand: 17.09.2019
• 26 Ronneberger O, Fischer P, Brox T. U-Net: Convolutional networks for biomedical image segmentation. In: International Conference on Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015. 2015: 234-241 Im Internet: https://link.springer.com/chapter/10.1007/978-3-319-24574-4_28 Stand: 17.09.2019
• 27 Weiss K, Khoshgoftaar TM, Wang D. A survey of transfer learning. Journal of Big Data 2016; 3: 9 doi:10.1186/s40537-016-0043-6
• 28 McCarey BE, Edelhauser HF, Lynn MJ. Review of corneal endothelial specular microscopy for FDA clinical trials of refractive procedures, surgical devices and new intraocular drugs and solutions. Cornea 2008; 27: 1-16 doi:10.1097/ICO.0b013e31815892da
• 29 Al-Fahdawi S, Qahwaji R, Al-Waisy AS. et al. A fully automated cell segmentation and morphometric parameter system for quantifying corneal endothelial cell morphology. Comput Methods Programs Biomed 2018; 160: 11-23 doi:10.1016/j.cmpb.2018.03.015
• 30 Apostolopoulos S, Zanet SD, Ciller C. et al. Pathological OCT retinal layer segmentation using branch residual U-shape networks. In: International Conference on Medical Image Computing and Computer Assisted Intervention – MICCAI 2017. 294-301 Im Internet: https://link.springer.com/chapter/10.1007/978-3-319-66179-7_34 Stand: 17.09.2019
• 31 Colonna A, Scarpa F, Ruggeri A. Segmentation of corneal Nerves using a U-net-based convolutional neural Network. In: Stoyanov D, Taylor Z, Ciompi F, Xu Y. eds. Computational Pathology and ophthalmic medical Image Analysis. Cham (Schweiz): Springer Nature; 2018: 185-192
• 32 Daniel MC, Atzrodt L, Bucher F. et al. Automated segmentation of the corneal endothelium in a large set of ‘real-world’ specular microscopy images using the U-Net architecture. Sci Rep 2019; 9: 4752 doi:10.1038/s41598-019-41034-2
• 33 Devalla SK, Renukanand PK, Sreedhar BK. et al. DRUNET: a dilated-residual U-Net deep learning network to segment optic nerve head tissues in optical coherence tomography images. Biomed Opt Express 2018; 9: 3244-3265 doi:10.1364/BOE.9.003244
• 34 Dos Santos VA, Schmetterer L, Stegmann H. et al. CorneaNet: fast segmentation of cornea OCT scans of healthy and keratoconic eyes using deep learning. Biomed Opt Express 2019; 10: 622-641 doi:10.1364/BOE.10.000622
• 35 Mathai TS, Lathrop K, Galeotti J. Learning to segment corneal tissue interfaces in OCT images. Preprint arXiv: 1810.06612 2018. Im Internet: https://arxiv.org/abs/1810.06612 Stand: 17.09.2019
• 36 Quellec G, Charrière K, Boudi Y. et al. Deep image mining for diabetic retinopathy screening. Med Image Anal 2017; 39: 178-193 doi:10.1016/j.media.2017.04.012
• 37 Roy AG, Conjeti S, Karri SPK. et al. ReLayNet: retinal layer and fluid segmentation of macular optical coherence tomography using fully convolutional networks. Biomed Opt Express 2017; 8: 3627-3642 doi:10.1364/BOE.8.003627
• 38 Shah A, Zhou L, Abrámoff MD. et al. Multiple surface segmentation using convolution neural nets: application to retinal layer segmentation in OCT images. Biomed Opt Express 2018; 9: 4509-4526 doi:10.1364/BOE.9.004509
• 39 Vigueras-Guillén JP, Sari B, Goes SF. et al. Fully convolutional architecture vs. sliding-window CNN for corneal endothelium cell segmentation. BMC Biomed Eng 2019; 1: 4 doi:10.1186/s42490-019-0003-2
• 40 Zhang T, Elazab A, Wang X. et al. A novel technique for robust and fast segmentation of corneal layer interfaces based on spectral-domain optical coherence tomography imaging. IEEE Access 2017; 5: 10352-10363 doi:10.1109/ACCESS.2017.2712767
• 41 Deng J, Dong W, Socher R. et al. ImageNet: a large-scale hierarchical Image Database. In: 2009 IEEE Conference on Computer Vision and Pattern Recognition 2009. 248-255 Im Internet: https://ieeexplore.ieee.org/document/5206848 Stand: 17.09.2019
• 42 Bartschat A, Stegmaier J, Allgeier S. et al. Augmentations of the bag of visual words approach for real-time fuzzy and partial image classification. In: Proceedings. 27. Workshop Computational Intelligence, Dortmund 2017. 227-242 doi:10.5445/KSP/1000074341
• 43 Prodanova N, Stegmaier J, Allgeier S. et al. Transfer learning with human corneal tissues: an analysis of optimal cut-off layer. Preprint arXiv: 1806.07073 2018. Im Internet: https://arxiv.org/abs/1806.07073 Stand: 17.09.2019