Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Influence of background preprocessing on the performance of deep learning retinal vessel detection

Influence of background preprocessing on the performance of deep learning retinal vessel detection Abstract.Purpose: Segmentation of the vessel tree from retinal fundus images can be used to track changes in the retina and be an important first step in a diagnosis. Manual segmentation is a time-consuming process that is prone to error; effective and reliable automation can alleviate these problems but one of the difficulties is uneven image background, which may affect segmentation performance.Approach: We present a patch-based deep learning framework, based on a modified U-Net architecture, that automatically segments the retinal blood vessels from fundus images. In particular, we evaluate how various pre-processing techniques, images with either no processing, N4 bias field correction, contrast limited adaptive histogram equalization (CLAHE), or a combination of N4 and CLAHE, can compensate for uneven image background and impact final segmentation performance.Results: We achieved competitive results on three publicly available datasets as a benchmark for our comparisons of pre-processing techniques. In addition, we introduce Bayesian statistical testing, which indicates little practical difference (Pr  >  0.99) between pre-processing methods apart from the sensitivity metric. In terms of sensitivity and pre-processing, the combination of N4 correction and CLAHE performs better in comparison to unprocessed and N4 pre-processing (Pr  >  0.87); but compared to CLAHE alone, the differences are not significant (Pr  ≈  0.38 to 0.88).Conclusions: We conclude that deep learning is an effective method for retinal vessel segmentation and that CLAHE pre-processing has the greatest positive impact on segmentation performance, with N4 correction helping only in images with extremely inhomogeneous background illumination. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Medical Imaging SPIE

Influence of background preprocessing on the performance of deep learning retinal vessel detection

Owler, James; Rockett, Peter
Journal of Medical Imaging , Volume 8 (6) – Nov 1, 2021
Download PDF
19 pages

Loading next page...
 
/lp/spie/influence-of-background-preprocessing-on-the-performance-of-deep-1VoAf0G8S3

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
SPIE
Copyright
© 2021 Society of Photo-Optical Instrumentation Engineers (SPIE)
ISSN
2329-4302
eISSN
2329-4310
DOI
10.1117/1.jmi.8.6.064001
Publisher site
See Article on Publisher Site

Abstract

Abstract.Purpose: Segmentation of the vessel tree from retinal fundus images can be used to track changes in the retina and be an important first step in a diagnosis. Manual segmentation is a time-consuming process that is prone to error; effective and reliable automation can alleviate these problems but one of the difficulties is uneven image background, which may affect segmentation performance.Approach: We present a patch-based deep learning framework, based on a modified U-Net architecture, that automatically segments the retinal blood vessels from fundus images. In particular, we evaluate how various pre-processing techniques, images with either no processing, N4 bias field correction, contrast limited adaptive histogram equalization (CLAHE), or a combination of N4 and CLAHE, can compensate for uneven image background and impact final segmentation performance.Results: We achieved competitive results on three publicly available datasets as a benchmark for our comparisons of pre-processing techniques. In addition, we introduce Bayesian statistical testing, which indicates little practical difference (Pr  >  0.99) between pre-processing methods apart from the sensitivity metric. In terms of sensitivity and pre-processing, the combination of N4 correction and CLAHE performs better in comparison to unprocessed and N4 pre-processing (Pr  >  0.87); but compared to CLAHE alone, the differences are not significant (Pr  ≈  0.38 to 0.88).Conclusions: We conclude that deep learning is an effective method for retinal vessel segmentation and that CLAHE pre-processing has the greatest positive impact on segmentation performance, with N4 correction helping only in images with extremely inhomogeneous background illumination.

Journal

Journal of Medical Imaging – SPIE

Published: Nov 1, 2021

Keywords: retinal vessel segmentation; deep learning; U-Net; fundus imaging; Bayesian hypothesis testing; image background correction

References

  • Global prevalence and major risk factors of diabetic retinopathy
    Yau, J. W. Y.
  • Global report on diabetes
  • Retinal imaging and image analysis
    Abràmoff, M. D.; Garvin, M. K.; Sonka, M.
  • Advances in retinal imaging and applications in diabetic retinopathy screening: a review
    Fenner, B. J.
  • A personal identification system using retinal vasculature in retinal fundus images
    Köse, C.; İki˙baş, C.
  • Artificial intelligence in medicine
    Hamet, P.; Tremblay, J.
  • The impact of artificial intelligence in medicine on the future role of the physician
    Ahuja, A. S.
  • The potential for artificial intelligence in healthcare
    Davenport, T.; Kalakota, R.
  • Performance comparison of publicly available retinal blood vessel segmentation methods
    Vostatek, P.
  • An ensemble classification-based approach applied to retinal blood vessel segmentation
    Fraz, M. M.
  • Deep learning models for retinal blood vessels segmentation: a review
    Soomro, T. A.
  • Deep learning
    LeCun, Y.; Bengio, Y.; Hinton, G.
  • A survey on deep learning in medical image analysis
    Litjens, G.
  • Handwritten digit recognition: applications of neural network chips and automatic learning
    LeCun, Y.
  • ImageNet classification with deep convolutional neural networks
    Krizhevsky, A.; Sutskever, I.; Hinton, G. E.
  • Segmenting retinal blood vessels with deep neural networks
    Liskowski, P.; Krawiec, K.
  • U-Net: convolutional networks for biomedical image segmentation
    Ronneberger, O.; Fischer, P.; Brox, T.
  • Fully convolutional networks for semantic segmentation
    Long, J.; Shelhamer, E.; Darrell, T.
  • The importance of skip connections in biomedical image segmentation
    Drozdzal, M.
  • Retinal vessel segmentation based on fully convolutional neural networks
    Oliveira, A.; Pereira, S.; Silva, C. A.
  • Joint segment-level and pixel-wise losses for deep learning based retinal vessel segmentation
    Yan, Z.; Yang, X.; Cheng, K.
  • Deep retinal image segmentation with regularization under geometric priors
    Cherukuri, V.
  • A review of machine learning methods for retinal blood vessel segmentation and artery/vein classification
    Mookiah, M. R. K.
  • Applications of deep learning in fundus images: a review
    Li, T.
  • Trainable COSFIRE filters for vessel delineation with application to retinal images
    Azzopardi, G.
  • Adaptive histogram equalization and its variations
    Pizer, S. M.
  • Luminosity and contrast normalization in retinal images
    Foracchia, M.; Grisan, E.; Ruggeri, A.
  • Retinal blood vessels extraction using probabilistic modelling
    Kaba, D.
  • N4ITK: improved N3 bias correction
    Tustison, N. J.
  • The use of the area under the ROC curve in the evaluation of machine learning algorithms
    Bradley, A. P.
  • Why rankings of biomedical image analysis competitions should be interpreted with care
    Maier-Hein, L.
  • Retinal vessel segmentation using the 2-D Gabor wavelet and supervised classification
    Soares, J. V. B.
  • Blood vessel segmentation of fundus images by major vessel extraction and sub-image classification
    Roychowdhury, S.; Koozekanani, D.; Parhi, K.
  • Robust retinal vessel segmentation via locally adaptive derivative frames in orientation scores
    Zhang, J.
  • Is the UCI repository useful for data mining?
    Soares, C.
  • Time for a change: a tutorial for comparing multiple classifiers through Bayesian analysis
    Benavoli, A.
  • Comparing multiple comparisons: practical guidance for choosing the best multiple comparisons test
    Midway, S.
  • The Bonferonni and Šidák corrections for multiple comparisons
    Abdi, H.; Salkind, N. J.; Rasmussen, K.
  • Bayesian estimation supersedes the t-test
    Kruschke, J. K.
  • Ridge-based vessel segmentation in color images of the retina
    Staal, J.
  • Locating blood vessels in retinal images by piecewise threshold probing of a matched filter response
    Hoover, A. D.; Kouznetsova, V.; Goldbaum, M.
  • Measuring retinal vessel tortuosity in 10-year-old children: validation of the computer-assisted image analysis of the retina (CAIAR) program
    Owen, C. G.
  • Rethinking the inception architecture for computer vision
    Szegedy, C.
  • Batch normalization: accelerating deep network training by reducing internal covariate shift
    Ioffe, S.; Szegedy, C.
  • Understanding the difficulty of training deep feedforward neural networks
    Glorot, X.; Bengio, Y.
  • Deep ensembles: a loss landscape perspective
    Fort, S.; Hu, H.; Lakshminarayanan, B.
  • ADAM: a method for stochastic optimization
    Kingma, D. P.; Ba, J.
  • A threshold selection method from gray-level histograms
    Otsu, N.
  • Bayesian hypothesis testing in machine learning
  • Ignorance, myopia, and naiveté in computer vision systems
    Jain, R. C.; Binford, T. O.