A Semi-Supervised Learning Approach for Automatic Segmentation of Retinal Lesions Using SURF Blob Detector and Locally Adaptive Binarization
In the clinical ophthalmology, the retinal area is routinely investigated from the retinal images, by the naked eyes. Such subjective assessment may be apparently influenced by ineligible inaccuracies. Therefore, objective assessment of the retinal image records plays an important role for the clinical evaluation and treatment planning. Retinal lesions in premature born children represent one of the most frequent retinal findings which may endanger their vison. These findings are mostly connected with the Retinopathy of Prematurity (RoP). In this paper, we have proposed a novel segmentation model utilizing the SURF blob detector and locally adaptive binarization. The proposed model is able to autonomously detect, and consequently classify retinal lesions. In the result, we obtain a segmentation model of the retinal lesions, where the retinal posterior is effectively separated. As a part of the proposed analysis, we have done objectification and quantitative comparison of the proposed method against some of the state of the art segmentation models by selected evaluating parameters. The proposed method has a potential to be used in the clinical practice as a feedback for the automatic evaluation of the retinal lesions, and also for dynamic retinal lesion’s features extraction.
KeywordsSemi-Supervised learning Image segmentation Retinal lesions RetCam 3 Optical disc
The work and the contributions were supported by the project SV4508811/2101Biomedical Engineering Systems XIV’. This study was also supported by the research project The Czech Science Foundation (GACR) 2017 No. 17-03037S Investment evaluation of medical device development run at the Faculty of Informatics and Management, University of Hradec Kralove, Czech Republic. This study was supported by the research project The Czech Science Foundation (TACR) ETA No. TL01000302 Medical Devices development as an effective investment for public and private entities.
- 2.Kubicek, J., et al.: Optical nerve segmentation using the active shape method. Lékař a technika-Clinician Technol. 46(1), 13–20 (2016)Google Scholar
- 4.Zhang, J., Hu, J.: Image segmentation based on 2D Otsu method with histogram analysis. In: 2008 International Conference on Computer Science and Software Engineering. IEEE (2008)Google Scholar
- 9.Zhu, X.: Semi-supervised learning literature survey. Comput. Sci. Univ. Wis. Madison 2(3), 4 (2006)Google Scholar
- 10.Bandyopadhyay, T., Mitra, S., Mitra, S., Rato, L.M., Das, N.: Analysis of pancreas histological images for glucose intolerance identification using wavelet decomposition. In: Satapathy, S.C., Bhateja, V., Udgata, S.K., Pattnaik, P.K. (eds.) Proceedings of the 5th International Conference on Frontiers in Intelligent Computing: Theory and Applications. AISC, vol. 515, pp. 653–661. Springer, Singapore (2017). https://doi.org/10.1007/978-981-10-3153-3_65CrossRefGoogle Scholar
- 11.Hartigan, J.A., Wong, M.A.: Algorithm AS 136: a k-means clustering algorithm. J. Roy. Stat. Soc. Ser. C (Appl. Stat.) 28(1), 100–108 (1979)Google Scholar
- 15.Kubicek, J., Timkovic, J., Krestanova, A., Augustynek, M., Penhaker, M., Bryjova, I.: Morphological segmentation of retinal blood vessels and consequent tortuosity extraction. J. Telecommun. Electron. Comput. Eng. 10(1–4), 73–77 (2018)Google Scholar
- 16.Kubicek, J., Kosturikova, J., Penhaker, M., Augustynek, M., Kuca, K.: Segmentation based on gabor transformation with machine learning: Modeling of retinal blood vessels system from retcam images and tortuosity extraction. Front. Artif. Intell. Appl. 297, 270–283 (2017)Google Scholar