Skin lesion segmentation based on spatial group convolutions in dermoscopic images
2
조회
0
다운로드
2
인용
0
참고문헌
키워드
초록
Semantic image segmentation is the process of labeling each image pixel with its associated class label. Most of the image segmentation task use an encoder-decoder-based method, like U-Net and its variants which is a popular strategy for solving segmentation tasks. In this paper, we propose an encoder-decoder-based approach with novel spatial group convolutions (SGC) for skin lesion segmentation. The SGC employ multiple-sized convolution kernels such as 1×1, 3×3, 5×5, 7×7, where each of these k×k kernel operations is decomposed as the combination of k×1 convolution followed by 1×k convolution. The large kernel using two 1-D kernels is likely to have better spatial information of the features and shows the improved model learning with the limited learnable parameters and more discriminative feature extraction. The SGC-based proposed model shows improved segmentation accuracy with a significantly reduced number of model parameters. The mean Jaccard index for the proposed method is 89%, the mean dice score is 93.84%, and the mean accuracy value is 95.99 %, with 0.4 million parameters for the proposed method on the PH2 test dataset.
Abstract
Semantic image segmentation is the process of labeling each image pixel with its associated class label. Most of the image segmentation task use an encoder-decoder-based method, like U-Net and its variants which is a popular strategy for solving segmentation tasks. In this paper, we propose an encoder-decoder-based approach with novel spatial group convolutions (SGC) for skin lesion segmentation. The SGC employ multiple-sized convolution kernels such as 1×1, 3×3, 5×5, 7×7, where each of these k×k kernel operations is decomposed as the combination of k×1 convolution followed by 1×k convolution. The large kernel using two 1-D kernels is likely to have better spatial information of the features and shows the improved model learning with the limited learnable parameters and more discriminative feature extraction. The SGC-based proposed model shows improved segmentation accuracy with a significantly reduced number of model parameters. The mean Jaccard index for the proposed method is 89%, the mean dice score is 93.84%, and the mean accuracy value is 95.99 %, with 0.4 million parameters for the proposed method on the PH2 test dataset.
참고문헌 (31)
- Siegel, R. L., "Cancer statistics, 2017", CA: A Cancer Journal for Clinicians, 67(1), pp.730~ (2017)
- O. J. Wisco, "Prognostic factors for melanoma", Dermatologic Clinics, 30(3), pp.469~485 (2012)
- M. Balch, "Final version of 2009 AJCC melanoma staging and classification", J. clinical oncology, 27(36), pp.6199~6206 (2009)[DOI]
- P.Lameski, "Computer-aided detection of melanoma, a case study", Proceedigns of ETAI 2018, ETAI Society of Macedonia (2018)
- X. Zhang, "Melanoma segmentation based on deep learning’", Computer Assisted Surgery, 22, pp.267~277 (2017)
- Sharma, N., "Automated medical image segmentation techniques", Journal of Medical Physics, 35(1), pp.3~14 (2010)[DOI]
- S. Kurtek, "Statistical analysis of manual segmentations of structures in medical images", Computer Vision and Image Understanding, 117(9), pp.1036~1050 (2013)
- R. Garnavi, "Automatic segmentation of dermoscopy images using histogram thresholding on optimal color channels", Int. J. Med. Med. Sci, 1(2), pp.126~134 (2010)
- G. Sforza, "Using adaptive thresholding and skewness correction to detect gray areas in melanoma in situ images", IEEE Trans. Instrum. Meas, 61(7), pp.1839~1847 (2012)
- D. Mumford, "Optimal approximations by piecewise smooth functions and associated variational problems", Commun. Pure Appl. Math, 42, pp.577~685 (1989)
- M. Emre Celebi, "Border detection in dermoscopy images using statistical region merging", Skin Res. Technol, 14(3), pp.347~353 (2008)
- M.E. Celebi, "Lesion border detection in dermoscopy images", Comput. Med. Imaging Graph, 33(2), pp.148~153 (2008)
- M. Zortea, "A simple weighted thresholding method for the segmentation of pigmented skin lesions in macroscopic images’", Pattern Recogn, 64, pp.92~104 (2017)
- Yading Yuan, "Automatic skin lesion segmentation with fully convolutional deconvolutional networks", CoRR (2017)
- V. Badrinarayanan, "Segnet: A deep convolutional encoder-decoder architecture for image segmentation", IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(12), pp.24812495~ (2017)
- P. Lameski, "Weed detection dataset with rgb images taken under variable light conditions", International Conference on ICT Innovations, pp.112~119 (2017), Springer
- Yu, L., "Automated melanoma recognition in dermoscopy images via very deep residual networks", IEEE transactions on medical imaging, 36(4), pp.994~1004 (2017)
- A. Esteva, "Dermatologistlevel classification of skin cancer with deep neural networks", Nature, 542, pp.115~118 (2017)
- Burdick, J., "Rethinking skin lesion segmentation in a convolutional classifier", Journal of digital imaging, 31(4), pp.435~440 (2018)
- Menegola, A., "'RECOD titans' at ISIC challenge 2017", arXiv preprint arXiv:1703.04819, ~ (2017)
- F. Ccero, "Deep learning and convolutional neural networks in the aid of the classification of melanoma", In Conference on Graphics, Patterns and Images, SIBGRAPI 2016 (2016)
- Codella, "Deep learning ensembles for melanoma recognition in dermoscopy images", IBM Journal of Research and Development, 61(4), pp.5~1 (2017)
- Yuan, Y., "Improving dermoscopic image segmentation with enhanced convolutional-deconvolution networks", IEEE Journal of Biomedical and Health Informatics (2017)[DOI]
- Qian, Chengyao, "A Detection and Segmentation Architecture for Skin Lesion Segmentation on Dermoscopy Images", arXiv:1809.03917, ~ (2018)
- Ronneberger, O., "U-net: Convolutional networks for biomedical image segmentation", International Conference on Medical image computing and computer-assisted intervention, pp.234~241 (2015), Springer
- L. Chen, "Drinet for medical image segmentation", IEEE transactions on medical imaging, 37(11), pp.2453~2462 (2018)
- J. Long, "Fully convolutional networks for semantic segmentation", Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), pp.3431~3440 (2015)
- Z. Zhou, Z., "UNet++ : A nested U-Net architecture for medical image segmentation", arXiv, 3~11 (2018)
- T. Mendonça, "PH2-A dermoscopic image database for research and benchmarking", 2013 35th Annual International Conference of the IEEE Engineering in Medicine and Biology Society(EMBC), pp.5437~5440 (2013)
- L. R. Dice, "Measures of the amount of ecologic association between species", Ecology (1945)
- P. Jaccard, "The distribution of the flora in the alpine zone", New Phytologist (1912)