Melanoma Recognition in Dermoscopy Images via Deep Residual Network
Li Hang, Yu Zhen, Ni Dong, Lei Baiying, Wang Tianfu*
Department of Biomedical Engineering, School of Medicine, Shenzhen University, National-Regional Key Technology Engineering Laboratory for Medical Ultrasound, Guangdong Key Laboratory for Biomedical Measurements and Ultrasound Imaging, Shenzhen 518060, Guangdong, China
Abstract：Malignant melanoma is one of the most common and deadly skin cancers. Clinically, dermoscopy is a routine method for early diagnosis of malignant melanoma. However, human's visual examinations are laborious, time-consuming, and highly dependent on dermatologist’s clinical experience. Therefore, it is important to design an algorithm for recognizing melanoma automatically in dermoscopy images. This study proposed a novel framework for the evaluation of dermoscopy images, using deep learning to generate more discriminative features with limited training data. Specifically, we first extracted the intermediate convolutional features of each skin lesion image using a very deep residual neural network including 152 network layers (i.e. Res-152) which was pre-trained on a large natural image dataset, and the final deep representation was obtained by averaging the spatial feature maps into single feature vector, then, the support vector machine (SVM) was used to classify the melanoma. By using the proposed method 248 melanoma images and 1031 non-melanoma images in published ISBI 2016 challenge datasets of skin lesion images were evaluated, obtaining accuracy rate of 84.96% and AUC of 84.18%. In addition, in order to demonstrate the effect of neural network depth on the classification results, we compared the different depth of the model framework. Our approach, which could solve large variations in melanoma and small differences between melanoma and non-melanoma with the limited training data, can produce more discriminative representations than existing studies using hand-crafted features (i.e. the BoF models based on densely sampled SIFT (DSIFT) descriptors) or only to extract features from the fully connected layers.
李航,余镇,倪东,雷柏英,汪天富. 基于深度残差网络的皮肤镜图像黑色素瘤的识别[J]. 中国生物医学工程学报, 2018, 37(3): 274-282.
Li Hang, Yu Zhen, Ni Dong, Lei Baiying, Wang Tianfu. Melanoma Recognition in Dermoscopy Images via Deep Residual Network. Chinese Journal of Biomedical Engineering, 2018, 37(3): 274-282.
[1]Kasmi R,Mokrani K. Classification of malignant melanoma and benign skin lesions: implementation of automatic abcd rule[J]. Iet Image Processing, 2016, 10(6): 448-455. [2]Celebi M, Kingravi H, Uddin B, et al. A methodological approach to the classification of dermoscopy images[J]. Computerized Medical Imaging & Graphics, 2007, 31(6): 362-373. [3]Codella N, Cai Junjie, Abedini M, et al. Deep learning, sparse coding, and svm for melanoma recognition in dermoscopy images[C]//Machine Learning in Medical Imaging: 6th International Workshop(MLMI 2015). Munich: Springer International Publishing, 2015: 118-126. [4]Stolz W, Reimann A,Cognetta A. Abcd rule of dermatoscopy: a new practical method for early recognition of malignant melanoma[J]. European Journal of Dermatology, 1994, 4(7): 521-527. [5]Henning J, Dusza S, Wang S, et al. The cash (color, architecture, symmetry, and homogeneity) algorithm for dermoscopy[J]. Journal of the American Academy of Dermatology, 2007, 56(1): 45-52. [6]Lee T, Ng V, Gallagher R, et al. Dullrazor: A software approach to hair removal from images[J]. Computers in Biology & Medicine, 1997, 27(6): 533-543. [7]Sáez A, Serrano C,Acha B. Model-based classification methods of global patterns in dermoscopic images[J]. IEEE Transactions on Medical Imaging, 2014, 33(5): 1137-1147. [8]Bi Lei, Kim J, Ahn E, et al. Automatic melanoma detection via multi-scale lesion-biased representation and joint reverse classification[C]//2016 IEEE 13th International Symposium on Biomedical Imaging. Prague: IEEE, 2016: 1055-1058. [9]Iyatomi H, Celebi M, Schaefer G, et al. Automated color calibration method for dermoscopy images[J]. Computerized Medical Imaging & Graphics, 2011, 35(2): 89-98. [10]Schaefer G, Rajab M, Celebi M, et al. Colour and contrast enhancement for improved skin lesion segmentation[J]. Computerized Medical Imaging & Graphics, 2011, 35(2): 99-104. [11]Shimizu K, Iyatomi H, Celebi M, et al. Four-class classification of skin lesions with task decomposition strategy[J]. IEEE Trans Biomed Eng, 2015, 62(1): 274-283. [12]Capdehourat G, Corez A, Bazzano A, et al. Toward a combined tool to assist dermatologists in melanoma detection from dermoscopic images of pigmented skin lesions[J]. Pattern Recognition Letters, 2011, 32(16): 2187-2196. [13]Song Youyi, Zhang Ling, Chen Siping, et al. Accurate segmentation of cervical cytoplasm and nuclei based on multiscale convolutional network and graph partitioning[J]. IEEE Transactions on Biomedical Engineering, 2015, 62(10): 2421-2433. [14]Krizhevsky A, Sutskever I,Hinton G. Imagenet classification with deep convolutional neural networks[C]//The 19th International Conference on Neural Information Processing Systems (ICONIP 2012). Doha: Curran Associate Inc, 2012: 1097-1105. [15]Lecun Y, Bengio Y,Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436-444. [16]Ronneberger O, Fischer P,Brox T. U-net: convolutional networks for biomedical image segmentation[C]//The 18th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2015). Munich: Springer International Publishing, 2015: 234-241. [17]Razavian A, Azizpour H, Sullivan J, et al. Cnn features off-the-shelf: an astounding baseline for recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition Workshops. Columbus: IEEE Computer Society, 2014: 512-519. [18]Donahue J, Jia Y, Vinyals O, et al. Decaf: A deep convolutional activation feature for generic visual recognition[J]. 2013, 50(1): 1-647. [19]Deng Jia, Dong Wei, Socher R, et al. Imagenet: a large-scale hierarchical image database[C]//IEEE Computer Vision and Pattern Recognition (CVPR 2009). Florida: IEEE Computer Society, 2009: 248-255. [20]He Kaiming, Zhang Xiangyu, Ren Shaoping, et al. Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2016). Las Vegas: IEEE Computer Society, 2016: 770-778. [21]Yu Lequan, Chen Hao, Dou Qi, et al. Automated melanoma recognition in dermoscopy images via very deep residual networks[J]. IEEE Transactions on Medical Imaging, 2017, 36(4): 994-1004. [22]Yosinski J, Clune J, Bengio Y, et al. How transferable are features in deep neural networks?[C]//Advances in Neural Information Processing Systems. Montreal: MIT Press, 2014: 3320-3328.
