-
摘要: 随着大数据时代的到来, 深度学习技术在图像分类、检测等任务中相对传统模式识别方法均取得了令人瞩目的突破。2017年1月, 斯坦福大学人工智能实验室采用深度学习方法对皮肤镜和临床皮损图像进行自动分类, 并在《自然》杂志上发表了相关研究成果, 代表了皮肤图像自动分析领域的最新研究进展。本文从数据库建立、研究方法设计以及试验结果分析等角度对这一研究工作进行解读, 并分析国内皮肤影像计算机辅助诊断的研究现状, 以及未来多源皮肤影像大数据分析与智能辅助诊断的发展空间, 以期推进我国皮肤疾病的医疗诊断水平。Abstract: With the advent of the big data era, deep learning has made remarkable breakthroughs compared with traditional pattern-recognition methods in many tasks, such as image classification and detection. In January 2017, the artificial intelligence laboratory of Stanford University applied deep learning to automatic classification of clinical skin images and dermoscopy images, and published the research results in Nature, which represents the latest research progress in the field of automatic analysis of skin images. Herein, we interpret this research from the aspects of database establishment, design of research methods and analysis of the experimental results; we also elaborate the current research status of computer-aided diagnosis of skin images in China, as well as the future developmental prospect of multi-source big data analysis of skin images and intelligent auxiliary diag-nosis, in order to promote the medical diagnostic level of skin diseases in China.
-
图 1 皮肤疾病分类树形结构示意图[10]
图 2 GoogLeNet Inception-v3结构图[10]
-
[1] Schmid-Saugeona P, Guillodb J, Thirana JP.Towards a computeraided diagnosis system for pigmented skin lesions[J]. Comput Med Imaging Graph, 2003, 27:65-78. doi: 10.1016/S0895-6111(02)00048-4 [2] Abbas Q, Celebi ME, García IF. Hair removal methods:a comparative study for dermoscopy images[J]. Biomed Signal Process Control, 2011, 6:395-404. doi: 10.1016/j.bspc.2011.01.003 [3] Xie F, Bovik AC. Automatic segmentation of dermoscopy images using self-generating neural networks seeded by genetic algorithm[J]. Pattern Recognition, 2013, 46:1012-1019. doi: 10.1016/j.patcog.2012.08.012 [4] Sáez A, Serrano C, Acha B. Model-based classification methods of global patterns in dermoscopic images[J]. IEEE Trans Med Imaging, 2014, 33:1137-1147. doi: 10.1109/TMI.2014.2305769 [5] Isasi AG, Zapirain BG, Zorrilla AM. Mendez. Melanomas Non-invasive Diagnosis Application Based on the ABCD Rule and Pattern Recognition Image Processing Algorithms[J]. Comput Biol Med, 2011, 41:742-755. doi: 10.1016/j.compbiomed.2011.06.010 [6] Lecun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521:436-444. doi: 10.1038/nature14539 [7] Krizhevsky A, Sutskever I, Hinton GE. ImageNet classifica-tion with deep convolutional neural networks[C]. Interna-tional Conference on Neural Information Processing Systems, 2012: 1097-1105. [8] Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the Inception Architecture for Computer Vision[J]. Computer Science, 2015:2818-2826. doi: 10.1109/CVPR.2016.308 [9] Ren S, He K, Girshick R, et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE Trans Pattern Anal Mach Intell, 2017, 39:1137-1149. doi: 10.1109/TPAMI.2016.2577031 [10] Esteva A, Kuprel B, Novoa RA, et al. Dermatologist-level classification of skin cancer with deep neural networks[J]. Nature, 2017, 542:115. doi: 10.1038/nature21056 [11] Masood A, Aljumaily AA. Computer aided diagnostic support system for skin cancer:a review of techniques and algorithms[J]. Int J Biomed Imaging, 2013, 2013:323268. https://www.hindawi.com/journals/ijbi/2013/323268/ [12] Ramlakhan K, Shang Y. A mobile automated skin lesion classification system[C]. IEEE International Conference on TOOLS with Artificial Intelligence, 2011: 138-141. [13] https://www.ericsson.com/res/docs/2016/ericsson-mobility-report-2016.pdf (2016). [14] Han S, Mao H, Dally WJ. Deep compression:compressing deep neural networks with pruning, trained quantization and huffman coding[J]. Fiber, 2016, 56:3-7. https://arxiv.org/abs/1510.00149