Adaptive melanoma diagnosis using evolving clustering, ensemble and deep neural networks
作者：
Highlights：
• We conduct skin lesion segmentation and classification in dermoscopic images.
• A modified PSO model fine-tunes K-Means cluster centres for lesion segmentation.
• Evolving deep CNNs with optimal topologies and hyper-parameters are devised.
• PSO-based feature selection for ensemble lesion classification is also conducted.
• Our systems depict a superior capability in image segmentation and classification.
摘要
•We conduct skin lesion segmentation and classification in dermoscopic images.•A modified PSO model fine-tunes K-Means cluster centres for lesion segmentation.•Evolving deep CNNs with optimal topologies and hyper-parameters are devised.•PSO-based feature selection for ensemble lesion classification is also conducted.•Our systems depict a superior capability in image segmentation and classification.
论文关键词：Skin lesion segmentation and classification,Feature selection,Clustering,Evolutionary algorithm,Evolving convolutional neural network and ensemble classifier
论文评审过程：Received 9 December 2018, Revised 16 June 2019, Accepted 19 June 2019, Available online 25 June 2019, Version of Record 18 November 2019.
论文官网地址：https://doi.org/10.1016/j.knosys.2019.06.015
