Genetic programming for automatic skin cancer image classification
作者：
Highlights：
• Proposed methods construct new models with texture, color, and wavelet features.
• Evolved features are highly informative to discriminate between skin image classes.
• New features improve classification accuracy, efficient in real-time clinic situation.
• Identify prominent visual features to help the dermatologist in making a diagnosis.
• Achieved 86.77% accuracy on difficult dataset, outperforming the state-of-the-arts.
摘要
•Proposed methods construct new models with texture, color, and wavelet features.•Evolved features are highly informative to discriminate between skin image classes.•New features improve classification accuracy, efficient in real-time clinic situation.•Identify prominent visual features to help the dermatologist in making a diagnosis.•Achieved 86.77% accuracy on difficult dataset, outperforming the state-of-the-arts.
论文关键词：Image classification,Genetic programming,Dimensionality reduction,Feature selection,Feature construction
论文评审过程：Received 30 April 2020, Revised 2 February 2022, Accepted 12 February 2022, Available online 24 February 2022, Version of Record 2 March 2022.
论文官网地址：https://doi.org/10.1016/j.eswa.2022.116680
