Detection of bladder cancer with feature fusion, transfer learning and CapsNets
作者：
Highlights：
• A system to predict and classify bladder cancer according to its stage was proposed.
• It was proven the usability of features and DL approaches to classify bladder cancers.
• A feature fusion approach of LBP, DWT and new color-dependent HOG features was used.
• Capsule Networks showed potential to overcome CNNs weaknesses.
• 92% and 93%–96% accuracies were reached for Ensemble and DL approaches, respectively.
摘要
•A system to predict and classify bladder cancer according to its stage was proposed.•It was proven the usability of features and DL approaches to classify bladder cancers.•A feature fusion approach of LBP, DWT and new color-dependent HOG features was used.•Capsule Networks showed potential to overcome CNNs weaknesses.•92% and 93%–96% accuracies were reached for Ensemble and DL approaches, respectively.
论文关键词：Bladder tumor,Capsule neural networks,Decision fusion,Ensemble learning,Transfer learning
论文评审过程：Received 29 June 2021, Revised 22 February 2022, Accepted 2 March 2022, Available online 6 March 2022, Version of Record 8 March 2022.
论文官网地址：https://doi.org/10.1016/j.artmed.2022.102275
