Ensemble of fine‐tuned convolutional neural networks for urine sediment microscopic image classification. Issue 1 (20th January 2020)

• Record Type: Journal Article
• Authors: Liu, Wenqian; Li, Weihong; Gong, Weiguo
• Is Part Of: IET computer vision. Volume 14:Issue 1(2020)
• Page Start: 18
• Page End: 25
• Publication Date: 2020-01-20
• Subjects: image classification -- medical image processing -- feature extraction -- learning (artificial intelligence) -- convolutional neural nets; urine sediment microscopic image processing -- pre-trained CNNs -- urine sediment microscopic image dataset -- learning rate -- cascading features -- convolutional layers -- fully connected neural network -- fine-tuned convolutional neural networks -- urine sediment microscopic image classification -- CNN training -- impurity interference -- classification accuracy; Computer vision -- Periodicals; Pattern recognition systems -- Periodicals; 006.37
• DOI: 10.1049/iet-cvi.2018.5829
• Languages: English
• ISSNs: 1751-9632

Abstract

Abstract : In this study, an ensemble of fine‐tuned convolutional neural networks (CNNs) is proposed. As CNN training requires large annotated data, which are lacking in the field of urine sediment microscopic image processing, the authors first pre‐trained the CNNs, including ResNet50 and GoogLeNet, and developed AlexNet on an ImageNet dataset. Thereafter, some of the weights of the pre‐trained CNNs were transferred to the urine sediment microscopic image dataset. To guide fine‐tuning of the learning rate and cascading features, the hierarchical nature of features in different convolutional layers was investigated by visualising the CNN. Then, they combined three CNNs as an ensemble of CNNs to decrease the differences and impurity interference among features of urine sediment microscopic image. These fusion features were employed to train the fully connected neural network for classification. In this study, they improved the accuracy of each CNN by an average of 2.2% through fine‐tuning of the learning rate and cascading features. Moreover, the better experimental results were achieved compared with other state‐of‐the‐art methods and indicated that a 97% classification accuracy can be attained.