Classification method for nailfold capillary images using an optimized sugeno fuzzy ensemble of convolutional neural networks.

Journal: Computers in biology and medicine
PMID:

Abstract

This study developed a novel binary classification method for analyzing nailfold capillary images associated with the risk of developing sclerosis. The proposed approach combined a Sugeno fuzzy integral inference system with an ensemble of convolutional neural networks (CNNs), including GoogLeNet, ResNet, and DenseNet. Nailfold capillary images are highly valuable for diagnosing and monitoring various systemic diseases. They can reveal early indicators of systemic sclerosis, such as capillary enlargement, loss, or hemorrhages. The study obtained nailfold capillary images from a hospital in Taiwan, with 80 % allocated for model training and the remaining 20 % reserved for testing purposes. The proposed method achieved a high performance with an accuracy of 85 %, a recall of 81.82 %, a precision of 90 %, and an F1 score of 85.17 %. In comparison, individual CNN models (GoogLeNet, ResNet, and DenseNet) achieved accuracies of 73.33 %, 67.96 %, and 70.83 %, respectively. These results demonstrate that the proposed integrated method outperforms single-model approaches in classifying nailfold capillary images more accurately and efficiency. Using CNN models as a novel application opens new avenues for research in related image analysis fields.

Authors

  • Chiao-Chi Ou
    Department of Industrial Engineering and Enterprise Information, Tunghai University, Taichung City, Taiwan; Department of Medical Research, Taichung Veterans General Hospital, Taichung, Taiwan; Division of Allergy, Immunology, and Rheumatology, Department of Internal Medicine, Taichung Veterans General Hospital, No.1650, Sec. 4, Taiwan Blvd., Xitun Dist., Taichung City, Taiwan. Electronic address: cactus3854@hotmail.com.
  • Yun-Chi Liu
    Department of Industrial Engineering and Enterprise Information, Tunghai University, Taichung City, Taiwan. Electronic address: a881124a@gmail.com.
  • Kuo-Ping Lin
    Department of Industrial Engineering and Enterprise Information, Tunghai University, Taichung City, Taiwan; School of Accounting, University of Economics Ho Chi Minh City, Ho Chi Minh City, Viet Nam. Electronic address: kplin@thu.edu.tw.
  • Tsai-Hung Yen
    Division of Allergy, Immunology, and Rheumatology, Department of Internal Medicine, Taichung Veterans General Hospital, No.1650, Sec. 4, Taiwan Blvd., Xitun Dist., Taichung City, Taiwan; Division of General Medicine, Department of Internal Medicine, Taichung Veterans General Hospital, Taichung, Taiwan. Electronic address: mi552296@gmail.com.
  • Wen-Nan Huang
    Division of Allergy, Immunology, and Rheumatology, Department of Internal Medicine, Taichung Veterans General Hospital, No.1650, Sec. 4, Taiwan Blvd., Xitun Dist., Taichung City, Taiwan; School of Medical, College of Medicine, National Yang Ming Chiao Tung University, Taipei, Taiwan; Department of Post-Baccalaureate Medicine, College of Medicine, National Chung Hsing University, Taichung, Taiwan. Electronic address: gtim5555@vghtc.gov.tw.