Conference paper

Pediatric pneumonia detection using vision transformer

Denis Eka Cahyani, Faisal Farris Setyawan, Anjar Dwi Hariadi, Sapti Wahyuningsih, Samsul Setumin

AIP Conference Proceedings , Vol. 3189 (1) · 2024

2Citations
10Highest percentile
5Authors

Summary

Summary Pediatric pneumonia detection using vision transformer
Key quote

The Vision Transformer model generated good performance for pediatric pneumonia detection based on CXR.

Cahyani et al., 2024

Details

Abstract

Pneumonia is the leading cause of death in children and one of the leading causes of death worldwide. Early detection of pneumonia in pediatric is needed so that children get proper treatment and avoid wider transmission. This study aims to detect pneumonia in children based on chest X-ray (CXR) images by comparing the Vision Transformer (ViT), InceptionV3, and Xception models. Vision Transformer (ViT) is a new transformer model that trains neural networks on image data. Meanwhile, the InceptionV3 and Xception models are transfer learning models in a convolutional neural network. This study uses a dataset called the Pediatric Chest X-ray Pneumonia, which contains two classes, namely normal and pneumonia. Vision Transformer achieved the best accuracy, precision, recall, and F1-measure values of 97.78, 97.46, 97.88, and 97.16. Furthermore, the second model, namely InceptionV3, obtained accuracy, precision, recall, and F1-measure values of 97.18, 96.06, 96.87, and 96.45. Finally, the Xception model obtained accuracy, precision, recall, and F1-measure values of 97.01, 95.85, 96.65, and 96.24. So, the conclusion of this research is that the Vision Transformer model can generate good performance for pediatric pneumonia detection based on CXR. © 2024 Author(s).

Affiliations

Department of Mathematics, Faculty of Mathematics and Natural Sciences, Universitas Negeri Malang, Malang, Indonesia; School of Electrical Engineering, Universiti Teknologi MARA, Cawangan Pulau Pinang, Malaysia

American Institute of Physics · ISSN 0094243X