A Lightweight Convolutional Neural Network Model for Child Pneumonia Classification

EasyChair Preprint 4416

Abstract

Pneumonia is still a serious threat for children including newborns. Each year many children died of pneumonia. Physicians diagnose pneumonia through some process including reviewing chest X-rays of patients. While reviewing, a single diagnostic mistake may cause a serious threat to patients. In recent years, Computer-aided detection system (CAD) and medical image classification are progressively turning into another research territory. CAD can reduce physician's effort to review chest X-ray. Currently, Researchers build various models to detect pneumonia. However, there is still a lack of computationally efficient models to diagnose pediatric pneumonia. Further, some off-the-self or pre-trained models are not always suitable for mobile and embedded vision applications since these models are not lightweight. In our research, we built a lightweight convolutional neural network model from scratch which able to learn lung texture features and detect pediatric pneumonia. We compared our proposed model performance with some off-the-shelf models. Our proposed model achieved the best AUC (98.5), test accuracy (94.0), F1 (94.1), Precision (92.3), Specificity (92.1), NPV (95.7) scores. We employed several data augmentation algorithms to increase the model's classification ability.

Keyphrases: AUC, Artificial Intelligence, Bioinformatics, CNN, Chest X-ray, Pediatric Pneumonia, deep learning, image classification, medical image

@booklet{EasyChair:4416,
  author    = {Khan Fashee Monowar and Md. Al Mehedi Hasan and Jungpil Shin},
  title     = {A Lightweight Convolutional Neural Network Model for Child Pneumonia Classification},
  howpublished = {EasyChair Preprint 4416},
  year      = {EasyChair, 2020}}