Classifying Pneumonia among Chest X-Rays Using Transfer Learning

Abstract

Chest radiography has become the modality of choice for diagnosing pneumonia. However, analyzing chest X-ray images may be tedious, time-consuming and requiring expert knowledge that might not be available in less-developed regions. therefore, computer-aided diagnosis systems are needed. Recently, many classification systems based on deep learning have been proposed. Despite their success, the high development cost for deep networks is still a hurdle for deployment. Deep transfer learning (or simply transfer learning) has the merit of reducing the development cost by borrowing architectures from trained models followed by slight fine-tuning of some layers. Nevertheless, whether deep transfer learning is effective over training from scratch in the medical setting remains a research question for many applications. In this work, we investigate the use of deep transfer learning to classify pneumonia among chest Xray images. Experimental results demonstrated that, with slight fine-tuning, deep transfer learning brings performance advantage over training from scratch. Three models, ResNet-50, Inception V3 and DensetNet121, were trained separately through transfer learning and from scratch. The former can achieve a 4.1% to 52.5% larger area under the curve (AUC) than those obtained by the latter, suggesting the effectiveness of deep transfer learning for classifying pneumonia in chest X-ray images.

Chest X-ray analysis empowered with deep learning: A systematic review

“…For instance, Irfan et al. [23] have carried out a deep transfer learning with ResNet-50, InceptionV3 and DenseNet121 models separately for Pneumonia classification using chest X-ray images. They have achieved higher accuracy using transfer learning compared to training from the scratch for each model.…”

“…Moreover, Irfan et al. [23] , have carried out a deep transfer learning with ResNet-50, Inception V3 and DenseNet121 models separately, to classify chest X-ray images for Pneumonia, where the ResNet-50 model has achieved better accuracy compared to Inception V3.…”

“…A study for Pneumonia classification by Irfan et al. [23] , which was carried out a deep transfer learning with ResNet-50, Inception V3 and DenseNet121 models separately, has achieved the highest accuracy from DenseNet121. Varshani et al.…”

“…Another fact to be noticed is that the VGG model is mostly used with transfer learning compared to other studies. As indicated in the reviewed studies, the use of transfer learning can be due to several reasons such as usage of small datasets [54] , need to achieve better performance [23] and the need to improve the training time [69] . However, the mean accuracy value of the studies that have used transfer learning is 92.2%, while the mean accuracy of the studies that have not used transfer learning is 92.5%.…”

“…Among many DL techniques, CNNs have proved to be highly effective in the implementation of medical image analysis frameworks. Therefore, we focus on the applicability of CNNs with different approaches such as building CNNs from scratch [22] , applying transfer learning with popular models [23] and using ensembles of multiple models [24] . This systematic review provides a comprehensive list of datasets with COVID-19 and Pneumonia images and analyzes the state-of-the-art techniques used in the literature in each phase in the DL process.…”