Efficient Pneumonia Detection in Chest Xray Images Using Deep Transfer Learning

Abstract

Pneumonia causes the death of around 700,000 children every year and affects 7% of the global population. Chest X-rays are primarily used for the diagnosis of this disease. However, even for a trained radiologist, it is a challenging task to examine chest X-rays. There is a need to improve the diagnosis accuracy. In this work, an efficient model for the detection of pneumonia trained on digital chest X-ray images is proposed, which could aid the radiologists in their decision making process. A novel approach based on a weighted classifier is introduced, which combines the weighted predictions from the state-of-the-art deep learning models such as ResNet18, Xception, InceptionV3, DenseNet121, and MobileNetV3 in an optimal way. This approach is a supervised learning approach in which the network predicts the result based on the quality of the dataset used. Transfer learning is used to fine-tune the deep learning models to obtain higher training and validation accuracy. Partial data augmentation techniques are employed to increase the training dataset in a balanced way. The proposed weighted classifier is able to outperform all the individual models. Finally, the model is evaluated, not only in terms of test accuracy, but also in the AUC score. The final proposed weighted classifier model is able to achieve a test accuracy of 98.43% and an AUC score of 99.76 on the unseen data from the Guangzhou Women and Children’s Medical Center pneumonia dataset. Hence, the proposed model can be used for a quick diagnosis of pneumonia and can aid the radiologists in the diagnosis process.

Intelligent Pneumonia Identification From Chest X-Rays: A Systematic Literature Review

“…The current best performance with a low computational requirement was provided by Sousa et al [87]. However, the works of Hashmi et al [136] and Chouhan et al [131] can also be considered remarkable if computational power is not constrained. Therefore, the main challenges to be addressed in the future are listed below:…”

“…Hashmi et al [136] proposed a weighted classifier comprising different DL models (ResNet18, Xception, InceptionV3, DenseNet121, and MobileNetV3), which can be used to identify pneumonia from CXRs [2]. The experimental results confirmed a remarkable accuracy of 98.43% and precision and recall values of 98.26% and 99.0%, respectively.…”

Deep Learning Assisted Localization of Polycystic Kidney on Contrast-Enhanced CT Images

“…Secondly, we conducted subject-wise testing. Both testing evaluations were carried out by comparing our method with other well-established DL pre-trained object detection models including CNN backbones, which have been adopted in similar current approaches for pneumonia detection in chest X-ray [ 28 ] and for malignant pulmonary nodule detection in CT scans [ 29 ]. These models and backbones were SSD Inception V2, Faster-RCNN, ResNet, and MobiNet.…”

Which Current and Novel Diagnostic Avenues for Bacterial Respiratory Diseases?

“…Artificial intelligence (AI), which refers to systems that imitate human thought and actions, is now broadly used to help interpret CXRs for the diagnosis of respiratory infections. X-ray imaging is the most common and available diagnostic technique used in the world, however specialist workers are not necessarily trained in advanced analysis and this led to the development of AI-aided strategies that support clinicians, with the advantage of limited cost ( Hashmi et al, 2020 ). Machine learning (ML) and deep learning (DL) are part of AI and have garnered a lot of attention over the past 2 years.…”

“…They consist in multi-layer neural networks that recognize visual patterns from pixel images ( Shin et al, 2016 ). Among the different tools recently developed, the existing CNNs CXNet-m1 ( Xu et al, 2019 ), CheXNeXt ( Rajpurkar et al, 2018 ), VGG16 and VGG19 ( Toǧaçar et al, 2020 ), AlexNet ( Rahman et al, 2020 ; Rajaraman and Antani, 2020 ; Toǧaçar et al, 2020 ), ResNet18 ( Rahman et al, 2020 ; Rajaraman and Antani, 2020 ), DenseNet201 ( Rahman et al, 2020 ), SqueezeNet ( Rahman et al, 2020 ), VGGNet ( Rajaraman and Antani, 2020 ), GoogLeNet ( Saraiva et al, 2019 ), Lastly, Hashmi and collaborators proposed the most accurate and precise model regarding previous developed programs ( Hashmi et al, 2020 ) using ResNet18, Xception, InceptionV3, DenseNet121 and MobileNetV3 CNN algorithms. They could develop a robust model for bacterial pneumonia detection with the help of hospital-scale CXR and CT databases provided respectively from Wang et al (2017) (named ChestX-ray 14) and Kermany et al (2018) .…”