Lightweight deep learning models for detecting COVID-19 from chest X-ray images

Abstract

Deep learning methods have already enjoyed an unprecedented success in medical imaging problems. Similar success has been evidenced when it comes to the detection of COVID-19 from medical images, therefore deep learning approaches are considered good candidates for detecting this disease, in collaboration with radiologists and/or physicians. In this paper, we propose a new approach to detect COVID-19 via exploiting a conditional generative adversarial network to generate synthetic images for augmenting the limited amount of data available. Additionally, we propose two deep learning models following a lightweight architecture, commensurating with the overall amount of data available. Our experiments focused on both binary classification for COVID-19 vs Normal cases and multi-classification that includes a third class for bacterial pneumonia. Our models achieved a competitive performance compared to other studies in literature and also a ResNet8 model. Our binary model achieved 98.7% accuracy, 100% sensitivity and 98.3% specificity, while our three-class model achieved 98.3% accuracy, 99.3% sensitivity and 98.1% specificity. Moreover, via adopting a testing protocol proposed in literature, our models proved to be more robust and reliable in COVID-19 detection than a baseline ResNet8, making them good candidates for detecting COVID-19 from posteroanterior chest X-ray images.

Medical imaging and computational image analysis in COVID-19 diagnosis: A review

“… Quan et al (2021) [ 258 ] Classification and segmentation of COVID-19 lesions CXR CNN 9432 781 90.7 The DenseCapsNet has been proposed. Karakanis and Leontidis (2021) [ 259 ] Automatic COVID-19 diagnosis CXR CNN and GAN 435 145 98.7 The GAN model has been used for data augmentation. Jin et al (2021) [ 260 ] Automatic COVID-19 diagnosis CXR CNN 1743 543 98.64 A hybrid ensemble model, including a pre-trained AlexNet as feature extractor and an SVM classifier as the classifier, has been proposed.…”

COVID-19 Diagnosis in Chest X-rays Using Deep Learning and Majority Voting

“…While we used the same dataset, the problem that we address in this paper is more challenging since it also considers a third class of other pneumonia cases. More recently, in order to mitigate the problem of the small size of available COVID-19 datasets, Karakanis and Leontidis [22] used a conditional generative adversarial network (cGAN [23]) for data augmentation. Accordingly, they generated realistic synthetic images only for the under-represented COVID-19 class, since the two other classes had a sufficient number of original images.…”

WP-UNet: Weight Pruning U-Net with Depth-wise Separable Convolutions for Semantic Segmentation of Kidney Tumours

“…WP-UNet block organized with sequence of layers like depth-wise separable convolutional layer, lters, ReLu, batch normalization, and max pool layers as show in gure 3. Here depth-wise separable convolutinoal layers are used which is much more commonly used in deep learning (e.g Mobile Net and Xception) for embedded devices (Karakanis 2020 ). The proposed model with an input image of size H x W x D, if we do depth-wise separable convolution (stride =1, padding=0) with Nc kernals of size e x e x d, where e is even, then the multiplications in transformation for depthwise separable convolution is ( e x e + Nc) x D x ( H -e +1 ) x ( W -e +1) which is less with 2D convolution transformation Nc x e x e x D x (H-e+1) x (W-e+1) .…”

“…Currently, there is an increased need to deploy deep learning solutions on mobile handheld devices (Hooman Vaseli, 2019), embedded systems (Karakanis al., 2020), or machines with minimal resources.…”