A mask-guided attention deep learning model for COVID-19 diagnosis based on an integrated CT scan images database

Abstract

The global extent of COVID-19 mutations and the consequent depletion of hospital resources highlighted the necessity of effective computer-assisted medical diagnosis. COVID-19 detection mediated by deep learning models can help diagnose this highly contagious disease and lower infectivity and mortality rates. Computed tomography (CT) is the preferred imaging modality for building automatic COVID-19 screening and diagnosis models. It is well-known that the training set size significantly impacts the performance and generalization of deep learning models. However, accessing a large dataset of CT scan images from an emerging disease like COVID-19 is challenging. Therefore, data efficiency becomes a significant factor in choosing a learning model. To this end, we present a multi-task learning approach, namely, a mask-guided attention (MGA) classifier, to improve the generalization and data efficiency of COVID-19 classification on lung CT scan images. The novelty of this method is compensating for the scarcity of data by employing more supervision with lesion masks, increasing the sensitivity of the model to COVID-19 manifestations, and helping both generalization and classification performance. Our proposed model achieves better overall performance than the single-task baseline and state-of-the-art models, as measured by various popular metrics. In our experiment with different percentages of data from our curated dataset, the classification performance gain from this multi-task learning approach is more significant for the smaller training sizes. Furthermore, experimental results demonstrate that our method enhances the focus on the lesions, as witnessed by both attention and attribution maps, resulting in a more interpretable model.

Deep-Learning-Based COVID-19 Diagnosis and Implementation in Embedded Edge-Computing Device

“…For the training and testing, we used two datasets, COVIDx CT-2 [ 11 , 12 ] and the integrated CT scan dataset [ 43 ]. The COVIDx CT-2 dataset has two diverse, large-scale datasets named COVIDx CT-2A and COVIDx CT-2B, and we chose COVIDx CT-2A.…”

“…To demonstrate the improved discriminative performance of the VGG19 model, we conducted training and testing on a moderately large dataset and compared it with the original VGG19, Nair et al [ 39 ], Chaudhary et al [ 40 ], Perumal et al [ 41 ], Garg et al [ 42 ], and Maftouni et al [ 43 ], as shown in Table 5 . The results presented in Figure 9 indicate that the improved VGG19 outperforms VGG19 in both CAP and COVID-19 categories, achieving a 4.77% higher recall rate.…”

“…Figure 2 shows the workflow of the proposed COVID-19 classification method. Firstly, the input images are selected from datasets [ 11 , 12 , 43 ] and preprocessed using data augmentation and image enhancement. Secondly, the processed images are input into the improved VGG19 to extract image features, and then the classifier is used to classify normal, pneumonia, and COVID-19 cases.…”

“…Garg et al [ 42 ] proposed a three-stage method for diagnosing COVID-19, first using a fine-tuned ResNet50 model for feature extraction, then extracting features for training for different classifications, and finally classifying normal, COVID-19, and CAP, achieving classification accuracy of 95.76%. Maftouni et al [ 43 ] used DenseNet-121 and Residual Attention-92 models to detect CAP and COVID-19 in CT images, pre-trained on the ImageNet and CIFAR-10 datasets, respectively, and finally used a fully connected layer and SVM to output the classification results, achieving an accuracy rate of 95.31%. Verma et al [ 20 ] developed an artificial intelligence (AI)-driven android application for detecting COVID-19 from CT images.…”

On the differences between CNNs and vision transformers for COVID-19 diagnosis using CT and chest x-ray mono- and multimodality

“…Figures 1-3 showcase samples from the three datasets listed for each respective class. Curated COVID-CT (Maftouni et al, 2021), or COVID19 C, is a large lung CT dataset curated from seven different publicly available datasets. It comprises images of 7,593 COVID, 6,893 normal and 2,618 community-acquired pneumonia cases.…”