Efficient and visualizable convolutional neural networks for COVID-19 classification using Chest CT

Abstract

With COVID-19 cases rising rapidly, deep learning has emerged as a promising diagnosis technique. However, identifying the most accurate models to characterize COVID-19 patients is challenging because comparing results obtained with different types of data and acquisition processes is non-trivial. In this paper we designed, evaluated, and compared the performance of 20 convolutional neutral networks in classifying patients as COVID-19 positive, healthy, or suffering from other pulmonary lung infections based on Chest CT scans, serving as the first to consider the EfficientNet family for COVID-19 diagnosis and employ intermediate activation maps for visualizing model performance. All models are trained and evaluated in Python using 4173 Chest CT images from the dataset entitled “A COVID multiclass dataset of CT scans,” with 2168, 758, and 1247 images of patients that are COVID-19 positive, healthy, or suffering from other pulmonary infections, respectively. EfficientNet-B5 was identified as the best model with an F1 score of 0.9769±0.0046, accuracy of 0.9759±0.0048, sensitivity of 0.9788±0.0055, specificity of 0.9730±0.0057, and precision of 0.9751± 0.0051. On an alternate 2-class dataset, EfficientNetB5 obtained an accuracy of 0.9845±0.0109, F1 score of 0.9599±0.0251, sensitivity of 0.9682±0.0099, specificity of 0.9883±0.0150, and precision of 0.9526 ± 0.0523. Intermediate activation maps and Gradient-weighted Class Activation Mappings offered human-interpretable evidence of the model’s perception of ground-class opacities and consolidations, hinting towards a promising use-case of artificial intelligence-assisted radiology tools. With a prediction speed of under 0.1 seconds on GPUs and 0.5 seconds on CPUs, our proposed model offers a rapid, scalable, and accurate diagnostic for COVID-19.

Deep feature fusion classification network (DFFCNet): Towards accurate diagnosis of COVID-19 using chest X-rays images

“…To demonstrate the effectiveness of the DFFCNet method, we compared it with eight state-of-the-art methods: ECOVNet [43] , Fused-DenseNet-Tiny [23] , BCNN_SVM [44] , COVNet [45] , InceptionV3 [46] , DTL-V19 [47] , ResNet152V2 [48] , and VGG16 [49] . All methods used the unified dataset and MDA preprocessing methods, experiments were performed on the testing set.…”

“…ECOVNet [43] , Fused-DenseNet-Tiny [23] , COVNet [45] and DTL-V19 [47] are the proposed methods for COVID-19 diseases. BCNN_SVM [44] , InceptionV3 [46] , ResNet152V2 [48] and VGG16 [49] are better classification networks proposed in recent years, and these methods are very representative.…”

“…However, due to the limited network performance, the network had fewer layers and was prone to overfitting. ECOVNet [43] and COVNet [45] were trained for classification using EfficientNetB3 and ResNet50, respectively. These models were simple, lack the fusion of features, and their performance was inferior to our proposed network.…”

DAFLNet: Dual Asymmetric Feature Learning Network for COVID-19 Disease Diagnosis in X-Rays

“…To demonstrate the proposed method's contribution, the proposed DAFLNet method was compared with 8 state-of-the-art approaches: ECOVNet-EfficientNetB3 base [ 28 ], BCNN_SVM [ 29 ], COVNet [ 30 ], DTL-V19 [ 9 ], DenseNet121 [ 31 ], Resnet50V2 [ 32 ], Xception [ 33 ], and MobileNetV2 [ 34 ]. The same dataset with all methods and the results shown are the mean and standard deviation of 5 runs.…”

COVID-19 contagion forecasting framework based on curve decomposition and evolutionary artificial neural networks: A case study in Andalusia, Spain

“…In Tamal et al (2021) , a set of radiomics features from CXR images were selected and used to train three classical ML classifiers, providing an accurate, fast and automatic method that can be integrated with standard X-ray reporting systems. Most recently, in Garg, Salehi, Rocca, Garner, and Duncan (2022) , the performances of 20 different CNNs trained for classifying patients into three and two classes using chest CT images achieved an accurate and very efficient classification model.…”