One of the promising methods for early detection of Coronavirus Disease 2019 (COVID-19) among symptomatic patients is to analyze chest Computed Tomography (CT) scans or chest x-rays images of individuals using Deep Learning (DL) techniques. This paper proposes a novel stacked ensemble to detect COVID-19 either from chest CT scans or chest x-ray images of an individual. The proposed model is a stacked ensemble of heterogenous pre-trained computer vision models. Four pre-trained DL models were considered: Visual Geometry Group (VGG 19), Residual Network (ResNet 101), Densely Connected Convolutional Networks (DenseNet 169) and Wide Residual Network (WideResNet 50 2). From each pre-trained model, the potential candidates for base classifiers were obtained by varying the number of additional fully-connected layers. After an exhaustive search, three best-performing diverse models were selected to design a weighted average-based heterogeneous stacked ensemble. Five different chest CT scans and chest x-ray images were used to train and evaluate the proposed model. The performance of the proposed model was compared with two other ensemble models, baseline pre-trained computer vision models and existing models for COVID-19 detection. The proposed model achieved uniformly good performance on five different datasets, consisting of chest CT scans and chest x-rays images. In relevance to COVID-19, as the recall is more important than precision, the trade-offs between recall and precision at different thresholds were explored. Recommended threshold values which yielded a high recall and accuracy were obtained for each dataset.

To accurately diagnose multiple lung diseases from chest X-rays, the critical aspect is to identify lung diseases with high sensitivity and specificity. This study proposed a novel multi-class classification framework that minimises either false positives or false negatives that is useful in computer aided diagnosis or computer aided detection respectively. To minimise false positives or false negatives, we generated respective stacked ensemble from pre-trained models and fully connected layers using selection metric and systematic method. The diversity of base classifiers was based on diverse set of false positives or false negatives generated. The proposed multi-class framework was evaluated on two chest X-ray datasets, and the performance was compared with the existing models and base classifiers. Moreover, we used LIME (Local Interpretable Model-agnostic Explanations) to locate the regions focused by the multi-class classification framework.

No abstract