Classification of COVID-19 and Nodule in CT Images using Deep Convolutional Neural Network

Distinguishing between coronavirus disease 2019 (COVID-19) and nodule as an early indicator of lung cancer in Computed Tomography (CT) images has been a challenge that radiologists have faced sinceCOVID-19 was announced as a pandemic. The similarity between these two infections is the main reason that brings dilemmas for them and may lead to a misdiagnosis. As a result, manual classification is not as efficient as automated classification. This paper proposes an automated approach to classify COVID-19 infections from nodules in CT images. Convolutional Neural Networks (CNNs) have significantly meliorated automated image classification tasks, particularly for medical images. Accordingly, we propose a refined CNN-based architecture through modifications in the network layers to reduce complexity. Furthermore, to vanquish the lack of training data, data augmentation approaches are utilized. In our method, Multi Layer Perceptron (MLP) is obligated to categorize the feature vectors extracted from denoised input images by convolutional layers into two main classes of COVID-19 infections and nodules. To the best of our knowledge, other state-of-the-art methods can only classify one of the two classes listed above. Compared to the mentioned counterparts, our proposed method has a promising performance with an accuracy of 97.80%.