Convolutional Neural Networks: A Simple and Functional Approach for COVID-19 Severity Prediction

Abstract

Background and Aims: The Coronavirus disease 2019 (COVID-19) pandemic began in 2020. A major problem during COVID-19 was determining the clinical severity. There are a variety of markers for assessing the COVID-19 severity and outcome. So, this study aims to introduce a new approach for determining the disease severity based on the laboratory data obtained by machine learning algorithms.
 Materials and Methods: In this study, we used 100 patients for modeling. We used demographical, background disease, and laboratory data of COVID-19 patients as parameters for training the convolutional neural network model to evaluate disease severity and tried to create a predictive algorithm for future data. The sequential neural network from the Keras library by TensorFlow was used for prediction. The clinical validation of prediction by model was evaluated by the receiver operating characteristic (ROC) curve.
 Results: The mean F1 score for our current model was 0.62 (in the range of 0-1). The F1 scores for the severe group and the mild group were 0.8 and 0.45, respectively. The ROC curve for clinical validity revealed an acceptable Area Under Curve (0.085) for both severe and mild categories.
 Conclusion: The current study introduces a simple machine learning algorithm as tool for determining COVID-19 severity of by acceptable ROC. This study can lead us to use such algorithms more often in laboratory medicine and clinical decision-making. Furthermore, the present study is just a preliminary study and highlights the need for further research to validate and refine the proposed model.