ABSTRACT
Background and Objective: Severe courses of COVID-19 disease can lead to long-term complications. The post-acute phase of COVID-19 refers to the persistent or new symptoms. This problem is becoming more relevant with the increasing number of patients who have contracted COVID-19 and the emergence of new virus variants. In this case, preventive treatment with corticosteroids can be applied. However, not everyone benefits from the treatment, moreover, it can have severe side effects. Currently, no study would analyze who benefits from the treatment. Methods: This work introduces a novel approach to the recommendation of Corticosteroid (CS) treatment for patients in the post-acute phase. We have used a novel combination of clinical data, including blood tests, spirometry, and X-ray images from 273 patients. These are very challenging to collect, especially from patients in the post-acute phase of COVID-19. To our knowledge, no similar dataset exists in the literature. Moreover, we have proposed a unique methodology that combines machine learning and deep learning models based on Vision Transformer (ViT) and InceptionNet, preprocessing techniques, and pretraining strategies to deal with the specific characteristics of our data. Results: The experiments have proved that combining clinical data with CXR images achieves 8% higher accuracy than independent analysis of CXR images. The proposed method reached 80.0% accuracy (78.7% balanced accuracy) and a ROC-AUC of 0.89. Conclusions: The introduced system for CS treatment prediction using our neural network and learning algorithm is unique in this field of research. Here, we have shown the efficiency of using mixed data and proved it on real-world data. The paper also introduces the factors that could be used to predict long-term complications. Additionally, this system was deployed to the hospital environment as a recommendation tool, which admits the clinical application of the proposed methodology.
ABSTRACT
Pulmonary fibrosis is one of the most severe long-term consequences of COVID-19. Corticosteroid treatment increases the chances of recovery; unfortunately, it can also have side effects. Therefore, we aimed to develop prediction models for a personalized selection of patients benefiting from corticotherapy. The experiment utilized various algorithms, including Logistic Regression, k-NN, Decision Tree, XGBoost, Random Forest, SVM, MLP, AdaBoost, and LGBM. In addition easily human-interpretable model is presented. All algorithms were trained on a dataset consisting of a total of 281 patients. Every patient conducted an examination at the start and three months after the post-COVID treatment. The examination comprised a physical examination, blood tests, functional lung tests, and an assessment of health state based on X-ray and HRCT. The Decision tree algorithm achieved balanced accuracy (BA) of 73.52%, ROC-AUC of 74.69%, and 71.70% F1 score. Other algorithms achieving high accuracy included Random Forest (BA 70.00%, ROC-AUC 70.62%, 67.92% F1 score) and AdaBoost (BA 70.37%, ROC-AUC 63.58%, 70.18% F1 score). The experiments prove that information obtained during the initiation of the post-COVID-19 treatment can be used to predict whether the patient will benefit from corticotherapy. The presented predictive models can be used by clinicians to make personalized treatment decisions.
