Artificial-Intelligence-Driven Algorithms for Predicting Response to Corticosteroid Treatment in Patients with Post-Acute COVID-19

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.

Corticosteroid treatment prediction using chest X-ray and clinical data

“…Compared to our previous work [35] , this approach has improved the capability of CS treatment prediction for more than 6% accuracy. Here, it can be concluded that additional information, such as CXR is useful in CS treatment prediction.…”

“…The metrics used are accuracy, precision, recall, balanced accuracy, and F1-score, which are usually used in evaluating classification tasks. The definitions of the mentioned metrics are shown below [35] : where TN – number of True Negative cases, TP – True Positives, FP – False Positives, FN – False Negatives.…”

“…One of our latest research works [35] is primarily focused on CS treatment prediction, but it is based only on clinical data of patients. This approach compared several ML algorithms, such as Logistic Regression, k-NN, Decision Tree, XGBoost, Random Forest, SVM, MLP, AdaBoost, and LGBM, for the mentioned problem.…”

The effectiveness of glucocorticoid treatment in post-COVID-19 pulmonary involvement

“…One possible approach to improve clinical decision making may be the use of artificial intelligence driven algorithms, as demonstrated by an affiliated collective of Myska et al [ 34 ], who have used the data from Olomouc post-COVID dataset to test the performance of 9 different machine learning algorithms in predicting improvement in the patients indicated for CS therapy. The best results were achieved by the Decision Tree approach, reaching 73.52% balanced accuracy on a validation portion of the dataset.…”

Liability of Health Professionals Using Sensors, Telemedicine and Artificial Intelligence for Remote Healthcare

“…As far as possible, when sensors and/or AI are used, there should be a document precisely defining the responsibility of each participant. For example, remote telemonitoring has been widely used during the COVID-19 pandemic in order to obtain early hospital discharge, allowing beds to be saved for other patients, and it will again be used to follow the numerous patients suffering from long COVID [ 11 , 67 , 68 , 69 ]. This is a simple use case, but doctors would have been responsible if one parameter (blood saturation decrease related to worsening of pneumopathy secondary to virus spreading) were not to be detected or not transmitted early enough to the doctor.…”