An improved machine learning-based prediction framework for early detection of events in heart failure patients using mHealth

Acute decompensated heart failure (ADHF) is the most prevalent cause of acute respiratory distress worldwide, accounting for the majority of new cases and associated fatalities according to global statistics, making it a serious public health issue at present. The prognosis and probability of survival can be greatly enhanced by an early diagnosis of ADHF since it encourages patients to receive prompt clinical care. Over the past decade, there have been notable advancements in the use of artificial intelligence to interpret cardiovascular data from echocardiograms, and cardiac magnetic resonance imaging to ascertain hazard manifestations or future risks of cardiovascular disorders.

In this paper, a model is devised to forecast events in outpatients with heart failure. This analysis utilized ten classification models to estimate the patient's prognosis. The order of the importance of the features is determined based on the recursive feature elimination technique by using all of the aforementioned approaches as a base model. The top five features such as gender, diabetes mellitus, systolic blood pressure, sodium, and heart rate were selected from the average ranking of the classifier.

The experimental results demonstrate that the Gaussian Naive Bayes classifier is superior to other models on an average performance basis, whereas K-Nearest Neighbor provided the best results over other classifiers with precision, recall, and F1 scores of 0.98, 0.91, and 0.94, respectively.

Finally, a web-based mHealth application is built to estimate the probability of heart failure depending on the accuracy level.