Identification of Myocardial Infarction (MI) Probability from Imbalanced Medical Survey Data: An Artificial Neural Network (ANN) with Explainable AI (XAI) Insights

Abstract

In the healthcare industry, many artificial intelligence (AI) models have attempted to overcome bias from class imbalances while also maintaining high results. Firstly, when utilizing a large number of unbalanced samples, current AI models and related research have failed to balance specificity and sensitivity – a problem that can undermine the reliability of medical research. Secondly, no reliable method for obtaining detailed interpretability has been put forth when addressing large numbers of input features. The present research addresses these two critical research gaps with a proposed lightweight Artificial Neural Network (ANN) model. Using 43 input features from the 2021 Behavioral Risk Factor Surveillance System (BRFSS) dataset, the proposed model outperforms prior models in producing balanced outcomes from markedly unbalanced large survey data. The efficacy of this proposed ANN model is attributed to its simplified design, which reduces processing demands, and its resilience in identifying the probability of myocardial infarction (MI). This is demonstrated by its 80% specificity and 77% sensitivity, and is substantiated by a Receiver Operating Characteristic Area Under the Curve (AUC) of 0.87. The outcomes across the scopes of each specified data domain were also separately represented, thus demonstrating the proposed model’s robust sensitivity. The interpretability of the model, as measured by Shapley values, reveals substantial correlations between myocardial infarction (MI) and its risk factors, including long-term medical conditions, socio-demographic factors, personal health habits, economic and social status, healthcare availability and affordability, as well as impairment statuses, providing valuable insights for improved cardiovascular risk assessment and personalized healthcare strategies.