Experience in Using Neural Networks to Predict the Outcomes of Ischemic Stroke: A Literature Review

Abstract

Introduction. Ischemic stroke is a structurally complex disease based on various pathogenetic mechanisms. In view of the complexity of this pathology and its structure, the medical community has established various assessment scales based on different signs. The scales were created in order to predict possible conditions of a patient at various stages of treatment. The objective of our research was to determine the relevance of applying the system of predicting outcomes of ischemic stroke based on neural networks to improve ischemic stroke treatment and management. Materials and methods: We reviewed scientific and medical literature devoted to the development and use of forecasting systems based on artificial neural networks to predict outcomes of ischemic stroke and analyzed the most common assessment scales currently used in therapeutic practices. Results. The analysis of effectiveness of available scales revealed that their main drawback was a subjective component in the assessment of a patient’s condition. The use of neural networks, in its turn, minimizes the subjective component in predicting the outcome of ischemic stroke since neural networks are capable of processing large amounts of data and can, therefore, establish implicit correlation between research objects. Conclusion. The analysis of domestic and foreign literary sources proves that the presence of a forecasting system based on a neural network is a major advantage for a health care facility. Yet, neural networks have not fully passed clinical trials that would confirm their superiority over current methods of predicting disease outcomes, which impedes their extensive use in clinical practice.