Accurate estimation of stroke risk with fuzzy clustering and ensemble learning methods

Objectives: Stroke is a serious health condition that is among the leading causes of death and permanent disability worldwide. Despite this, a generally valid, effective treatment method has not been found in the struggle against stroke, leaving preventive treatment as the most viable option. Due to its acute nature, it is practically impossible to conduct pre-symptomatic diagnosis of stroke events while they occur. For this reason, determining the patients with high stroke risk is considered as the first step towards taking necessary precautions, which is the main objective of this study. Materials and methods: In this study, a novel model consisting of a fuzzy clustering stage followed by an ensemble learning based estimation stage is proposed for stroke risk estimation. The proposed model is trained and tested using a novel dataset with 124 features and 3000 patients, formed retrospectively within the context of this study. Stroke events are handled in 4 types and risk of each stroke type is estimated separately. Results: The accuracy of the model in predicting general stroke, ischemic stroke, hemorrhagic stroke and tran-sient ischemic attack events is obtained as 99.85%, 99.89%, 99.77% and 99.24%; respectively. Conclusions: The proposed model showed almost perfect performance. In this context, it has the ability to perform risk estimation accurately for 4 different types of stroke while considering a large set of risk factors. Therefore, it possesses the potential to aid the medical experts by functioning as a decision support system.

Automated summary

This study proposes a novel model for stroke risk estimation. The model shows excellent accuracy and is able to accurately estimate the risk of different types of stroke.