Ischemic Stroke Post-treatment Prediction using Machine Learning to develop Web Application for Healthcare Centers in Thailand

Stroke attack is one of the most serious concerns in an emergency room. Expert diagnosis and accurate treatment are crucial for the improvement, death, or impairment of the patients. This work aims at using big data analytics for developing a system to provide the probability for each patient whether he would improve after medications. Records of N=1700 volunteers were gathered for non-linear classification training. Random Forest outperforms seven other methods compared in this study. The AI model was deployed as a web application for a rural hospital to assist medical personnel. This can be considered an important step towards AI-assisted emergency medications for stroke. Under the situation where the NIH Stroke Score was unavailable on time, the system could still provide the improvement probability via a web application based on the AI model trained in this work.

I. Introduction

Since 2017, cerebrovascular disease has remained the leading cause of death and disability around the globe. The abnormalities may occur in the central nervous system. Blood vessels in the brain and includes brain regions such as subarachnoid hemorrhage (SAH) and intracerebral hemorrhage (ICH). A million people worldwide were diagnosed with cerebrovascular disease. On average, every 6 seconds, a person dies due to a stroke worldwide. Furthermore, there is an increasing trend in these deaths [1]. In 2021, 12.2 million people worldwide died as a result of a stroke attack. It has been reported in the past that adults aged 25 and older had a 1/6 chance of developing cerebrovascular disease. The situation nowadays is getting worse, the chance of having a stroke has increased to 1/4 in the adult population with an age >= 25. The statistics of death and stroke incidence are increasing in the same direction. The exact causes and factors that promote stroke are inconclusive [2].